Bolds are best reads, *Bolds are best of the best.

Soviet / Russian History & the Cold War

*To the Success of Our Hopeless Cause: The Many Lives of the Soviet Dissident Movement, Benjamin Nathans, 2024. Covered in an upcoming podcast. A masterful book. For a taste:

Samizdat provided not just new things to read, but new modes of reading. There was binge reading: staying up all night pouring through a sheath of onion-skin papers because you'd been given twenty-four hours to consume a novel that Volodia was expecting the next day, and because, quite apart from Volodia's expectations, you didn't want that particular novel in your apartment for any longer than necessary. There was slow-motion reading: for the privilege of access to a samizdat text, you might be obliged to return not just the original but multiple copies to the lender. This meant reading while simultaneously pounding out a fresh version of the text on a typewriter, as a thick raft of onion-skin sheets alternating with carbon paper slowly wound its way around the platen, line by line, three, six, or as many as twelve deep. "Your shoulders would hurt like a lumberjack's," recalled one typist.

Experienced samizdat readers claimed to be able to tell how many layers had been between any given sheet and the typewriter's ink ribbon. There was group reading: for texts whose supply could not keep up with demand, friends would gather and form an assembly line around the kitchen table, passing each successive page from reader to reader, something impossible to do with a book. And there was site-specific reading: certain texts were simply too valuable, too fragile, or too dangerous to be lent out. To read Trotsky, you went to this person's apartment; to read Orwell, to that person's.

However and wherever it was read, samizdat delivered the added frisson of the forbidden. Its shabby appearance—frayed edges, wrinkles, ink smudges, and traces of human sweat—only accentuated its authenticity. Samizdat turned reading into an act of transgression. Having liberated themselves from the Aesopian language of writers who continued to struggle with internal and external censors, samizdat readers could imagine themselves belonging to the world's edgiest and most secretive book club. Who were the other members, and who had held the very same onion-skin sheets that you were now holding? How many retypings separated you from the author?

*The Gulag Archipelago, Vol. 1: An Experiment in Literary Investigation, Aleksandr Solzhenitsyn, 1973. Survivor and stylist. Energy and humanity sparks though the prose.

Look around you—there are people around you. Maybe you will remember one of them all your life and later eat your heart out because you didn't make use of the opportunity to ask him questions. And the less you talk, the more you'll hear. Thin strands of human lives stretch from island to island of the Archipelago. They intertwine, touch one another for one night only in just such a clickety-clacking half-dark car as this and then separate once and for all. Put your ear to their quiet humming and the steady clickety-clack beneath the car. After all, it is the spinning wheel of life that is clicking and clacking away there.

*To Run the World: The Kremlin’s Cold War Bid for Global Power, Sergey Radchenko, 2024. Covered in a two-parter.

Unwanted Visionaries: The Soviet Failure in Asia at the End of the Cold War, Sergey Radchenko, 2014.

Apple in China. Patrick McGee, 2025. Covered in a pod here.

China & East Asia

*The Party’s Interests Come First: The Life of Xi Zhongxun, Father of Xi Jinping, Joseph Torigian, 2025. A China book at the level we get only a few times a decade. A true must read for anyone hoping to deeply understand the CCP and Chinese 20th century history. Covered in a podcast here.

A Dragon’s Head and a Serpent’s Tail: Ming China and the First Great East Asian War, 1592-1598, Kenneth M. Swope, 2009. Inspired by the Yasheng Huang shows.

Hideyoshi, man. A letter he sends to the Koreans before he invades them…

Whenever and against whomever I have waged war, the victory has always been mine. The lands and districts invaded by me have always been conquered. Now our empire has entered upon a period of peace and prosperity, and the people are enjoying a benevolent rule. Lonely old men and forlorn widows are all well provided for. Both the national wealth and that of individuals has been so greatly augmented that it is unparalleled in our history. Since the nation’s founding, our empire has never before witnessed such glory as that of our imperial court and such splendor as that of our imperial capital.

However, human life in this world is brief. . . . I am not willing to spend the remaining years of my life in the land of my birth. According to my idea, the empire that I would create should not be separated by mountains and seas, but should include them all. In starting my conquest, I plan that our forces should proceed to the country of the Great Ming and compel the people there to adopt our customs and manners. Then that vast country, consisting of more than four hundred provinces, would enjoy our imperial protection and benevolence for millions of years to come. I have in mind a plan of conquest which shall surely be carried to a successful ending. Your kingdom has taken the lead among the continental states by sending an envoy to our court, thus showing reverence to our throne. You have acted in accordance with the wise saying of the ancients that one who has foresight and is humble and cautious will always be free from grief and worry. . . . You, King of Korea, are hereby instructed to join us when we proceed to [the country of the] Great Ming at the head of all your fighting men. You may thereby further renew your pledge of service due to us as a neighboring state. Our sole desire is to have our glorious name revered in the three countries [of China, Korea, and Japan].

Swope writes: “while some of Hideyoshi’s statements to foreign rulers are overbearing to the point of laughter, it seems likely that he, like so many great conquerors, truly believed in his destiny and ability to overcome all odds.”

People used to be so petty too!

Hideyoshi at that time was on campaign, so the envoys had to await his return. But this was not the first indignity they had suffered. When they first got to Tsushima, the Koreans were treated discourteously in their eyes, for Sō Yoshitoshi’s retainers had brought him into a banquet hall in a palanquin, then Sō proceeded to sit in an elevated position. This infuriated Kim Sŏngil, who stormed out, exclaiming, “These barbarians really have no sense of propriety and cannot distinguish between higher and lower officials.

Under the Nuclear Shadow: China’s Information-Age Weapons in International Security, Fiona S. Cunningham, 2025. Covered in a pod here.

Beyond Power Transitions: The Lessons of East Asian History and the Future of U.S.-China Relations, David Kang and Xinru Ma. Covered in a pod here.

House of Huawei: The Secret History of China’s Most Powerful Company, Eva Dou, 2025. Covered in a pod here.

Israel & Jewish Studies

Rebbe: The Life and Teachings of Menachem M. Schneerson, the Most Influential Rabbi in Modern History, Joseph Telushkin, 2014.

Fascinating less for any particular advice he gave people or viewpoints about Judaism or politics and more rather how and why people held him in such esteem. We already have something close to a Rebbe we can consult 24/7 in our pockets but no-one feels like asking ChatGPT a personal question deserves the weight and reverence that a private meeting with or letter from the Rebbe does.

Why? Meetings were scheduled months in advance and all after dark, with many occuring after midnight following a multi-hour long build up as you wait in line. “You saw the Rebbe when the rest of the world was asleep.” All who went through this ordeal came primed with the understanding that the Rebbe has access to deep spiritual knowledge and decades of learning of a religion you hold in high regard.

When you finally get your meeting, for the time you’re together he interacts with you as if you’re the only other person in the world. “The Rebbe’s most striking physical characteristic, the one most often commented upon, particularly by those who met him in yechidus, were his eyes; in Swados’s case, he recalled how ‘the pale blue eyes remain fixed upon [me] with an unblinking directness that could be disconcerting.’”

Clinton could do this too: Game recognize game, from Obama in 2009: “Now, I think everyone knows what it’s like when Bill Clinton asks you to make a commitment. He looks you in the eye … he makes you feel like you’re the only person in the room.”.

McChrystal is the only person I’ve sat in the same room with whose gaze made me feel this way. Six weeks after getting fired in Afghanistan he’s teaching a seminar at Yale and sitting down with me to talk about some sophomore about a mediocre paper idea. I don’t think this is a thing people can fake—this breed is that curious and able to live in the moment.

The fact that someone important took the time to care about your situation means something that you can’t get from Scarlett Johannsen in Her.

The case involved a seventeen-year-old girl who was having extreme emotional difficulties, involving issues of rebellion and serious doubts about faith and religious practice—‘the typical kinds of things that you find among teenagers, but in this case quite severe and for some reason.’ The young woman herself would often write to the Rebbe and then share and review with the young rabbi the responses she received. Kaplan was amazed at the speed with which the Rebbe responded to the girl’s letters—sometimes within hours, never longer than a day. One exchange of letters remains vivid to Kaplan: the young woman had written an extended letter describing the continuing emotional turmoil and constant anguish she was experiencing, and the Rebbe, in turn, responded that he felt her pain. Kaplan recalls the impact of these words: the girl felt that there was a person who actually did feel what she was going through and truly wanted to help her; from that point on, her life slowly started to turn around. To this day, Kaplan remains staggered that the Rebbe—a man known to receive hundreds of letters a day—placed himself in this situation with such intensity, ‘answering the girl’s letters back and forth…and doing so within a day, and sometimes within hours.’

Like , the Rebbe “was known to personally open all letters addressed to him.”

Productivity hacks brought to you by the Rebbe.

His father-in-law taught him the secret of “success with time,” a technique that the Frierdiker Rebbe had learned from his father, the Rebbe Rashab: The first prerequisite for fulfilling one’s responsibilities is to fully grasp that a person can never add to the amount of time in the day or the night. Since time is finite, the only way we can carry out all that we need to do is to utilize whatever time we do have to its full capacity; this means giving our entire focus—our full concentration—to whatever we are doing at that moment.

Therefore, while working on one task, “we must regard anything else we have done before and anything that we are planning to do later as totally insignificant.”

This was a good line. Or maybe I just don’t consume enough chicken soup.

Rabbi Zev Segal, a community activist, encountered a similar, though subtly different, response. After a long career in the rabbinate, Segal went to work for the Memorial Foundation for Jewish Culture. Once, as he was getting ready to depart on an overseas mission, he received a call from Rabbi Hodakov, with a request from the Rebbe that he fulfill a certain assignment. Segal agreed to do so, although he eventually found the task requested by the Rebbe to be much more difficult than he had anticipated, and even a bit dangerous.

When he returned, he said, “I came back and gave the Rebbe a report and I concluded that the Rebbe should know that this was not an easy task for me. It was, rather, very difficult. The Rebbe looked at me quizzically and said, ‘Rabbi Segal, since when did you make a contract with the Almighty for an easy life?’

At a speech delivered after the Rebbe’s death, Segal explained how this one, seemingly throw-away line, uttered in fewer than ten seconds, permanently affected him. Even though a task will not be easy, each of us must do what we know we were put on earth to do.

“If the Nazis had murdered Jews with Zyklon B gas because it was cheap, a few pennies for each person they suffocated, the Rebbe wanted to elevate the value of human life…I recall him saying something to the effect that even capturing the Kotel ha‑Ma‘aravi (Western Wall) was not worth losing one innocent Israeli soldier’s life.”

And I found this quirk endearing.

For the Rebbe, the desire to choose positive words was so deeply ingrained that he hesitated to use terms like “evil,” even when describing something that was…The Rebbe’s search for non-negative language went well beyond anything traditional Jewish texts might have intended. He apparently believed that words with bad connotations could trigger harmful associations even in the most innocent contexts. He avoided the word “undertake,” lest it evoke “undertaker.” And, no matter how great the pressure to finish a project, he never referred to the due date as a “deadline.” Once “deadline” is removed from one’s vocabulary, a natural alternative is “due date,” with “deadline” connoting death and “due date” connoting birth.

*Eichmann in Jerusalem: A Report on the Banality of Evil, Hannah Arendt, 1963. I would have read this years ago if anyone told me how funny it was! As she wrote to a friend, “You are the only reader to understand what otherwise I have never admitted—namely that I wrote Eichimann in Jerusalem in a curious state of euphoria.” It shows and is so much better for it.

World-historically funny.

Extended excerpts here in my Tel Aviv writeup.

I paired this with Radio Treason: The Trials of Lord Haw-Haw, the British Voice of Nazi Germany, Rebecca West, 2025.

Rise and Kill First: The Secret History of Israel's Targeted Assassinations, Ronen Bergman, 2018. Love how sourced this book is, mostly by people who did cool secret shit for years, retired, and now want to tell the world how cool the secret shit they did was. His conclusion that the incredible tactical successes have helped cover up for massive strategic failues seems a little premature in light of Israel eliminating Hezbollah and tbd on Iran…though if Iran ends up with the bomb in two years the thesis will have been all too true. Deserves a new edition!

World War II

Eastern Front

*The Stahel Arc

Operation Barbarossa and Germany’s Defeat in the East, David Stahel, 2009
Operation Typhoon: Hitler’s March on Moscow, October 1941, David Stahel, 2013
Kiev 1941: Hitler’s Battle for Supremacy in the East, David Stahel, 2012
The Battle for Moscow, David Stahel, 2015
Retreat from Moscow: A New History of Germany’s Winter Campaign, 1941-1942, David Stahel, 2019
Hitler’s Panzer Generals: Guderian, Hoepner, Reinhardt and Schmidt Unguarded, David Stahel, 2023

Operational history does not get more grounded and gripping than this. The enormity of the Eastern Front from a manpower, space, and materiel perspective even after consuming all these books is still too much for my brain to get around. He also weaves in enough human color to let you start to develop a feel for the experience of the soldiers and officers.

Once Moscow survived the initial push in 1941, which Stahel convincingly argues wasn’t really even all that close-run a thing due to the Nazis burning their mechanized army and didn’t really have any contingency plan…

Industrial weight and, as Williamson Murray called it, the Nazis “strategic myopia” completely washes out always beats out tactical and operational creativity if you’re committed to staying in the fight over a long enough time horizon.

In November 1941 Stalin confidently exclaimed: ‘Modern war is a war of motors. The war will be won by the one who produces the most motors. The combined motor production of the USA, Britain, and the USSR is at least three times that of Germany.’ In fact the combined motor production of the three Allied powers was far in excess of Stalin’s three-fold estimate. The remaining years of the war continued to see a commanding Allied lead in armament production, dooming Germany to eventual defeat by sheer weight of arms. The ebb and flow of bat­tlefield successes affected only the length of the war, not its eventual outcome.

Oh, and it’s never a good look when you feel like you had it worse than a Napoleonic soldier.

Brilliant page of analysis here.

A death cult means living soldiers aren’t important.

The most nauseating bit of all this military history was reading Heinz Guderian’s love letters.

War is the fucking worst.

The Soviet Union at War 1941-1945, David R. Stone, 2010 + lots of David Stone articles.

The Military History of the Soviet Union, Robin Higham, 2002.

All this I cover in an upcoming podcast with Stahel and David Stone together (that’s still looking for a sponsor! Hit me up!)

Espionage & Intelligence

The Billion Dollar Spy: A True Story of Cold War Espionage and Betrayal, David E. Hoffman, 2015

The Illegals: Russia’s Most Audacious Spies and Their Century-Long Mission to Infiltrate the West, Shaun Walker, 2025.

Military History & Theory

Men Against Fire: The Problem of Battle Command in Future War, S. L. A. Marshall, 1947. The guy has a weird relationship with the truth, but still worth reading for the unique perspective on WWII infantry combat. Lots of aphorisms that sound right?

The most common cause of psychological shock, however, is a partial victory. The adage that "the weakest point fol­lows success" is a fundamental truth of minor tactics and the danger is always greatest when the success is easily won.

Success is disarming. Tension is the normal state of mind and body in combat. When the tension suddenly relaxes through the winning of a first objective, troops are apt to be pervaded by a sense of extreme well-being and there is apt to ensue laxness in all of its forms and with all of its dangers.

The ideal relation­ship between a commander and his subordinate is nowhere better illustrated than in a passage from the letter of instruc­tion wherein Grant told Sherman to proceed to the destruc­tion of Johnston's Army: "I do not propose to lay down for you a plan of campaign: but simply to lay down the work it is desirable to have done and leave you free to execute it in your own way."

Great Captains Unveiled, B. H. Liddell Hart, 1927.

Reputations, 10 Years After, B.H. Liddell Hart, 1928. Hart at 33 years old delivers a hundred pages of scorching takes on the leading WWI generals. France’s Joffre “not a general but a national nerve-sedative”. Von Falkenhayn “The ablest and most scientific general, ‘penny-wise, pound-foolish,’ who ever ruined his country.” Luttendorf was “the Robot Napoleon.” Hart would have killed it on substack.

A Savage War: A Military History of the Civil War, Williamson Murray and Wayne Wei-Siang Hsieh, 2016. Covered in a podcast to come!

Economics, Technology & Innovation

Abundance, Ezra Klein and Derek Thompson, 2025. Covered in a pod with Dan Wang, Ezra, and Derek.

Learning by Doing: The Real Connection between Innovation, Wages, and Wealth, James Bessen, 2015

Chokepoints: American Power in the Age of Economic Warfare, Edward Fishman, 2025. Covered in a pod.

Traffic: Genius, Rivalry, and Delusion in the Billion-Dollar Race to Go Viral, Ben Smith, 2023. If we thought the race to clicks was brain-warping, just wait until we have AI friends that optimize for engagement and microtransactions.

Kissinger the Negotiator: Lessons from Dealmaking at the Highest Level, James K. Sebenius and Nicholas Burns, 2018. Written almost like a business book, “what lessons can we learn from Kissinger the negotiator,” but if you get over that there’s good diplomatic archival work in here. Covered briefly in a pod here.

Best bits were about the 4D chess he used on Rhodesia to try to get them to change their ways.

Loved this bit about how trilateral leverage unlocked a ton in the US-China relationship.

Dobrynin also was probably more into Kissinger than he should have been. Zhou Enlai would never be caught glazing like this, clearly Kissinger’s superior on the political mindgames front.

The 1920s

Only Yesterday: An Informal History of the 1920s, Frederick Lewis Allen, 1931. did a great job summarizing this book here.

Bubble in the Sun: The Florida Boom of the 1920s and How It Brought on the Great Depression, Christopher Knowlton, 2020. The best part about history books is even if they aren’t great as long as someone has spent real time in the archives and has a decent eye you’re bound to learn things! Good sociological details on really rich people in the 1920s and architecture.

At the very top of the social heap were those who arrived in Florida by private yacht or, better yet, in their own private railcar. During this period, which fell at the end of the Gilded Age, “The near­est thing to a real aristocracy in America was the private-car peerage,” ac­cording to the New Yorker writer Alva Johnston, “and Palm Beach had the world’s most snobbish rail road yard—a Newport-on-wheels, an exalted trailer camp for the gold-encrusted Pullmans of the New World nobility.”

Nothing changes.

Popular resorts, it was commonly thought, had an almost Darwinian evolution: first, they were discovered by writers, artists, and academics; following them came the “nice” millionaires, who didn’t flaunt their wealth and valued their privacy; these were then followed by the “naughty” millionaires, who did flaunt their wealth by building extravagant houses, opening social clubs, throwing grand balls, and, in general, consuming conspicuously. The latter was the moment when an architect on the scene could succeed beyond his wildest imaginings—and Florida was rapidly approaching that stage. In the social historian Cleveland Amory’s view, there was one final step in this evolution, which he simply, and aptly, described as “trouble.”…

Restrictions against buying property or frequenting the hotels often applied to Jews as well as blacks. Anti-Semitism had been on the rise since the arrival in the United States of large numbers of Jewish immi­grants from Eastern Europe starting in the 1880s. Exceptions to the restric­tions were often made for wealthy or prominent “assimilated” Jews who ran investment banks or successful corporations. Fisher’s correspondence in later years is full of letters of exception he wrote for friends such as yeast company president Julius Fleischmann, taxicab magnate John Hertz, and department store mogul Bernard Gimbel, granting them access to his ho­tels or privileges at his clubs. Fisher liked to think of himself as a fair per­son, staunchly unbiased, but for business purposes, he was quite willing to share in the prevailing bigotry of the day. It wasn’t until 1948 that the US Supreme Court eventually ruled such race covenants unenforceable. The Fair Housing Act of 1968 would ban the practice altogether.

Cool kids got divorced in Paris.

Elizabeth left for Europe around September 5 on the liner Homeric, accompanied by her best friend, presumably to vacation for a month and then to establish domicile in Paris for a quickie Paris divorce—a favored method for the wealthy to part ways during those years. As the lawyer Frank Shutts explained to Carl Fisher, whose marriage was in similar trou­ble, “the French court will grant divorce decrees up on the general proposi­tion that two people do not want to live together anymore.”

Retvrn actually looks like advertising “social supremacy” in your real estate brochures.

‘Social supremacy and fi­nancial solidarity are assured to Boca Raton by the wealth and standing, the character and achievement of its proprietary sponsors. These men and women of world-standing in society, finance, and affairs know that the best of democracy is the flower of genuine aristocracy. They therefore invite men and women of substance and standing to participate in their unique undertaking.’

You used to be able to get sued for ‘breach of promise’ if a guy renegged on marrying you.

They built ‘foreign villages’ in Coral Gables just like Huawei did with its campus. Here’s a photo of the Chinese one refurbished today:

Bruce Catton on the 50’s, good line but not surprising from a Civil War historian: “All the old rules seemed to be vanishing in the twenties. In exchange came a strange new world both gaudy and sad.”

Ex-Wife, Ursula Parrott, 1929. Vivid and gripping proto-Nora Ephron. You really have to wait until the 1920s to read prose that feels completely modern, and this alongside some early Fitzgerald passes the threshold for me.

Grab Bag

They Flew: A History of the Impossible, Carlos M. N. Eire, 2023. Fun read of a careful and writerly historian who takes seriously claims of levitation in the late Middle Ages.

Every age and culture has its own unquestionable beliefs, and our own tends to prize the rationality and superiority of unbelief as one of its core beliefs, especially in regard to denying the existence of a supernatural dimension. Such unquestionable pervasive beliefs—Troelstch’s “social facts”—which William Blake called “mind-forged manacles” in 1794 and Max Weber spoke of as the “steel-hard casing” or an “iron cage” a century later, are difficult to detect and acknowledge, for they frame our thinking and are very much like the air we breathe, which we take for granted as much as an octopus takes water for granted. And even when perceived for what they are—as difficult as that is to do—these manacles and cages are even harder to discard or annihilate.

Monasticism itself is deeply rooted in holy foolishness because of its rejection of worldly values, which can make every monk or nun seem insane, or even a raving lunatic. The Cistercian luminary Saint Bernard of Clairvaux (1090–1153), who greatly admired Antony the Great, captured the essence of monastic foolishness brilliantly in one of his letters:

For what else do worldlings think we are doing but playing about, when what they desire most on earth, we flee, and what they flee, we desire? We are like jesters and tumblers, who, with heads down and feet up, exhibit extraordinary behaviour by standing or walking on their hands, and thus draw all eyes to themselves. But ours is not the play of children or of the theatre, which excites lust and represents sordid acts in the effeminate and shameful contortions of the actors. No, ours is a joyous game, decent, grave, and admirable, delighting the gaze of the heavenly onlookers.

Strong writing here:

Discerning the difference between the natural and the supernatural or between genuine and fraudulent miracle claims can become immensely difficult, if not impossible, for the clerical elites in charge of ensuring the purity of the faith, as well as for the laity, including crowned monarchs. Inevitably, given the devil’s reputation as the ultimate trickster—the fact that he was “a liar and the father of lies,”1 ever eager to cause trouble—he, too, could more easily wheedle his way into the picture.

Such was the dilemma faced by early modern Catholicism when miracles became a highly valued feature of Catholic identity as well as a polemical weapon to wield against Protestants and skeptics of all stripes. It was a vexing conundrum, and a painful one, for it required doubting, and doubt always rubs faith raw. Sorting the genuinely divine from the fraudulent or demonic was an ordeal that also required intellectual, emotional, psychological, and spiritual fortitude on the part of all involved in the process. Teresa of Avila, Joseph of Cupertino, and María de Ágreda were liminal avatars of the impossible, suspended between the divine and demonic, their sanctity revered and questioned simultaneously, perfectly poised to play the role of tricksters acting as agents of the devil, the ultimate trickster. All three bore the brunt of doubt and survived their ordeal. But many other liminal “living saints” who claimed similar impossible feats did not survive intense scrutiny.

We will never know how many, exactly, since so many of the Inquisition’s records have been lost, but it is undeniable that in the surviving records, those found guilty of fraud or diabolical mischief do outnumber those who were not. Consequently, to fully understand the context in which belief in impossible feats was forged, one must also consider cases of ostensibly holy individuals whose impossible feats failed to be recognized as genuinely divine; that is, one must take into account cases in which doubt and reason trumped faith.

And it was dangerous to go for sainthood! Tough as its one of the few ways for lay women to have an exciting life. See the story of Sor María, who was on the sainthood track until she got got.

All in all, [the Insquisition] concluded, María’s wounds were a miracle of “supernatural origin and beyond the ability of all human artifice.” The chief lesson to be learned from their report, they concluded, was that the miracles made possible by the Catholic faith testify to its genuinely divine character and that Sor María’s miracles had been ordained by God “to awaken those who are asleep during these times when malice reigns supreme.”

María had prevailed once again. After receiving this report, Master General Sixto Fabri declared all the accusations against María to be false. She was no fraud, after all, but a genuinely holy stigmatic and worthy of admiration. Shortly after this pronouncement was made, María was reelected prioress of La Anunciada, apparently well on her way to heaven, ultimately, and to a privileged place of honor in the Catholic Church. But Fabri had made one decision that undermined the results of the past three investigations: he refused to punish or expel those nuns who had “falsely” accused María of fraud.107 With her enemies still surrounding her, María had no real chance of avoiding conflict. Complaints kept flowing, many of which had more to do with her lack of humility and her poor leadership than with feigned miracles.

But she got re-investigated thanks to politics…

If the Spanish Armada blessed by María had been successful rather than a humiliating disaster, she might have escaped further scrutiny, at least for a few years. But the Armada’s annihilation was doubly harmful for her.

First, it raised new questions about her holiness and miracle-working powers since her blessing of the armada had obviously failed. Then, by giving the Portuguese some hope for their own cause—including the possibility of a successful English invasion led by the contender Antonio—that colossal defeat pulled María into the realm of conspiracies, willingly and unwillingly.

And it didn’t end well! The Inquisition, who contrary to reputation was remarkably diligent and even scientific in looking into miracle claims looking for multiple witnesses and weighing higher the views of not particularly pious people, took another look and concluded…

All of her prolonged ecstasies and trances had been faked; the stigmata on her head were small self-inflicted wounds; the stigmata on her hands, feet, and chest had been painted on; the cross-shaped bloodstains she produced were carefully crafted using blood from self-inflicted wounds; her levitations had been accomplished with the aid of thick-soled footgear known as chapines and wooden poles hidden from view under her habit; and her halos and luminescence had all been produced through the manipulation of lamps and mirrors. And so on. Not a single miracle had been genuine, even those healings that had supposedly occurred through her agency, which she now attributed to the faith of those who believed in her.115 In addition, she also confessed that she had feigned her pain during previous inspections of her wounds, knowing that this ruse would prevent the examiners from scouring them vigorously enough to take off the paint, and that she never thought anyone would ever dare to give her a scrubbing such as the one that revealed her duplicity.

Unlike Magdalena de la Cruz, who had immediately blamed the devil for her duplicity as soon as her fraudulence was discovered, Sor María insisted that she alone deserved blame. When pressed to ferret out the devil, she repeatedly held firm: nothing she had done whatsoever was demonically induced or the result of some pact with the devil. All she would admit was that she had excelled as an actress and illusionist and that her sole aim had been to gain admiration as a saint and mystic, for purely selfish reasons.

On November 7, 1588, the Inquisition passed sentence on María, declaring her guilty of “trickery and deceit” through her own “artifice and invention.” Consigning her to “perpetual incarceration” at a convent outside Lisbon, the inquisitors ordered her to spend the rest of her life continually doing penance, shut off from contact with the outside world. In addition, they decreed that all physical objects and pseudo-sacramentalia related to her were to be destroyed—images, books, manuscripts, crosses, beads, cloths, and any such items—to make it seem “as if they had never existed.”

When Brains Dream: Exploring the Science and Mystery of Sleep, Antonio Zadra and Robert Stickgold, 2021

The Scapegoat: The Brilliant Brief Life of the Duke of Buckingham, Lucy Hughes-Hallett, 2024. I will read anything Lucy writes.

Ex Supra, Tony Stark, 2022. Covered in a pod here.

Well, we’ve seen this before…

Lily Ottinger reports:

You’ve probably heard the recent rumors that Xi Jinping is facing the threat of imminent replacement by PLA and party elites. As a longtime observer of internal Russian politics, I’ve become desensitized to this type of speculation — rumors about Putin facing serious illness or internal overthrow first surfaced in 2012, then again in 2015, 2016, 2020, 2022, and 2023. How often has Xi gotten similar treatment? As far as I can tell, this is the third sustained instance of replacement rumors over the course of Xi’s 12-year tenure. A rundown:

September 2022: India Pumps ‘Xi Under House Arrest’

In September 2022, Premier Li Keqiang attended the annual meeting of the UN General Assembly in Xi’s place. This absence was followed by a period of about 10 days where Xi didn’t appear in public, which sparked rumors that he was under house arrest and would be replaced by General Li Qiaoming 李桥铭. This coincided with mass changes to commercial flights through China, with more than 9,000 flights being cancelled on September 21st. Rumors of a coup spread quickly across Indian news media in particular, and an Indian politician with more than 10 million followers tweeted about Xi’s supposed house arrest.

Flights returned to normal by the following Monday, and Xi resumed public appearances not long after. Rather than being under house arrest, it seems Xi was observing the zero-covid quarantine requirement after traveling to Uzbekistan for a Shanghai Cooperation Organization summit on September 16th. In December of 2022, Li Qiaoming was promoted to commander of the PLA Ground Force.

August 2024: Xi About to Die?

In August 2024, a Falun Gong-run Chinese media outlet called “Secret China” published a sensational report claiming Xi was gravely ill and that he would step down from his position as General Secretary at the annual Beidaihe leadership retreat, and only retain his position as President of China. The story alleged that party officials had decided to replace Xi with Hu Chunhua 胡春华, citing a cryptic tweet from a princeling. A YouTube video also cited in the report observed that, when Xi had met with the leaders of Vietnam, Italy, and East Timor in July 2024, the placards describing Xi’s position translated to “President of the PRC” instead of “General Secretary” or “Chairman of the Central Military Commission.”

Summer vacation came and went, and Hu Chunhua was not promoted.

May 2025 to Today — Third Time’s Probably not the Charm

This most recent round of rumors began in May, when Xi didn’t appear in public for about two weeks. Like the previous cases, rumors began with commentators in the diaspora — this time it was dissident journalist Cai Shenkun of the Falun Gong-run Epoch Times, citing an open letter detailing 28 of Xi’s policy mistakes and calling for his resignation that was allegedly emailed to Cai in December 2024.

The rumors gained momentum after Li Qiang attended the BRICS summit in Xi’s place on July 6th. Another parallel with previous rumors is that Indian news outlets like the Hindustan Times have been quick to disseminate this speculation for clicks. Yao Cheng, a former lieutenant colonel in the PLA Navy living in exile, alleges that Xi will be forced to retire at the 21st National Congress in October 2027 — so the rumors will probably stick around until then.

Bill is Right

has an excellent riff on the rumors on his latest Sharp China podcast. I (Jordan) couldn’t agree more.

Whatever we talk about here, just remember that when you get into early August and someone writes a story saying Xi has disappeared, it happens every year at this time.

What I will say is, first, we don't know. Chinese elite politics is a black box, so we cannot say with 100% confidence that nothing is going on. That's my initial caveat. However, when you look at the round of rumors and the way I approach them—this is something I've done in my newsletter for a long time.

My master's was in looking at Chinese politics. My master's advisor was at the time one of the US government's top experts on Chinese politics (now retired). I talk to lots of people who spend time focusing on elite Chinese politics. People get things wrong, but I do feel comfortable expressing opinions in this area more than in other areas.

When you see a rumor that Xi is in trouble, there are a few things you should ask. First, what is that being based on? Some of these reports were based on there being no Politburo meeting in May, so therefore Xi is in trouble. Someone says that to you without mentioning that there was also no Politburo meeting announced in May 2023. Actually, during this Party Congress, the number of monthly Politburo meetings that have been missed or not announced is much higher than in previous Party Congresses. If they don't give you that caveat, it tells me they're not actually doing the work of trying to see what is anomalous and what is not.

If someone says Xi Jinping disappeared for 13 days in May and that's a big problem, without saying that he has disappeared for similar periods previously—not around that early August break, but in other parts of the year—it's strange. We don't know why, but it has happened before and then he came back. Again, that's not evidence there is a problem.

Third, if somebody says Xi Jinping looks like he's being mentioned less in People's Daily or state media, that is patently not true. Demonstrably untrue. [See the China Media Project’s coverage on this topic.] Someone who says that is not doing the work. If someone cites a Twitter account or YouTube account, another red flag. That's a problem without understanding their track record of predictions and without checking whether they're linked to Falun Gong accounts. There are dozens of Falun Gong accounts spouting crap about all sorts of things, especially Chinese politics.

For example, there was an op-ed in The Spectator by a professor of Japanese history. The Spectator article, like the New York Post article, got a lot of traction. The guy cited some of the things I talked about without any background, then he cited accounts that were clearly Falun Gong accounts. But people read this—"Oh, it's The Spectator. Oh, it's the New York Post. It must be real." It cycles up and gets amplified.

I understand why a lot of people want to believe it's true. Hopium is a strong drug. The idea that different leadership—this mythical reformer—might come in, like the alleged reformer Wang Yang (poor Wang, he's retired). He's probably not happy that all these rumors about him taking over from Xi keep circulating during his retirement. The idea that Wang Yang is somehow this fantasy that people keep having—that there's just a reformer over there somewhere. As a friend of mine who worked on this stuff for a long time in the government said, "People are always looking for someone to mail the cake to."

I understand why people want to believe this, because the idea that they'll come in and fix the economy, maybe change the approach to Russia, maybe change the approach to trade—it's much easier to hope that's happening than to deal with the reality.

The actuarial tables give Xi a 3% chance of dying or becoming incapacitated this year, and the premium for holding a position for six months on polymarket is around 4%. Even a 2% chance of a coup seems high to me.

For Actual Insight into Xi

Check out our interview with Joseph Torigian on his monumental scholarly achievement: The Party’s Interest Comes First: The Life of Xi Zhongxun, Father of Xi Jinping, easily a contender for one of the best China books of the decade. Joseph’s goal, in his own words, was to “shine as much light into the darkness of the past as possible” to understand the nature of authoritarian politics, and he succeeds beyond my wildest expectations.

This biography gives me a feel for Chinese politics that I honestly thought I’d never have. It does an incredible job of digging deep to shed light on some of the most consequential moments in CCP history, as well as conveying what it was like to live as a senior official under Mao and Deng. Reading it was a powerful experience at both an intellectual and human level.

We get memorable vignettes, like 15-year-old Xi Zhongxun attempting to assassinate a teacher, or General Peng Dehuai using his shoe to silence Xi Zhongxun’s snoring in their shared bunk.

In this interview, we discuss:

• What we can learn about authoritarianism, the CCP, and China’s future from studying Xi’s father,

• Torigian’s methodology for uncovering hidden Party history,

• How the Party became an existential source of meaning, and how it weaponized suffering to paradoxically deepen political loyalty,

• The interplay of family, love, and career under the all-encompassing shadow of the Party,

Listen now on iTunes, Spotify, YouTube, or your favorite podcast app. Transcript here if you prefer!

‘Containing Xi’, or refining party rule?

What does the latest politburo meeting reveal about Chinese politics?

Dr. is Lecturer in Politics at the University of Glasgow. This piece was originally posted by the Council on Geostrategy.

The Investigator | No. 15/2025

New regulations on central decision making bodies

Speculation about the June politburo meeting readout is feeding rumours of Xi Jinping, General Secretary of the Chinese Communist Party (CCP)’s ‘weakening’ hold on power. The readout announced that the politburo had reviewed new ‘Regulations on Party Centre Decision Making, Discussion and Coordinating Body Work’ [‘党中央决策议事协调机构工作条例’].

This speculation hangs on two ideas. The wilder of the two is that some elusive group of actors has decided to establish a new body above Xi. This is based on a misreading of the first line of the readout – a boilerplate introduction to the topic at hand, not a declaration of a new power centre being established. The second is that these regulations are designed to rein in Xi’s power by placing it under institutionalised constraint.

Both readings ignore basic facts and two ongoing, well-documented trends: centralised CCP decision making nudging out the state, and ‘rule-based rule of the party’ [‘依规治党’]. Linking the readout to speculation about Xi’s ‘decline’ misses an opportunity to examine the regulations’ significance for the People’s Republic of China’s (PRC) evolving political system. While analysis of Xi’s impact on the system rightly points to power ‘centralisation’ and ‘personalisation’, less is known about how the CCP and its leader use power once it has been centralised. The regulations – not yet released – are likely part of the party centre’s ongoing task of modifying institutions to help it put centralised power into action.

To be clear, the politburo meeting readout contains nothing shocking. The party centre has long been planning to formulate the regulations in question. We know this because they were mentioned in April 2023’s Central Intra-Party Regulations Formulation Work Plan Outline (2023-2027) (hereinafter ‘Party Regulations Formulation Outline’).

What is ‘centralised power’ without rules to use it?

These planned regulations are part of Xi’s longstanding pet project to govern the CCP and the PRC by rules – his macro-plan combines ‘rule-based rule of the party’ and ‘law-based rule of the country.’ On coming to power, Xi spoke of the party’s need for rules to rein in the party’s excesses and avoid the existential threat of a breakdown in the party-people relationship, and he has been making rules ever since.

Xi has made rules about making rules – stipulating how party regulations can be formulated and by whom – and has made rules to put his fellow party members’ authority ‘in a cage’ [‘把权力关在笼子里’]. He has used rules to his advantage, giving himself maximum flexibility for pushing through decisions and to remould the party’s internal workings, creating new incentive structures, and making himself the ‘core’ with whom all must ‘align.’ He has used rules for everything; from dictating how many dishes cadres may serve when entertaining to delineating party powers to manipulate the state.

Xi takes rules, and their use for creating governance mechanisms, seriously. He regards them as integral to ‘modernising’ Chinese governance. In 2019, a Central Committee Plenum document resolved to ‘better translate China’s institutional strengths into national governance efficacy.’ The document called for ‘strengthening the role’ of the bodies involved in the controversial politburo readout, improving mechanisms for ensuring party centre decisions are implemented and ‘strictly enforcing’ the instruction requesting and reporting system – a little-discussed system of longstanding importance to the party’s internal workings which is discussed later in this article.

Xi has used rules not only to maximise his concentration of power, but also to facilitate its use. The Central Military Commission ‘Chairman in Charge’ system [‘中央军委主席负责制’] is a case in point. While analysts stress Xi’s being ‘in charge’ – and his ultimate authority to make decisions – his attempt to create a whole set of rules and mechanisms to facilitate this decision-making role go under the radar. For instance, the ‘three mechanisms’ developed through intra-party rules seek to serve the Chairman’s decision making capabilities.

What are Party Centre DDC bodies?

Party Centre Decision Making, Discussion and Coordination bodies (DDC bodies) form a level of authority which shifts major decision making upwards, away from state institutions. Some were originally ‘Leadership Small Groups’ transformed into commissions under the 2018 institutional reform. Their upgrade sought to ‘strengthen the Party Centre’s centralised, unified leadership over major work.’ Others, such as the Central Science and Technology Commission (CSTC), were established directly as commissions.

Though researchers refer to the new regulations as the ‘regulations on the Work of the Party Central Committees [DDC bodies]’, the document’s official title uses ‘Party Centre’ [‘党中央’], not ‘Central Committee’ [‘中央委员会’]. This reflects the nature of Party Centre DDC bodies: they are beholden to the 24 men of the politburo or the seven men of its Standing Committee (PBSC) and not to the larger Central Committee. Xi himself, as General Secretary, has direct control over the topics of DDC body meetings, either deciding on, or giving the go-ahead to, a meeting.

On the functions of Party Centre DDC bodies, the language from the June politburo readout is almost verbatim that of the aforementioned Party Regulations Formulation Outline. The Outline stated that such regulations are needed to help DDC bodies fulfil their functions of: ‘top-level design, choreographing and coordinating, integrated promotion, and monitoring and urging implementation’ [‘顶层设计, 统筹协调, 整体推进, 督促落实’] of major work. This is precisely the language used in the 2020 Central Committee Work Regulations, which stipulate the Central Committee’s power to create DDC bodies. The politburo readout uses the same ‘4x4’ character expression. This language is also used for the specific DDC bodies. The 2023 institutional reform plan describes the then-new Central Finance Commission’s functions in exactly the same way. The PRC Law on Foreign Relations grants the same functions to the Central Foreign Affairs Commission.

The new regulations will complement or adjust existing rules. Said rules themselves reveal something of these elusive power centres’ workings. Xi’s ten-year rule making spree has reinvigorated the Instruction Requesting and Reporting system [‘请示报告制度’] (IRR). The IRR lets lower ranking bodies request instructions from senior organisations and report back up on implementation. It enables party entities to respond to eventualities on the ground in their locality or policy field, although it can also create logjams.

At least since 2019, DDC bodies – and their leaders individually – have been permitted to play the role of IRR responders. This may be a way of delegating party centre authority while helping facilitate calibration of policies with central requirements. It could also help with the apparent spike in IRR requests resulting from a decade of relentless campaigns, pervasive punishments and strict demands for alignment with party centre policy, which has left officials ‘lying flat’ or looking for other ways to avoid culpability. DDC bodies have been ordered to create detailed and specific rules for implementing IRR in their own fields. This could amount to substantive power delegation to the heads of DDC bodies’ implementing offices.

The respective arrangements of the Central Foreign Affairs Commission (CFAC) and the CSTC serve as useful discussion points. Wang Yi, CFAC Office Head, holds triple roles, also serving as a politburo member and Minister of Foreign Affairs. This links CFAC decisions directly to the principal state implementing agency. With Wang heading both the Ministry of Foreign Affairs (MFA) and the CFAC Office, he can implement CFAC decisions through both the MFA and through coordinated action with other CFAC member ministries. Wang’s position as CFAC Office Head may be an authoritative lever to press other agencies to coordinate with the MFA. While the CFAC-MFA setup may be an outlier (due to the demise of Qin Gang, Wang’s predecessor), the CSTC’s arrangements are similar. Yin Hejun, Minister of Science and Technology (and Central Committee member), reportedly doubles up as CSTC Office Head, allowing CSTC decisions to flow directly to the Ministry of Science and Technology and giving Yin the lever of CSTC Office Head to help coordinate the implementation of CSTC decisions.

Xi ‘in decline’ or delegated powers being refined?

Returning to the politburo meeting readout, considering the powerful role of DDC Office Heads it is unsurprising that regulations should require DDC bodies to ‘coordinate, not stand in for, and perform as required, not overstep’ [‘统筹不代替、到位不越位’]. While speculation sees this as a ‘direct criticism of Xi’, it is more likely that the relevant provision seeks to regulate the power delegated by the Party Centre. Notably, Xi has himself used the expression in relation to the Central Comprehensive Law-based-rule Commission.

Overlooking subnational practice misses a chance to reflect on assumptions. Countering the notion that, from the politburo readout, ‘some language could reasonably be read as Xi being shunted aside’, provincial-level party committees are repeating that very same language in a promise to do better. Shanxi, Chongqing and Xinjiang, for example, declared that they will ‘study the spirit’ of Xi’s ‘important [politburo meeting] speech’ and improve the practices of their subnational DDC bodies. They will produce ‘realistic and effective policy measures’ [‘切合实际、行之有效的政策举措’] – a line also in the politburo meeting readout – and better regulate DDC body establishment and operation under their jurisdiction to ‘ensure implementation of Party Centre decisions.’ Provincial-level party committees have the power to make intra-party regulations. It appears communications about the Party Centre Regulations are prompting them to follow suit and make implementing documents to tighten up on local DDC body practice.

The regulations are likely less a signal of Xi’s waning star and more a run-of-the-mill move to hammer out the details of how the CCP uses the power it has centralised under Xi’s first two and a half terms. It may be that the Party Centre (or members thereof) will attempt to have the regulations’ content incorporated into the Party Charter at the 21st National Party Congress in 2027. This would give the CCP ‘constitutional’ credibility to the use of DDC bodies in governance, consolidating their place in the New Era Party bureaucracy.

Joseph Torigian’s The Party’s Interest Comes First: The Life of Xi Zhongxun, Father of Xi Jinping is a monumental scholarly achievement — easily a contender for one of the best China books of the decade. Joseph’s goal, in his own words, was to “shine as much light into the darkness of the past as possible” to understand the nature of authoritarian politics, and he succeeds beyond my wildest expectations.

This biography gives me a feel for Chinese politics that I honestly thought I’d never have. It does an incredible job of digging deep to shed light on some of the most consequential moments in CCP history, as well as conveying what it was like to live as a senior official under Mao and Deng. Reading it was a powerful experience at both an intellectual and human level.

We get memorable vignettes, like 15-year-old Xi Zhongxun attempting to assassinate a teacher, or General Peng Dehuai using his shoe to silence Xi Zhongxun’s snoring in their shared bunk.

In this interview, we discuss:

• What we can learn about authoritarianism, the CCP, and China’s future from studying Xi’s father,

• Torigian’s methodology for uncovering hidden Party history,

• How the Party became an existential source of meaning, and how it weaponized suffering to paradoxically deepen political loyalty,

• The arc of Xi Zhongxun’s life — from a young revolutionary to key advocate of reform — and his role during Tiananmen,

• The interplay of family, love, and career under the all-encompassing shadow of the Party,

• The role of “Surrogate fathers” and patronage in navigating political ascent,

• How literature shaped China’s early revolutionaries, and even impacted the Party as we know it today.

Co-hosting today is Jon Sine, former ChinaTalk intern.

Listen now on iTunes, Spotify, YouTube, or your favorite podcast app.

From Teenage Assassin to Tiananmen

Jordan Schneider: Joseph, how do you define the purpose of this book?

Joseph Torigian: There is inherent value in going to the past and telling stories about these people’s lives, especially because they suffered so much. Second, this book helps people understand authoritarian politics and how the party works, in a way that gives us special insight into Xi Jinping. Finally, the book allows you to draw your own conclusions in many ways.

One of the themes that emerges from the book is how much suffering people in the party experienced and how they brought different meanings to it. The party imposes one interpretation on the past, but what I wanted to do with my book is facilitate multiple different interpretations for people who want to read it closely.

Jordan Schneider: Generally, when people read biographies of major political figures, there’s a bias in authors to tie bows around the narrative and make some overwhelming moral judgment about whether their decisions were right or wrong, and whether this person was morally upright or lacking in how they behaved and their impact on the world.

Remarkably, this book stops before going there. This man, whose life was incredibly gray, complex, and shaded from many different dimensions, could have been interpreted through party narratives or through the lens of liberal outside reformers projecting their ideas onto him. As you were going through his life and trying to portray what you learned and discovered to your audience, how did you think about your role as the author?

Joseph Torigian: I wanted the evidence to lead me in the direction of what was most interesting about his life. I didn’t feel there was a missing part of the literature that needed to be explored. I just wanted to do justice to his life in a way that would allow people to make their own judgments. As you see in the book, people in his own life — and Xi Zhongxun himself — sometimes found it unclear which morality was most relevant to make a particular decision.

Narrative certainly plays a role in the book. I wanted to make people feel just how dramatic, exciting, and electric it was to be a member of this organization, especially in the times and places where Xi Zhongxun found himself. I wanted people to draw broader lessons about how the party worked.

But I also didn’t want to allow the need for it all to come together or for there to be a single argument to do violence to the evidence. As I said before, I wanted people to be able to draw their own conclusions about the meaning of his life. Certainly I needed to bring an interpretive lens and my own analysis because facts don’t speak for themselves. Even choosing which evidence to include in the book required some level of reflection and thought. In many ways, a lot of the conclusions are obvious, but I don’t spell them out because they’re obvious. Allowing people to think about the most striking parts of his life on their own is a better approach.

Jordan Schneider: Before we get too meta, I think it’s important to ground folks. Generally, when you read a political biography, it’s about the number one leader, maybe the number two. Xi Zhongxun, though he was not quite at that level, lived one of the most incredible lives I’ve ever engaged with in book format. Joseph, can you give us an introduction to the arc of this man’s life?

Joseph Torigian: He grew up in Shaanxi province, which was a fascinating place because his home village was near Xi’an, where the first unified state was forged by Qin Shi Huang, and dynasties had ruled from there for millennia. But by the time he was born, two years after the Qing collapsed, it had fallen into banditry, war, famine, and poverty. It was a place of extremes.

He’s trying to figure out a way to address these wrenching problems facing society — feudalism, imperialist encroachment. He’s attracted to the party, but he doesn’t really understand it well. In fact, he’s able to read this revolutionary literature because the Nationalists and the Communists are still getting along. But of course, that all changed in 1927 when Chiang Kai-shek betrayed the Communists and massacred them.

It’s in this milieu of violence that Xi Zhongxun is told to kill an academic administrator. He fails, instead getting a bunch of other teachers sick. He goes to prison and joins the party while incarcerated.

Jordan Schneider: At nineteen?

Joseph Torigian: Fifteen. He was fifteen years old. He became a founding member — although not one of the most important ones — of the base area in the northwest, which is where the Long March concludes. If it weren’t for that base area there, Mao Zedong and the other members of the Central Leadership would have needed to move even farther away from the interior Chinese areas.

He works in these fascinating places on the border of the base area, where he has to think about the United Front and manage relations with the Nationalists. They’re not at war with the Nationalists, but they’re not at peace either. It’s a very complicated liminal world he finds himself in.

In the late 1940s and early 1950s, he was head of the Northwest Bureau, which is a giant expanse of Chinese territory that includes not just Han areas, but also Uyghurs, Tibetans, and Hui. He was brought to the capital and worked as Minister of Propaganda. There, he was a witness to several power struggles — the target of the first great purge of the People’s Republic of China is a man named Gao Gang, who was also a Northwesterner quite close to Xi Zhongxun.

Xi Zhongxun worked for Zhou Enlai, the famous premier who ran the government at the State Council. Xi Zhongxun was purged in 1962 because of a novel. He spent 16 years in the political wilderness facing incarceration, humiliation, and exile.

When he went back to work, it was in Guangdong province, where the failures of the revolution are perhaps most obvious. Tens of thousands of people were fleeing from the Communist mainland to capitalist Hong Kong. He sees with his own eyes just how developed things are right across the river.

Then he went back to Beijing to work on the Secretariat, managing its daily affairs. He’s the right-hand man to Hu Yaobang 胡耀邦, the general secretary of the party. He has a fascinating set of tasks — he spends about 70% of his time on the United Front, returning to ethnic politics. But he also looks at Beijing’s relations with foreign leftist, revolutionary, and communist parties.

He witnessed this difficult relationship between Hu Yaobang and Deng Xiaoping, which mirrored Zhou Enlai’s relationship with Mao Zedong. It all comes together in June 1989 when Xi Zhongxun faces the question of how he’s going to react to these student protests. Ultimately, he goes along with the crackdown, even though it seems he had very intense doubts and skepticism about using violence to resolve the crisis.

He spent much of his later years in the south, only rarely returning to Beijing. It’s quite a dramatic life — marked in particular by persecution by his own party. On many occasions, the party hurt people close to him or forced him to betray people close to him. One of the central themes that emerges is how he balances his own thoughts, emotions, and feelings with a party to which he is totally loyal, but which often does things that are deeply, deeply distressing to him.

Jordan Schneider: This book was especially engaging because you were able to find sources that really brought you into all these moments in his life. They also gave you a sense of his interior monologue and psychology at all these moments of incredible stakes, trial, and national as well as personal drama.

You take us on this arc from this 15-year-old who tries to kill his teacher all the way through to Tiananmen Square, including the family life and the experience of the Great Leap Forward, the Cultural Revolution, and reform and opening. But it’s all through this man who is incredibly faithful but also scarred in ways that are almost impossible to comprehend as a coddled person who grew up in America over the past 30 years. This book is a tremendous accomplishment.

Jon Sine: You open the book saying that Xi Zhongxun can fill one of three roles at least — his legend in the party, the role of the father to Xi Jinping, and the role that you take him as, which is a lens to party history. Could you explain how this book evolved, and why you chose to focus on Xi Zhongxun? I’m sure you’ve experienced the immediate reaction from many readers is, “How does this explain Xi Jinping?” when the thrust of the book is how Xi Zhongxun illuminates party history through the 20th century. We’d love to hear more about that.

Joseph Torigian: The book was kind of an accident, actually. I was asked to participate in an edited issue of an academic journal and was asked to write about Xi Jinping. I wasn’t sure how to go about it. Then I decided that the best way would be through a historical lens. I envisioned an article that was half about Xi Zhongxun and half about Xi Jinping as a young person.

I had just finished a book on elite party history, and that had been a training session on how to do it, find sources, and interpret them. As I started working on it, I realized there were many more sources than I had envisioned, even though I had just finished this other project. But also that Xi Zhongxun was an interesting person, not just because he was the father of Xi Jinping, but precisely as you said — he is someone whose life illuminates in powerful ways party history and party culture. His life is really a microcosm of the party in the 20th century.

With the book, it’s certainly the case that people are reading it because of Xi Jinping, and I’m glad people are reading it. I’m not going to be nitpicky about what reason they pick up the book. But ultimately, it says two things about Xi Jinping.

One is that it says a lot about the party and how the party works, and how every leader has tried to manage dilemmas that have been with it from the beginning. These are dilemmas that can be managed, not problems that can be solved. Xi Jinping is bringing his own approach, and we can understand how he’s bringing that approach because party history gives us the context. That’s the clearest line between the book and Xi Jinping.

But of course, there’s another line which is psychological. Here, we need to be careful because even Xi Jinping’s siblings drew very different lessons from the meaning of their father’s life. Xi Jinping, like his father, is a product of multiple influences. He’s someone who changes his mind, someone who reacts often to the peculiarities of his situation as opposed to some dogmatic worldview that always makes him do the same thing.

I hope my book is one tool in a box of tools that will be useful as we think about contemporary China. I didn’t write the book so that people will know what he’s going to do during the tariff war, although these Taiwanese talk show hosts quite loudly thought that my book might be relevant in that regard. That might be true to some extent. But the real value of the book is to take a step back and just get a general sense of the party and the milieu from which Xi Jinping emerged.

The Politics of Suffering

Jordan Schneider: There’s a very striking quote you have at the end of your book from Xi Jinping, where he wrote a letter to his father on his 88th birthday. I guess he couldn’t visit because everyone in the party is too busy to be at their kids’ births and parents’ birthdays —

Joseph Torigian: The party’s interest comes first, Jordan.

Jordan Schneider: Exactly. But the line that stuck out to me was him saying, despite the Cultural Revolution, “I always stubbornly believed that my father was a great hero and that he was a father most worthy of feeling proud of.”

The central question in Xi Zhongxun’s life, as well as Xi Jinping’s life, is, “Why are you sticking with these guys?” The party was treating them terribly while they were giving their lives to this organization. On several occasions, the lesson seemed to be that the party does not care about you and will chew you up and spit you out in ways that are illogical and detrimental both for yourself and for the country.

But you write this line explaining it:

“Some may wonder why Xi Jinping would remain so devoted to an organization that severely persecuted his own father. But perhaps the better question is how could Xi Jinping betray the party for which his father sacrificed so much?”

That’s the central psychological question we explore over these 500 pages — how does this man relate to this organization, to which he has devoted his life, but is riddled with problems? What did the party mean to Xi Zhongxun?

Joseph Torigian: The party was the source of meaning in his life. It was the source of purpose. It was salvation. It was a place where a young boy growing up in a society wrecked by turmoil found an answer to solving that chaos. This was someone who didn’t really understand the intricacies of Das Kapital, but believed that only an organization as powerful as the Chinese Communist Party could respond to the challenges China was facing. It was the only way of organizing society that would allow China to return to its rightful place in the world and defeat the imperialists, the warlords, and all these other people who had brought China to the brink of collapse.

Sometimes when you join an organization like this, you don’t always fully understand what it’s going to be, but once you’re in it, it also changes you. You understand the principle of the organization, but then the organization has all these methods for changing your very soul. That’s not the language I’m using — it’s the language the Communists used. You have self-criticism, you have study sessions.

If you believe you are participating in a grand adventure, where you are a leading figure in a world historical force that is inevitable, that is a really exciting place to be. Also, think for a moment that you keep seeing the party failing over and over again, and then suddenly one particular person emerges in the form of Mao Zedong, who seems to finally have a way of getting it right. He takes this ragtag group of revolutionaries who had nearly been destroyed on many occasions and forges a new nation.

For Xi Zhongxun to reject someone like that would be rejecting his own self. When the party persecuted him, it didn’t turn him away from the party. It motivated him to ask, “How do I win back the party’s trust? How do I show that I’m better than other people? How do I show that, actually, precisely because I am going through this torment and still return to it, I am even more dedicated than other people who didn’t have to suffer?”

You see other people grumbling about the party, and Xi Zhongxun would say, “Yeah, well, I bet you didn’t go through this.” It was almost like a badge of honor that he could weaponize — this politics of suffering, of who went through the worst experiences, and that gives you political capital within the party. It’s an interesting way to think about life, but it does raise the question — what kind of suffering leads to dedication and what leads to alienation? This is really a fundamental question or puzzle.

Xi Zhongxun admitted that during the Cultural Revolution, he went through a period of doubt. But Xi Zhongxun says precisely because he wasn’t sure for a little while and nevertheless returned to the cause of the party, therefore his dedication is totally unshakable and even more powerful than someone who didn’t go through that experience.

That also raises a question for young people in China today: If you want them to eat bitterness, how do you do that without turning them away from the cause entirely? That’s a philosophical question in some ways, and it’s one that Xi Jinping thinks about all the time—something that keeps him up at night, I’m sure. It’s also at the very heart of the question of the future of the People’s Republic of China: whether or not they get that right.

Jordan Schneider: Yeah, there was this very evocative moment with Xi Zhongxun towards the end of his life. In the 80s, during this spiritual pollution campaign, all these old revolutionaries were excited for the future but also worried that the youth were going soft and didn’t understand. One of his answers was that they just needed the old comrades to go into the high schools and tell their stories — like the astronaut showing up in middle school and telling you how cool space is.

Joseph Torigian: Yeah. He talks about them as flowers in a hothouse garden, right? They need to go out into the real world so they can survive in the real world. Interestingly enough, he thinks that young people who went through the Cultural Revolution might be a good group of successors because they underwent these terrible experiences.

Of course, the Cultural Revolution was largely about succession issues at the very top and at the very bottom — getting rid of Liu Shaoqi, the revisionist who proved to be unsuitable to the task, but also toughening young people who had been growing up in the privilege of youth in a socialist nation. Many of those youth actually did go through a period of disillusionment. They wanted to have fun, they wanted to study overseas, they wanted to make money.

Xi Zhongxun consciously contrasted himself with those other youth, and he saw a lot of danger in this spiritual crisis that many young people in China found themselves in during the 1980s. Of course, this was one of the roots of what happened in 1989.

Jordan Schneider: Yeah, I mean, it is a religious attachment. There’s no other word to describe what someone like Xi Zhongxun or Xi Jinping has to the party. You have these little trials of faith, right?

The decision to go left or go right from Xi Zhongxun, as compared to all the other princelings who are having fun in Hong Kong and driving cars or whatnot. A wide swath of humanity would get disillusioned from the fact that you have all these people you idolize turn on you and say you’re a capitalist roader and making leftist deviations and being anti-party and whatnot.

The emotional effort to harmonize whatever you’ve seen and experienced and look back at that time in the wilderness from 1962 to 1978 — not as these people stealing 20 years of my life, but actually as a forging event that is making me a better party cadre, more well-honed to serve the people — requires some kind of deep spiritual belief. Even something as catastrophic as watching this party starve tens of millions of people or turn the country upside down in a Cultural Revolution is not able to shake it.

Joseph Torigian: We should remember though that even though he had this idealism, this conviction, these were not easy experiences to live through and he suffered. It was difficult for him to understand why this was happening to him. Part of it was this sense that the party was still the best future for China. But part of it was utilitarian, right? The sense that if we are going to move forward, we can’t think about the past too much.

There’s both an idealistic sense, but also this idea that if we are too reflective, if we’re too critical of the party, if we go too far, then we’re going to shake the whole house of cards and it’s going to come down with us. That was a central puzzle for the 1980s: how do we justify moving in any direction but still maintain this sense of faith even as we can’t even say what communism really is because we’re groping as we try to figure out a new direction?

That was the reason why there were so many zigs and zags in the 1980s. That was the shoal that Hu Yaobang crashed on. You could understand in theory how you could integrate reform with conservative principles because you needed to have stability to reform and you needed to have reform to have stability. But as Hu Yaobang said in his self-criticism, even though in theory that makes sense, practically it’s not easy and you need a lot of political skill. His inability to do that was one of the reasons that Deng Xiaoping scapegoated him.

Of course, Xi Zhongxun is this very close associate of Hu Yaobang and felt very strongly about that, even though he put the party’s interests first and went along even with Deng’s persecution of this very close associate of his.

Jon Sine: You cite Yuri Slezkine’s The House of Government several times, primarily for its generational analysis and its examination of how faith erodes over time. But Slezkine’s central thesis is that the old Bolsheviks functioned essentially as a millenarian sect. You don’t use much of that religious language in your book, but I’m curious whether you see similar parallels. As someone who studies both Soviet and Chinese history, what similarities and differences do you observe between the old Bolsheviks and these Chinese revolutionaries?

Joseph Torigian: Culturally, they’re very similar. Xi Zhongxun, on his very first trip overseas, goes to Russia and meets these sinologists. He talks about how he loves reading a Russian novel, What Is to Be Done? by Chernyshevsky. Of course, it has this very famous character, Rakhmetov, a professional revolutionary who would sleep on a bed of nails to inculcate his revolutionary élan. Xi Zhongxun said he wanted to mimic this person and walk around in the wind, in the rain, in the snow, and sleep on a hot kang stone bed with his shirt off. In that sense, there are some very powerful similarities.

Just like during the terror in the Soviet Union, people were sent to their execution yelling “Long live Stalin!” During the Cultural Revolution, these old revolutionaries had no idea what Mao wanted, didn’t understand what Mao wanted, and were going through intense physical and emotional torment. For them, the problem wasn’t Mao — it was that they weren’t keeping up with Mao. It was that they didn’t understand Mao. They were looking into themselves for the problem. They weren’t looking at Mao for the problem. That’s how they thought about the world.

More broadly speaking, what’s interesting too is sometimes when I give this talk, a person will come up to me and say, “You know, I was raised in a faith tradition. I went to Catholic school. A lot of what you’re saying, it’s not that unfamiliar.” Seeing meaning in suffering, redemption in suffering — it’s a universal thing.

Now, whether you call it a religion or not, or how similar these phenomena are, that’s a big discussion. Certainly, the contexts are very different. Class struggle is a message of violence. The gospel is a message of love. The gospel is also a message of individual conscience, while the party’s message is that conscience is the party’s conscience.

I don’t want to essentialize similarities or differences, but at the very least, even though many of the themes we’re discussing today seem very foreign to us, I think with radical empathy, it’s not impossible to understand these lives, and we should try to understand these lives. We should never underestimate how hard it is. We should never think that it’s easy to appreciate how culturally dissimilar or unique these situations can be, but nevertheless we’re all part of the human condition.

Jordan Schneider: I also think that one level down, at a tactical level, you see a lot of parallels in the moves that the church uses and the moves that the party uses. This whole confession, self-criticism thing, which is a central part of these guys’ lives, is this introspection and these doctrinal fights of who has the right line. Are we doing this right? The importance of words is just a really remarkable thing — they’re arguing about phrases and spending days and days trying to get the idea right, which is something that’s really important for religion and not really important for a lot of other parts of life. But it was literally life and death for these folks a lot of times.

Joseph Torigian: That’s one of the other ironies of this book. One of the communities in China that Xi Zhongxun was tasked with managing in the 1950s and then again in the 1980s was Catholics. He sees Catholics as the biggest problem for the party, partly because they’re in communion with the Pope and Xi Zhongxun sees the Pope as a tool of imperialism. But they’re also hierarchical, they’re very motivated, they’re very idealistic. They also have these texts that need to be interpreted. It’s not surprising that they have a very interesting collision both in the 50s and the 1980s, which people can read about in the book.

Jordan Schneider: But the Catholics figured out succession in a way that maybe the party hasn’t. Jon…

“Eating Sawdust” 吃木屑

Jon Sine: Let’s talk about the art and science of putting together a book like this, because I think a lot of people have a perception, maybe a misperception, that the regime today is hiding everything, that if you go to the government sources, sanctioned sources, you can’t really rely on them. But when you have the background that you do, there are things that you can pull from them and you can triangulate something — I think you call it “a mosaic.” There are about 2,000 endnotes in this book, and I know that you’ve probably had to cut quite a bit.

Joseph Torigian: A lot.

Jon Sine: Can you speak on the art and the science behind putting this together?

Joseph Torigian: Well, it wasn’t easy. We’ve talked a lot about ideals and conviction, and for Xi Jinping to achieve that, you need to have a single view of history. Xi Jinping was a surprise to a lot of people, but one of the first signs that he wasn’t going to be a reformer like people thought his father was came when he gave the speech about the collapse of the Soviet Union. He said one of the reasons they collapsed is they lost control of their history, and nobody took it seriously anymore.

If you want people to think that the party is an inevitability, that it just goes from one triumph to another, the more you talk about its dark past, the more dangerous that is for party rule. He has this term “historical nihilism” that he uses to describe people who talk about the party’s mistakes too much. For someone with that view to write about their father is certainly something that is challenging.

I don’t want to say that my book is the final verdict on Xi Zhongxun’s life. It really is just the latest draft. Whether or not it’s a meaningful draft has to do with whether or not there are enough new sources since the last person took a look and whether or not you’re doing enough justice to the sources that are available.

The question becomes: what are those sources? I don’t really think in terms of good and bad evidence. I think in terms of getting as much evidence as possible and then parsing it and then using it in different ways as that piece of evidence demands. You can’t just go to one archive and collect the materials and then write it up. You really need to have a detective sensibility. You need to be sensitive to possibility and limitation. You need to have a capacity for tedium.

There’s this expression “eating sawdust,” that China watchershands have used, becausewhere you need to eat a lot to get any protein. I’d probably say 99.99999% of everything I read is totally useless. But when I do find something that is really useful, it’s very, very exciting.

I was able to get some primary sources, of course. For a variety of reasons, a lot of documents or internally circulated material have become available at US institutions, as you saw in the book. Some very interesting document collections were also published in Hong Kong, as well as memoirs and party history written by insiders who had access to a lot of evidence. I did a lot of interviews for the book. I used a lot of material published in the mainland. There were these party history journals that were really pushing the envelope for a very long time that included subjects that were very sensitive — the Cultural Revolution, the Great Leap Forward.

I use official sources as well. That might strike people as surprising, but they are absolutely essential. There are cases where they changed the literal words that people said, but they’re quite rare. More often the problems are what they don’t include. But what that means is that you can treat what they do include as a hypothesis for what happened, at the very least. Then you look for other sources to back it up. It sensitizes you to things you might want to look at. Or, you use this phrase ’mosaic theory’ — they reveal a piece of evidence, unaware that I can connect it with all these other pieces of material.

The other thing I did was to make a list of every time Xi Zhongxun met with a foreigner. Then I went to speak to that foreigner, or I went to their archives, which are open. That allowed me to tell a story not just about Beijing’s relations with global communism, which I think is a very meaningful part of the book, a very significant part of the book, but even to get his understanding of Tibet, right? I interviewed the Dalai Lama. I was able to get these Tibetan language transcripts of negotiations between Dharamsala and Xi Zhongxun in Beijing and have them translated into English so I could use them in the book.

If we are taking these issues, these questions, these subjects with the seriousness they deserve — and because in a way it is a matter of life and death, right? I mean, people are dying and we’re writing about why these people are dying — we should feel a sense of moral obligation to do everything we can to get these stories right. I certainly didn’t in this book, as I say in the acknowledgments. I’m sure that in some cases somebody else will come and tell me I got something wrong, but I’m very open to that. I hope people will tell me.

I’m hoping my book will lead to further questions and other avenues for research because the world files its information very miscellaneously, which means that people may figure out that there was something I didn’t look at because I didn’t think of it. If there weren’t so many people who knew what I was working on, who said, “Oh, by the way, I stumbled onto this.”You never know what tiny, tiny breakthrough then leads to huge breakthroughs. That happened so many times in my book that there was some complete stroke of luck where I was looking at something completely different, or I was going for a long walk and I said, “You know, maybe I should go back and reread this.” Then you reread it and then it has a new meaning that wasn’t there before.

For example, the Li Rui Diaries. Li Rui (李锐) was a secretary to Mao Zedong. He was working for the Organization Department in the eighties, knew Xi Zhongxun very well, was the senior pro-reform constitutionalist comrade in Beijing during the 2000s and 2010s. His diaries are available at the Hoover Archive. Now they’re extraordinarily hard to use because it’s a diary. There’s a lot in there that’s wrong because he would write down rumors, and he wrote a lot in shorthand, which I couldn’t understand at all the first time I read it. But when you read other stuff, you go back, even if it’s just a few characters, and suddenly it has this huge meaning that was totally invisible to you the first time that you looked at it.

Jordan Schneider: For instance?

Joseph Torigian: One example was a curious incident in the 1980s where there was a debate about how to think about princelings and how to promote princelings, especially the ones that had engaged in really brutal violence as the so-called old Red Guards in the first months of the Cultural Revolution. Somebody wrote a letter and it was immediately approved by Chen Yun. It raised these questions about whether or not the whole thing had been planned ahead of time in the first place.

I saw the names had been mentioned in a conversation between Xi Zhongxun and Li Rui, and I wasn’t quite sure what that was all about. But I read a memoir written by the person who knew about this letter and then suddenly, because I knew that background, this very brief discussion between Xi Zhongxun and Li Rui that Li Rui didn’t include a lot of information about because it was his own diaries — suddenly I knew what Xi Zhongxun thought about this very interesting moment. That would be one example.

Jordan Schneider: That is just a tiny window into the amount of work that went into this book. Because you have that moment, but you actually have a 90-year life, and on every single page you can feel the amount of sawdust that you chewed through, Joseph.

Joseph Torigian: Each chapter has its own ecology of sources. I need to mention that I relied heavily on histories written by Chinese scholars — work published both on the mainland and in Hong Kong. For his family life, I inevitably needed to rely on commemorative essays written by the children about their father. You need to take these sources seriously, but one way to approach them is to understand the broader context.

What was it like to grow up in Beijing in the 1950s? What was it like to attend that particular school? Often, instead of researching Xi Jinping directly, I would take a different approach. If Xi Jinping went to August 1st School (北京市八一学校), I’d find every book I could about August 1st School. When Xi Zhongxun was exiled to a factory in Luoyang, I’d research the history of that factory. His name might not even appear in these sources, but you can at least understand what happened to him while he was there and what he would have experienced, even if we can’t say exactly what he did or thought about it.

Jordan Schneider: Those moments showcased your scholarship at its best. People will probably flip ahead to read the Tiananmen chapter first, but I appreciated how you condensed these experiences. In just a few paragraphs, you gave readers a sense of what these worlds were like — it’s some Robert Caro-level work. You’re not literally getting a job in a factory in Luoyang, but you’re doing the best you can with the available material.

Joseph Torigian: Robert Caro is my idol. His book on research methodology is perhaps the best guide to research ever written. The only other useful recommendation I have for aspiring researchers is watching old Gordon Liu kung fu movies like 36th Chamber of Shaolin or Fists of the White Lotus. Those training sequences offer a great way to think about learning research skills.

Caro writes about trying to understand LBJ but feeling like he couldn’t get close to his subject. He told his wife they were moving to the hill country in Texas. Only when he experienced how difficult life was there did he feel he understood who LBJ became — seeing the poverty and hardship firsthand. I often wonder how different my book would have been if I had lived in Shaanxi Province for two or three years. Those are the regrets you think about. You can only write the book that’s possible with the sources available to you, but it does make you wonder.

No Saints, Only Martyrs: Xi’s Early Years

Jon Sine: This seems like a good jumping-off point to discuss the book’s chronology. My first question is about getting into the mindset of someone like Xi Zhongxun, who was born in 1913 in a place of unimaginable suffering, death, battle, and tragedy — at least from a modern developed-world perspective. You might be uniquely positioned to answer this, having spent time reading dark literature from the Soviet archives and authors like Solzhenitsyn.

You open the book with Xi Zhongxun attempting to murder a teacher at age 14, which serves as a device to help us understand that this wasn’t particularly strange for the time — people of all ages were dying and being killed. How do you advise readers to get into the mindset of this world when Xi Jinping was growing up?

Joseph Torigian: One regret I have about this book is that people sometimes focus on the most salacious moments, like the one you described. You need to put these events in context. I’m not saying that understanding him means we should forgive him, but we should understand the conditions that led to this behavior. This book is as much about the party system as it is about Xi Jinping.

People like to think of the party as inhabited by good guys and bad guys, but it’s much more complicated. They’re all part of a system, trying to figure out what they can do in limited situations. This needs to be understood when we consider who these people were.

When Xi Zhongxun was 15 years old and attempted to kill this academic administrator, it occurred in a milieu of violence. The Nationalists were killing many of his comrades. This was an act of desperation by a frightened young person — a huge failure that wouldn’t have improved the party’s position even if it had succeeded. Not only did the administrator survive, but Xi Zhongxun and many others were arrested.

This illustrates a pattern Xi Zhongxun faced throughout his life: you need zeal and fearlessness, but too much zeal can lead to actions that go too far and trigger backlash, often making situations much worse. However, if you’re not aggressive enough, others will accuse you of being a rightist — an opportunist too frightened to use violence and risk to advance the revolutionary agenda.

There’s a famous communist text called “Left-Wing” Communism: An Infantile Disorder that recognizes being too radical can get you killed. But they don’t debate these questions rationally because of the ideological charge, which makes management extremely difficult and can lead to purges, arrests, and executions.

Jordan Schneider: The life-or-death stakes and emotional charge of those early years are hard to comprehend.

Joseph Torigian: Those were the old days, the bad days, the all-or-nothing days.

Jordan Schneider: Kids these days aren’t getting horses shot out from under them or taking bullets through the leg while running from Nationalists. But seriously —

Joseph Torigian: You can see how Xi Jinping in the 1980s looked at younger people born under socialism, after all that violence was gone. When they criticized the party, he wanted to tell them what it was really like. I can see how some students would be inspired by that, while others wouldn’t be. It depends on who listens and who is moved.

Jordan Schneider: Let’s stay with the 1920s and 1930s. There was this one guy accused of something serious who said, “I’m going to show you just how committed to the party I am.” He was reckless in battle, got shot, and died. That tension — if you’re redder than red, you might literally end up being killed.

Joseph Torigian: You’re referring to Xi Zhongxun’s first great mentor, Liu Zhidan (刘志丹). What’s fascinating about him is that even though they faced persecution and their cases hadn’t been fully rehabilitated, he would tell Xi Zhongxun that the party would eventually get things right and he should talk to others about this. Then this same person, Liu Zhidan, because he still had a black mark in his file and wanted to show his loyalty, got himself killed.

Jordan Schneider: It’s not like the Catholic Church as we know it today — this is early Christianity. There are martyrs, everything is life and death, the world is against you. To believe and commit your life to it, you need a truly revolutionary vision of how society will change.

Joseph Torigian: When you see people around you getting killed for these reasons, once the regime is established, you can understand wanting to do everything possible to ensure the regime persists not just during your lifetime, but for decades or even centuries into the future. This is something Xi Zhongxun thought about constantly — he would ruminate on how to make sure the party survives not just him, but deep into the future. This makes the fact that his son is now the leader of the country particularly interesting.

Jon Sine: There are a lot of novels in the book. When Xi Zhongxun is young, he read a novel with the same title as your second chapter (which is among my favorites of the 30 chapters, by the way). The novel is called The Young Wanderer 少年漂泊者. Xi Zhongxun sees a lot of himself in the protagonist. But I’m also reminded of Stalin. According to Kotkin’s biography, Stalin was very motivated by a novel he read before he was a committed Marxist-Leninist, which has a similar, though less tragic, heroic figure named Koba. So much so that Stalin uses Koba as a nickname for himself.

I’m curious both if you could explain the Xi Zhongxun example, but also what you think of the fact that so many of these revolutionaries were inspired by something like a novel before they even really had a conception of what they were going to be sacrificing their lives for.

Joseph Torigian: It’s interesting that his life as an early revolutionary was so deeply shaped by this book. In fact, he told Xi Jinping, his son, that he reread this novel after being released from jail to rekindle his love for the party. At the time of his release, he was desperate and couldn’t walk. He was covered in eczema and boils. His father had died, apparently because of the stress related to watching what his son had just gone through. Soon after, he lost his mother and two sisters. There was famine everywhere. He couldn’t link back up with the party because most of it had been destroyed. He couldn’t go back to school because he’s a wanted man. He read this book to get himself back in the mood, so to speak. He rededicated himself because of reading this novel. Then in 1962, he was thrown out of the leadership because of another novel. It’s an interesting bookend.

Then Xi Zhongxun, of course, returned to work in Beijing in the Secretariat. The very first thing that happens is people think that the country is going to fall into another Cultural Revolution because of a movie, a movie called Bitter Love (苦恋). It’s a movie that probably should speak to him in some ways. It’s a movie that is about someone who is deeply dedicated to the party and to the nation, but who the party persecutes. It includes this famous line — “You have always loved the motherland, but did the motherland love you?”

He doesn’t like stuff that criticizes the party, but he still sees the need to reflect on the Mao era, and he’s not quite sure how to criticize it without going too far. But repeatedly, we keep seeing that movies and novels have really fundamental impacts, not just on the party, but on Xi Zhongxun in particular. It speaks to this issue of communism being preoccupied with souls and for communist writers being “engineers of the soul.” And why in the 1980s, Xi Zhongxun, who in many ways was a reformer, would still say things like, “We also need to have a spiritual civilization."

People held different views on how to solve the same problems that Xi Jinping was witnessing. Some believed the solution lay in enlightenment, cultural change, science, and democracy. Mao Zedong disagreed. He thought the right path involved violence, class struggle, and forceful transformation.

In the 1980s, there was a reflection on these two different paths and how the May 4th movement represented one approach while the Yan’an Rectification represented something quite different. The Yan’an Rectification, as I write about in my book, addressed a fascinating problem — all these young people had hero complexes. They were young, entitled, and spoiled, yet nevertheless wanted to save the nation. They went to Yan’an and recognized they needed to become good party soldiers, but transforming into a party soldier proved extremely difficult. It was a wrenching, traumatic experience. However, when they emerged from this process of struggle sessions and self-reflection, they felt like completely new human beings and often experienced euphoria.

Jon Sine: Your book uses Xi Zhongxun as a lens to examine the party. However, for the first twenty years of his life, the party center hadn’t yet consolidated in Shaanxi and Yan’an. This raises a question about the story you’re telling, which relies heavily on Joseph Esherick’s excellent book Accidental Holy Land. How representative do you think this experience in Shaanxi was during these first twenty years of establishing the base area?

Joseph Torigian: The Northwesterners took great pride in the fact that almost all other base areas had been destroyed except theirs. This raises an interesting question. I’ve discussed how Xi Zhongxun was incarcerated in 1935 by his fellow communists and was still in prison when Mao arrived. Mao didn’t release them immediately and wasn’t the primary person involved in investigating. Nevertheless, if that base area hadn’t existed, it would have created a huge problem. Mao didn’t even know it was there. The Party Center only learned about this base area’s existence through a newspaper in some dusty, isolated town, which led them to investigate and eventually arrive. This raises the intriguing question: who saved whom? Did the Party Center save the Northwesterners, or did the Northwesterners save the Party? That situation was unique.

Another factor was the severe infighting in all these base areas, but it was especially bad in Shaanxi for various reasons. No clear leader emerged from the region. One potential leader I’ve already mentioned, Liu Zhidan, was killed for reasons we’ve discussed. Even before him, there was another prominent leader, Xie Zichang 谢子长, who was also killed. They weren’t represented by a single figure who could command the loyalty and affection of the entire group.

When Mao was in Yan’an, he needed to identify a representative who could speak for the Northwesterners’ group and whose promotion would signal to the Northwesterners that they were taken seriously. He chose Gao Gang. Many people disliked Gao Gang — in fact, they hated him. This left behind the potential for struggles that would emerge in subsequent years and continue to affect Xi Zhongxun’s life for decades.

Jon Sine: I was recently in Ruijin in Jiangxi, which some may know as the starting point for the Long March and the capital of that base area. When you’re there, you see the whole place covered with Mao’s little wax statues showing him standing and giving speeches. However, from what I understand of the history, he was actually marginalized for most of the period there after the Central Committee moved down around 1931.

I had a friend who recently went to Yan’an to visit the museum and see what role Xi Zhongxun plays and how he’s memorialized there. His view was the same as mine when I visited — the portrayal seems right-sized, which is pretty small. How would you right-size Xi Zhongxun’s importance, especially during the period before Mao arrived?

Joseph Torigian: He was primarily involved in civilian governance, mostly working with the Soviet structure. While that sounds important, what was even more crucial during this period were the military leaders. In some cases, these military figures who were actually commanding soldiers looked down on the civilians, and Xi Zhongxun was considered as such.

When Liu Zhidan, the commander of these forces, would interact with Xi Zhongxun and show deference to him as someone working for civilian government structures, Xi Zhongxun was quite struck by that respect. After the Party Center arrived, Xi Zhongxun was sent to some real backwaters — dangerous ones — and only gradually, as the party continued to review the 1935 persecution, was Xi Zhongxun placed in increasingly important positions.

What’s striking is that by the Seventh Party Congress in 1945, Xi Zhongxun was the youngest candidate member of the Central Committee. He benefited from Gao Gang’s rise because they were now on the right side of history, according to the view of the past that Mao was imposing on the party. You’re right that I wouldn’t place him as one of the leading figures who helped establish the base area. This was also important for Xi Zhongxun because later, when he was seen as an informal leader of Northwesterners within the leadership, he wasn’t as dominant over the rest of them as figures from other factions were seen as clear standard bearers with much more prestige and status.

Jon Sine: What about the factions that seem to emerge at this time? You have the Long Marchers, whom Mao derisively said something about — that just because you walked a lot doesn’t mean you have special status. Obviously, many of them certainly thought they did. Could you speak about the tension between the Long Marchers arriving and the people in the Shaanxi base area?

Joseph Torigian: The Long Marchers were the central leadership — the Central Committee. Many were very well educated with extensive experience. Many had studied in Moscow. Then they showed up in the Northwest and thought the locals were bumpkins. Yang Shangkun even said, “You consider this a city? I can’t believe you think this is an urban area given how undeveloped it is."

The Northwesterners had built the base area, and then the Long Marchers arrived and basically took over. People like Gao Gang and Xi Zhongxun continued to face the shadow of the 1935 purge — they weren’t immediately rehabilitated. The Long Marchers looked down on them, while the Northwesterners felt they had built this base area and were upset about the lack of respect they received.

Mao Zedong adopted the Northwesterners in quite striking ways. He gave them so many positions at the 1945 Seventh Party Congress that Mao’s associates from the Jiangxi base area became very jealous and quite unhappy. The story Mao told was essentially this: when he was in Jiangxi, dogmatic people who had learned in the Soviet Union told him to do wrong things and pushed him out through domineering behavior. He said that’s basically what happened in 1935 when people like Gao Gang and Xi Zhongxun were persecuted — because they were doing good, real, sinicized revolution and were persecuted by people who had no idea what they were doing.

For Gao Gang and Xi Zhongxun, this must have been thrilling. Here was Mao saying flattering things about the Northwesterners, claiming their historical experiences mirrored his, and promoting them rapidly at the expense of his old associates.

Then something interesting happened. In the early 1950s, Mao Zedong started complaining about Liu Shaoqi. He talked to Gao Gang, and Gao Gang harbored antagonism against party cadres who had spent most of their time in the “white areas” — meaning the cities doing underground work. He didn’t understand why they had all these cushy positions when it was the “red areas” — the base areas — that had done the bulk of work during the revolution. He also thought the Northwesterners needed more respect as reflected in official party history. He was motivated by how the party evaluated their contribution to the revolution.

Part of Gao’s machinations against Liu Shaoqi were based on Mao’s encouragement and certain things Mao was saying about Liu Shaoqi. But Gao also had this view of history that explains why he was acting dangerously. Xi Zhongxun was much more cautious. He said they wouldn’t be respected because they weren’t Long Marchers — they didn’t have that credibility, status, and prestige. Xi Zhongxun felt that people like Deng Xiaoping, who was a Long Marcher and knew Mao for a long time, were people he and Gao needed to be careful of. Ultimately, it was Deng Xiaoping who betrayed Gao Gang, went to Mao, and signaled Gao’s defeat.

The Party as Mother and Matchmaker

Jordan Schneider: We have this elite drama as well as human drama. There’s love, there’s death. What’s his relationship to his parents and his first wife?

Joseph Torigian: Xi Zhongxun’s parents died when he was quite young — fifteen or sixteen years old — leaving him an orphan for a while. After he was released from prison, he had to take care of his entire family. Then he left them to join the revolution, which led to an unflattering comment from an uncle. On one occasion, he actually referred to the party as “mother” — I don’t have that in the book, but it’s an interesting word choice. By fifteen or sixteen, the party was essentially his only family, or at least the most important family he had. He went back to his home village once or twice, but the entire emotional center of gravity of his life was the party, not his immediate relatives.

His first wife, Hao Mingzhu 郝明珠, was a very interesting figure. She came from a revolutionary family and was very capable. She did extensive work trying to explain to women why the revolution was something they needed to take seriously and why they shouldn’t be frightened by it. She became pregnant several times, but she hadn’t joined the revolution to get pregnant — she joined to make revolution. That tension apparently led to the break between them.

Jon Sine: Can we pause on his first wife for a second? One of the things that stuck with me most from your book—almost haunting—was a one-sentence line about her first husband who was killed and beheaded. She had to retrieve his body, which was severed from the head, and bury him. She was eighteen at the time. People should sit with that for a second to imagine what kind of normal setting this was, where this woman had to endure that.

Joseph Torigian: She’s also interesting because she went on to have a pretty good career in a very male-dominated institution like the Chinese Communist Party. She became a powerful individual in Shaanxi in subsequent decades.

Jon Sine: Despite this very grown-up tragedy she suffered at eighteen, you offer a very strange description — which I presume is completely factual — of the way Xi Zhongxun courts her.

Joseph Torigian: The stories in the book are funny from our perspective. They’re frightening and sad at the same time.

Jon Sine: I don’t know if you want to give that story, because it has all these elements — they’re twenty, adolescents, but also patriarchal, yet also progressive in some weird way.

Joseph Torigian: As I mentioned, one interesting thing that emerges from the book is how often the party’s own ideals and mission would bump up against something else — exigencies and political needs. You would have matchmaking sometimes to pursue political goals, even as you’re telling women not to bind their feet. It was quite a striking contradiction.

Jordan Schneider: Tell the story.

Jon Sine: I have it off the top of my head. She’s in this group, they basically show up, and they all start pointing to Xi Zhongxun and say…

Joseph Torigian: “This is the guy.”

Jon Sine: She has a shy girl reaction, runs away, and they chase her. Maybe you could pick it up from there — how that eventually ends rather swiftly in their marriage.

Joseph Torigian: There was extensive, forceful party matchmaking during this period. To put it in context, this was an organization built to achieve revolution, and these relationships were seen as something that should be built on principles of what facilitates the revolution. That led to pressure on people to marry people they didn’t know well or might not have even liked.

Jordan Schneider: How does getting married help revolution in the first place?

Joseph Torigian: To move on to the relationship between Xi Zhongxun and Qi Xin 齐心, it’s very striking how much he doesn’t want to be distracted by her, especially during dicey moments for the party, as the Nationalists were tearing up and breaking up much of what had been the Yan’an base area in the second half of the 1940s. When he meets Qi Xin, he tells her, “Why are you coming to see me? We’re in the middle of fighting this war. It’s just a distraction.”

The party was supposed to break down personalistic relationships so people wouldn’t put romance first — they put revolution first. This was something they all understood and intuited. I can’t remember if I left it in the book, but there was one time when Xi Zhongxun facilitated a meeting between Peng Dehuai and Peng Dehuai’s wife, and Peng yelled at him, saying he needed to focus on bigger things than setting up an occasion for him to see the woman he was married to.

Jordan Schneider: This reminded me of arranged marriages where the woman meets this man who’s in good standing with the party, and they get married the next day — like ISIS. ISIS would arrange marriages with women they captured as spoils. On one hand, Joseph, these guys like sex, which you bring up, but at the same time, you have these women who are incredibly conscious and want to contribute to the party. They’re also annoyed by the fact that they have to do all the childbearing. We end up in this weird dynamic where, because Xi Jinping’s mother is a very active person, they just don’t do parenting for six days a week. It shows how dedicated and religious they are — the family stuff and sex are there because they’re human beings, but they’re subordinating this to the broader goal of making revolution and having the party flourish.

Joseph Torigian: Some more context that’s not directly relevant to what you’re saying, but that we saw very commonly in Yan’an, is that the old revolutionaries had wives they had married years before getting to the base area. Many of those wives had their own seniority, prestige, and status because of when they had joined the revolution and what they had done for it. But then these men left their wives for younger party cadres — women often from more privileged backgrounds with more education. Xi Zhongxun’s two marriages were something quite common in Yan’an — they divorced their first wives and married younger wives. Sometimes that was quite controversial.

For example, in Yan’an, Mao was married to a woman who had contributed and sacrificed a lot for the revolution, and he married another person who was an actor from Shanghai who had been in movies and whose life had been affected by rumors about her love life. When Mao started courting her, some people were unhappy about that. This was another element of the social fabric of what was happening in Yan’an during this period.

Jordan Schneider: It’s funny because Xi Jinping almost goes in the other direction where the first wife is too educated and too into the Western world, ends up moving to the UK and living out a global cosmopolitan lifestyle. Then he finds this woman whose background and worldview much more echo what he had from the perspective of coming into the party and forging and suffering.

But we shouldn’t jump that far ahead. Jon, bring us back.

Jon Sine: On the topic of Xi Zhongxun maybe not being that high-ranking at the time, in the late thirties and forties, he’s maybe just an alternate Central Committee member. He’s the head of a county called Suide 绥德, a sub-region of the Northwest.

Joseph Torigian: It was northeast of Yan’an. The first big jump of his life was really in 1945 when he became a candidate member of the Central Committee — he was very young. Then the next huge jump was when he became head of the Northwest Bureau that same year. Peng Dehuai and He Long (贺龙) were also very important figures in the Northwest, but for someone that young to have a position that weighty is one of the more striking parts of his story. He repeatedly was the youngest person in these very delicate positions. He kept winning over older people to support him. He had this habit of collecting big brothers — that’s something that marked his life.

Jon Sine: I wanted to talk about the Suide period when he was there. You’re pulling out these threads of contradictions and tensions that exist within this individual, as opposed to other biographies that might have more of a clear arc for narrative reasons. But it’s a through line — is he a reformer, is he conservative? These tensions run through land reform there, they run through political rectification that he goes through. Sometimes he seems to tone it down, but sometimes he also seems to be turning it up, which just actually means purging, potentially killing, leading to the deaths of people. How do you see that period, if we linger on that, as a harbinger of some of these contradictions?

Joseph Torigian: During the rectification in 1942, at this period, it focused on transformation and education. Xi Zhongxun was the dean of a university in Yan’an for a few months. Then in 1943, he went to the Suide sub-region, which you just mentioned. Their rectification turned into something called the Confession Movement or the Rescue the Fallen Campaign. Now it was looking for spies — it wasn’t just about taking these young people and turning them into good communists, but finding which ones were spies so they could be rescued. That required a whole new level of investigation, persecution, sometimes even torture, and it completely went off the rails.

Xi Zhongxun had a reputation for being more humane and pro-reform, but the Confession Campaign — the Rescue the Fallen Campaign — was really a case of political radicalism that Xi Zhongxun seems to have been pushing and we should ask why that was the case. In some ways, it’s a puzzle because just a few years earlier, he himself had been persecuted largely on trumped-up charges that resulted from people being forced to give false confessions. He knew this would happen.

Part of it was that the party told him to find spies, and of course he was going to find spies. Part of it was war hysteria — there was fear in 1943 that the Nationalists were going to invade. Also, the Communists had their own spies, making it easy to imagine that the Nationalists were sending agents into this region.

Xi Zhongxun had reasons to be loyal to and grateful to Mao for what we just described — Mao’s new version of history that must have been thrilling to the Northwesterners, crediting them as being on the right side of history along with Mao Zedong.

But Xi went very far, and the form these campaigns took in Suide were very interesting because they focused on a school and included lurid details about many of the young women allegedly being spies sent by Nationalists to infiltrate the party by essentially being honey traps.

Jordan Schneider: To be clear, by young, we mean one of the women who ended up getting caught up in this purge was fourteen. She’s supposedly some prostitute for the Nationalists who can barely write her name.

Joseph Torigian: They had mass rallies where these young people went up and lied about having been spies and having been sent on these missions. What’s interesting is that the target of these campaigns in Suide was this school, and Qi Xin was a student at this school. That’s when Qi Xin and Xi Zhongxun first started to get to know each other. He showed off this inscription that Mao gave him that said, “The party’s interests come first” — something she never forgot.

Jordan Schneider: Yeah. Relationships that start in high school purges tend to be the ones that really see you through the decades.

Joseph Torigian: You can see it was an electric atmosphere. That’s one way of putting it. But he also told her to be careful, otherwise she would be caught up in it. On one occasion, she visits one of the students being persecuted and even interviews her. It was a messy scene.

Post-1945: Surrogate Fathers and Succession

Jon Sine: We can move forward to the part where the Japanese are defeated in 1945. At some point, maybe two or three years into the Civil War, the Nationalists chase the Communist leadership out of Yan’an. What happens to Xi Zhongxun at this point?

Joseph Torigian: 1947 was a really interesting moment. Yan’an had been the Communist capital for many years and was a symbol of the Chinese Communist Party. Then the Nationalist General Hu Zongnan (胡宗南) attacked, and Xi Zhongxun was on the front line. He screwed up a really key battle, and the party was facing a real crisis. Peng Dehuai, a very famous military figure in Chinese history, was plucked from the Central Military Commission and ordered to command the Northwest forces. Xi Zhongxun worked for him. In the subsequent battles, it looks like the two of them basically got along.

Later in the 1940s, Xi Zhongxun also spent time on governance, civilian work, logistics, and land reform, where he worked under He Long 贺龙. On the military side, he worked under Peng Dehuai, and on the civilian side, he worked under He Long. These were very weighty positions, especially for someone young.

I want to dwell on this moment for just another second — how emotional it must have been for Xi Zhongxun in March 1947. He’s in Yan’an, the Nationalists are coming, and Mao refuses to leave. Mao was really frightening the people around him by staying in the city even as the Nationalists were at the very gates. It was Peng Dehuai and Xi Zhongxun who finally saw him off and then tried to create this rearguard action to hold off the Nationalists to ensure the Central Committee could escape. To have spent that much time building up Communist forces in this area and then see them destroyed by the Nationalists must have been another very emotional moment in his life.

Jon Sine: Xi Zhongxun and Peng Dehuai, serving together, would be sleeping in the same room. Apparently Xi Zhongxun was quite the snorer. You write that Peng Dehuai would sometimes wake up in the middle of the night and put his shoe on Xi Zhongxun’s face to wake him up and stop him from snoring.

Joseph Torigian: Comrades-in-arms.

Jordan Schneider: Dude, that’s sleep apnea. That’s serious.

Joseph Torigian: Peng Dehuai was famously mean to people under him, and Xi Zhongxun sometimes acted as a peacemaker, a mediator between Peng Dehuai and whoever Peng was mad at. Xi Zhongxun’s ability to manage interpersonal relationships was emerging here. Even though Peng could really yell at people and get agitated, he also had this communist egalitarian ethos. In that sense, it’s not surprising that he would be sleeping right next to Xi Zhongxun and waking him up with his shoes. It was this interesting mix of hierarchy and egalitarianism that we see in Peng Dehuai.

Jordan Schneider: He does have this LBJ energy to him.

Joseph Torigian: It’s interesting — a few years later, there were tensions between Xi Zhongxun and Deng Xiaoping over Tibet. Xi Zhongxun was very close to Peng Dehuai and one of his protégés. Xi Zhongxun went to complain to Peng about something, and Peng said to him, “You need to be careful because Peng likes to fire cannons.” Then of course, Peng fired a cannon in 1959 at the Lushan Plenum that led to people doubling down on the Great Leap Forward in tragic ways. Xi Zhongxun could read people pretty well sometimes.

Jon Sine: People who have heard of Peng Dehuai, maybe in the China field, probably remember him from the Great Leap Forward, which we will get into. But it speaks to this — at least to me—because my conception of him is drawn from this idea of good versus bad guy. He was the good guy who tried to stop the Great Leap Forward, and people read this back into history. But when you go through the details, you realize he was a very complex person who would be very angry. Everyone at this time was a killer.

Joseph Torigian: Peng Dehuai, too, if you read this wonderful book by Yen-Lin Chung, which is a biography of Deng Xiaoping before the Cultural Revolution. Shortly before Peng Dehuai was targeted in 1959, there had been a campaign of anti-dogmatism in the People’s Liberation Army. Peng was just brutal to people. He especially went after this very famous general named Su Yu 粟裕. Peng was a victim, but very shortly before that, he was a victimizer — another very complicated figure.

Jordan Schneider: The LBJ thing I was going for was this cultivating of older men and having these surrogate father figures.

Joseph Torigian: I could give a whole list for Xi Zhongxun. Even later in his life, he was still making new friends older than him, like Ye Jianying 叶剑英, a man with whom he had no historic ties, no career ties. Suddenly, Ye Jianying took a shine to him, sent him to Guangdong, and became a patron to him. It’s very interesting. Apparently in 1959, the person who nominated Xi Zhongxun to become a vice premier was Chen Yi 陈毅, another very powerful marshal who apparently only really worked with Xi Zhongxun on the State Council for a few years in the 1950s but seems to have liked him. It’s really interesting.

There are reasons this might have been the case. Xi Zhongxun could be charming. These older men who liked to drink — Xi Zhongxun could roll with them in a lifestyle sense. He was good at his job, good at intuiting what people wanted, and could manage personal relations. That’s why he did United Front work. He was a Northwesterner — he wasn’t a Long Marcher. He had been orphaned, essentially, when Liu Zhidan, Xie Zichang, and then Gao Gang were gone. He was looking for big brothers. He needed big brothers. Maybe these big brothers didn’t really feel threatened by him because the Northwesterners as a group weren’t really that frightening.

Jordan Schneider: He was young, too.

Joseph Torigian: They were thinking about succession — how are we going to hand over the reins to somebody who knows what they’re doing? We need to test these people and give them an opportunity to show whether they’re good at it and gain experience. He was the youngest vice premier in 1959, the youngest candidate member of the Central Committee in 1945, and the youngest major figure from a regional bureau brought to Beijing in 1952. Then, when he went back to work in the 1980s, he never quite reached the top. In fact, he was sent to work for Hu Yaobang, who was much younger than him and had much less revolutionary prestige. But perhaps we’re getting ahead of ourselves.

Stay tuned for part two, where we’ll pick up the story in 1949.

Today we’re running a guest translation by the great of the Sinification substack which translates leading Chinese thinkers.

Below, Liu Shaoshan — a leading figure in China’s embodied AI research with a PhD from UC Irvine and an official state-designated “high-end overseas talent”— proposes a roadmap for Chinese AI dominance cueing off America’s successful diffusion of TCP/IP protocols in the late 20th century. Just as influence over the internet afforded the USA “a truly global mechanism of discursive control,” Liu argues, AI diffusion and the standards exported along with AI systems will be key to power projection in the 21st century.

He outlines four strategic levers for achieving open source dominance — technological competitiveness, open-source ecosystem development, international standard-setting, and talent internationalisation. Acknowledging that China’s engagement in open source AI still has a long way to go, he advocates for the creation of a comprehensive “China HuggingFace” that maximizes market share by publishing toolkits for model training, embodied AI implementation, and everything in between. Finally, the author argues that Beijing should encourage Chinese AI talent to live and work abroad, especially in Belt and Road participant countries, rather than encouraging them to come back to China.

This piece is particularly resonant at a time when leading White House AI advisors are tweeting stuff like this:

For the policy answer to the challenges Liu raises, check out Nathan Lambert’s call for action to replicate Deepseek in America.

Interconnects

The American DeepSeek Project

While America has the best AI models in Gemini, Claude, o3, etc. and the best infrastructure with Nvidia it’s rapidly losing its influence over the future directions of AI that unfold in the open-source and academic communities. Chinese organizations are releasing the most notable…

Read more

12 days ago · 80 likes · 1 comment · Nathan Lambert

1. US tariffs and export controls heighten global uncertainty but create a strategic opening for China’s AI industry to expand internationally and reshape “the global technological order”.

2. Global adoption of US or Chinese technology — not domestic technological prowess alone — is becoming the key battleground for great-power status in AI.

3. America’s success with TCP/IP’s global rise in the 1980s shows centrally-led government policies, open-source, mandatory standards and talent “exports” can turn national tech into the global default.

4. Rogers’ diffusion model suggests four steps for China: woo “innovators” with cutting-edge tech, attract “early adopters” through open-source, secure an “early majority” by setting international standards, and reach late adopters through Chinese talent going global.

5. Thus, China’s first objective should be to match US-level capabilities so that its AI-related technologies are credible and attractive to global “innovators” and “early adopters”.

6. China’s DeepSeek-R1 and other LLMs now rival OpenAI in maths, coding and/or reasoning, demonstrating their technical credibility for such global uptake.

5. Moreover, China’s embodied AI sector shows strong international competitiveness, with robust upstream manufacturing, rapidly improving midstream technologies and world-leading pilot deployments, forming a highly competitive end-to-end ecosystem that supports rapid technological advancement.

6. Nevertheless, China still trails the US in the maturity of its open-source community reach and its influence over international standards systems.

7. Recommended actions:

   1. Back talent-going-abroad schemes that help place Chinese experts in key positions in emerging markets such as in universities, labs and start-ups, thereby supporting the spread of Chinese technologies and standards internationally.

   2. Establish an “Open-Source Co-construction Fund” to boost China's influence in global technical governance and standards-setting.

   3. Shift from ISO-centric competition to a “code-as-standard” strategy: export products pre-loaded with open-source standards, ensuring that adoption effectively becomes standardisation.

   4. Promote collaborative hardware-software stacks that embed Chinese standards directly in already established code repositories like GitHub and Hugging Face, thereby easing global adoption.

   5. Build a “Chinese Hugging Face”: an integrated and globally influential open-source hub for models, middleware and applications, covering the full pipeline from model development to deployment.

Name: Liu Shaoshan (刘少山)
Year of birth: est. 1984 (age: 40/41)
Position: Director of the Embodied Intelligence Centre at the Shenzhen Institute of Artificial Intelligence and Robotics (AIRS); Founder and CEO of intelligent robotics company PerceptIn.
Previously: Technology Leadership Panel Advisory Group Member, National Academy of Public Administration (2023-4); Senior Autonomous Driving Architect, Baidu USA (2014-6); Senior Software Engineer, LinkedIn (2013-4); Software Development Engineer, Microsoft (2010-3)
Other: IEEE Senior Member; Recognised as a “national high-level overseas talent” (国家高层次海外人才) by the Chinese government (under its broader Overseas High-Level Talent Recruitment Programme, historically known as the Thousand Talents Plan)
Research focus: Embodied AI; Autonomous driving; Computing systems; Technology policy
Education: BSc, MSc and PhD (2010), UC Irvine; MPA, Harvard University.

As we moved into 2025, the Trump administration reintroduced tariffs across the board [统一关税] on global goods while simultaneously implementing stricter controls on technology exports, delivering a dual blow to global trade and technological ecosystems. This latest round of trade protectionism and technological blockade policies has significantly increased systemic uncertainty across the global economy and technology sectors. In March 2025, the Organisation for Economic Co-operation and Development (OECD) revised its forecast for global growth in 2025 from 3.3% down to 3.1%, and further downgraded it to 2.9% in June, citing “trade policy uncertainty” and “structural barriers” as factors dragging down global investment and supply chain stability. The United Nations Conference on Trade and Development (UNCTAD) also warned that escalating trade tensions and policy volatility could slow global growth to 2.3%, leading to stagnation in both investment and innovation. Against this backdrop, a strategic window of opportunity has opened for Chinese AI to “go global”. As the United States increases obstacles to trade and to the sharing and export of technology [技术壁垒], China should participate actively in the restructuring of the global technological ecosystem. It can do this by exporting its technology, building an open-source ecosystem, setting standards [标准制定] and encouraging the global mobility of Chinese AI talent [人才国际流动]. This would then mark a shift from a passive posture to a proactive strategic approach. This article offers a systematic analysis of the strategic pathways and concrete steps that China should take across these four key dimensions.

On 8 May 2025, during a hearing of the US Senate Committee on Commerce titled “Winning the AI Race”, Microsoft President Brad Smith issued a warning: “The number-one factor that will define whether the US or China wins this race is whose technology is most broadly adopted in the rest of the world.” His statement underscored a fundamental shift in the strategic landscape: against the backdrop of Trump's increasingly stringent technology and global trade policies, the extent to which [a country’s technology] is adopted globally has become the key determinant of [that country’s] great power status in AI. It is precisely within this climate of mounting restrictions that China’s AI industry has encountered a window of opportunity to “go global” and reshape the international technological ecosystem.

Looking back at the 1960s to 1980s, the United States successfully capitalised on the opportunity offered by its technological ascendancy in IT. First, in terms of technology export, in 1969 the US Defence Advanced Research Projects Agency (DARPA) launched ARPANET [a forerunner to the Internet], and by 1980 had incorporated TCP/IP [its new communications protocol] into its defence communications system. On 1 January 1983, a full network-wide transition to TCP/IP was completed, laying the foundation for the global internet. Next came the construction of the open-source ecosystem: DARPA’s TCP/IP protocols were incorporated into the open-source BSD system, thereby initiating an “open-source means of dissemination” [开源即传播] model. In 1986, the NSFNET project extended this model to the academic community. This made networking functions readily accessible at the operating system level and further stimulated broad participation from the research and developer communities. It was precisely this “code usable is code used” [代码可用即可用] design philosophy that accelerated the internet’s diffusion from the laboratory to commercial and civilian domains. The evidence demonstrates [实践证明] that only when core communication protocols move from closed-loop research to open-source community development can exponential technological breakthroughs occur — moving from source code to [forming the core of] global infrastructure [从源代码走向全球基础设施].

In addition to its advanced technology research and development (R&D) and open-source contributions, the United States also achieved widespread adoption of TCP/IP by standardising outputs and aligning government policy with open-source practice to establish a globally compatible basic infrastructure [全球兼容基础]. In March 1982, the US Department of Defence officially designated TCP/IP as the standard for military communications and announced a nationwide transition scheduled for 1 January 1983 — the so-called “flag day” policy. This was not merely a technical upgrade, but a form of “compatibility mandate”: all hosts connected to the network were required to support TCP/IP, or else they would be disconnected on the day of the switch. This compulsory standardisation not only enforced synchronised upgrades across the US military-industrial and research systems but also fostered a nationwide consensus on protocol standards.

More importantly, this standards rollout did not take place in a closed-off system [输出非封闭]. Rather, it advanced in tandem with the open-source ecosystem: DARPA awarded contracts in phases to institutions such as BBN, Stanford and Berkeley to develop TCP/IP implementations for major platforms including Unix BSD, IBM systems and VAX, subsequently incorporating the code into the 4.2BSD version of Unix for public release. In 1986, the NSFNET project further accelerated [进一步推动] the widespread deployment of this protocol across the national academic network, effectively achieving near 100% coverage.

This series of measures served collectively as the blueprint [机制范本] for the internationalisation of a US-developed communications standard. The government played a leading role by setting a compatibility timeline to guide synchronised upgrades to [this new] standard. Open-source practices facilitated multi-platform availability, enabling “use on demand” by research institutions and enterprises. The infrastructure for global system compatibility was deployed in parallel with the release of open-source code. This strategy — combining policy, open-source and platform integration — not only accelerated the adoption of the standard [缩短了标准推广路径], but also rapidly established TCP/IP as the default protocol for international communication.

From the 1960s to 1980s, the United States established its leading role through the export of technical standards. Of even greater long-term significance was [its strategy of] sending IT talent abroad [人才走出去], which profoundly influenced the global internet architecture. This experience offers valuable lessons for Chinese AI’s [ambition to] expand overseas [中国AI出海]. America’s engineers and researchers didn’t close themselves off from the world [封闭体系]— large numbers participated actively in international standards organisations and met with different parts of the academic community [社区会议]. For example, the International Network Working Group (INWG), founded in 1972 by American scholars such as Vint Cerf and Steve Crocker, played a key role in designing global network protocols and laid the groundwork for the birth of TCP/IP.

Subsequently, the Internet Engineering Task Force (IETF), established in 1986, held its first meeting in San Francisco [Note: its first meeting was held in San Diego, not San Francisco] with 21 American researchers and received government funding and support. These platforms became key frontlines through which American researchers exercised technological discourse power [成为美国人才传递技术话语权的前沿阵地]. American experts held core technical and leadership roles within the INWG, the IETF and its parent body, the Internet Architecture Board (IAB) [ — which together oversaw the development of the internet’s technical architecture]. Figures such as Vint Cerf, Jon Postel, and David Clark participated in meetings for many years, published RFC documents, and oversaw the registration of technical parameters and the standardisation process. These efforts not only ensured the professionalism of the adopted standards, but also reinforced the central role of the United States within global internet governance.

More importantly, the institutional design adopted by these organisations was one of open collaboration, enabling engineers from around the world to influence the direction of standards through voluntary participation. By leveraging their first-mover advantage and influence within the research community [社区影响力], American researchers led the development of key standards. In doing so, the US not only exported the technology itself, but also [American] internet culture, projecting its governance discourse power [overseas]. Furthermore, this strategy of internationalising talent meant that the US exported not just protocols, but also the capability to shape their evolution and the associated rule-making process — thereby establishing a truly global mechanism of discourse control over the Internet.

Technology diffusion is a systemic endeavour [系统工程]. In this section, I will draw on the theory of “innovation diffusion” to analyse the historical experience of US internet technology diffusion, offering systematic strategic recommendations for how Chinese AI can go global.

The theory of “innovation diffusion” was introduced by Everett Rogers in his 1962 book Diffusion of Innovations. Its main aim is to explain how new technologies and ideas spread and are adopted within a social system. This process does not happen instantaneously but unfolds gradually over time, spreading slowly between different groups in society through various channels.

The diffusion of innovation is the result of the interaction of multiple factors, including the innovation itself, the channels through which it spreads, the social environment and time. In his theory, Rogers proposed that members of society vary significantly in their willingness to adopt new ideas and can be categorised into five groups: “innovators”, “early adopters”, “early majority”, “late majority” and “laggards”. A technology’s influence gradually accumulates [缓慢累积影响] until it begins to be adopted by the “early majority”. Once it reaches this point of critical mass, its adoption accelerates rapidly. By observing the adoption process of key actors and leveraging their demonstration effect [借助他们的示范效应], an innovation can move beyond niche uptake and achieve widespread acceptance.

In fact, this theory emphasises the evolutionary process whereby an innovation adapts continuously to the needs of [different] social groups [创新自身不断适应群体需求]. It not only explains why some innovations enter the mainstream successfully while others stall due to lack of trust from “latecomers”, but also highlights the critical role of communication strategies, social networks and opinion leaders in the diffusion process.

In short, whether an innovation can truly achieve widespread adoption often depends on its ability to bridge the gap between marginal experimentation and mass adoption. This theory offers important insights for understanding technology export and international uptake: true diffusion is not a one-off success, but a gradual process of acceptance.

Step One: attract [the interest of] “innovators” through technological sophistication [通过技术先进性获取“先驱者”]. Innovators focus on the breakthrough advantages of the technology itself. Only by possessing technological capabilities [技术实力] on a par with the global frontrunners—which means, for example, having Large language models (LLMs) approaching the performance of OpenAI’s, or surpassing them in reasoning ability—can Chinese AI technologies be deemed worthy of adoption by [these early] global users.

Step Two: attract early adopters through open-sourcing. Early adopters are not only interested in how cutting-edge a technology is but also in usability and real-world application scenarios. During the internet era, the United States fostered an early ecosystem by open-sourcing the TCP/IP protocol, enabling the global research and engineering communities to use and experiment with it directly. If China, at this stage, open-sources large model code, interface documentation, inference APIs and the like, it would enable overseas developers to adopt and test these models quickly in local environments, thereby building recognition and participation in an international community.

Step Three: Reach the “early majority” by promoting global standards. These users place greater value on whether a technology has become an industry standard and whether it is widely compatible. The United States established global compatibility and interoperability by implementing a “flag-day” — when all connected systems automatically switched over to the new TCP/IP standard — thereby establishing it [统一] as the standard communication protocol. If Chinese AI could similarly embed interface protocols, model formats and evaluation metrics into international standard frameworks, it would help bridge the gap between the peripheral [early users] and the mainstream, thus accelerating widespread adoption.

Finally: [China needs to] gain the remaining user groups by exporting talent [人才输出]. This stage involves reaching the “late majority” and the “laggards”, and relies heavily on “talent going global” [人才走出去]. This is not limited to technical personnel going abroad to implement projects. A more important part is the active participation of Chinese AI engineers, researchers and policymakers in international standards bodies, community conferences and open-source projects—writing forum articles, submitting technical specifications and serving as working group leads. Just as Vint Cerf and Jon Postel once brought American protocol concepts into the IAB/IETF [which set technical standards for the internet], shaping global rules through RFCs and governance, China must also foster deep expert-level international engagement in order to shape norms at the conceptual level and establish its technologies as a credible global standard.

Based on the innovation diffusion theory discussed in the previous section, this next part explores whether Chinese AI technology is cutting-edge enough to “go global” [是否具备“走出去”的领先性], which is fundamental for technology diffusion.

First and foremost, DeepSeek has already clearly demonstrated its ability to compete alongside the world’s leading foundational models. Multiple independent benchmark tests show that DeepSeek-R1 performs comparably to OpenAI’s o1 model in mathematics, programming and reasoning tasks. In some tests, such as the MATH-500 question set, it even slightly outperforms it. [Additionally, based on DeepSeek’s own research], it also outperforms on the SWE [coding] question set. Its reasoning ability has been widely recognised by its community of users. Although DeepSeek-R1 falls slightly short of OpenAI’s o1 on some complex reasoning problems, its overall performance remains at a high level. As the first open-source large model with comparable capabilities — and with its code freely available and easily accessible — it is quickly gaining traction globally among developers. Even more important are its cost advantages and open approach. By using a mixture-of-experts architecture and reinforcement learning techniques, DeepSeek significantly reduces training and inference costs, achieving international top-tier performance not only in capability but also in efficiency.

Second, in the field of embodied AI enabled by foundational models, China has demonstrated clear global competitive advantages. The supply chain for embodied intelligence encompasses upstream core components, midstream system technologies—including foundation models and computational power—and downstream application scenarios. On the upstream side, the rise of the new energy smart vehicle industry has propelled China’s component manufacturing to a high level of localisation and mass production capacity, spanning key elements from sensors and LiDAR to servo motors. This supply chain advantage provides a solid foundation for the large-scale deployment of embodied intelligence systems.

Turning to midstream system technologies, as previously mentioned, products like DeepSeek are at the global cutting edge. In the area of computing chips, companies such as Huawei and Rockchip continue to make significant progress. Although they are still trailing behind their American counterparts for now, they are closing the gap rapidly.

In downstream application scenarios, China currently leads the world—particularly in large-scale pilot deployments of robotics projects initiated by local governments. A notable example is the “robotics district” in Longgang, Shenzhen. Robots are being deployed in urban management, manufacturing, logistics, elderly care and other sectors. Governments in cities like Shanghai and Shenzhen are providing subsidies and development platforms to support the implementation of these projects.

Overall, China has established a comprehensive closed-loop [完整闭环] AI ecosystem spanning upstream, midstream and downstream segments. Although midstream technologies still lag slightly behind, the gap is narrowing rapidly. This industrial structure enables Chinese products to be competitive in international markets and provides a solid foundation for the future diffusion of its AI technologies.

The previous section concluded that China’s AI technology is sufficiently cutting-edge to “go global”. This next section will examine the current state of China’s AI open-source ecosystem—specifically, whether the open-source infrastructure is sufficiently prepared to attract the world’s “early adopters”.

So far, China has achieved certain foundational milestones [取得初步成就] in building an open-source ecosystem for AI and embodied intelligence. However, it still lags significantly behind the United States in terms of global reach, community governance and platform influence—factors that constrain the depth and breadth of its technology’s international expansion. In terms of LLMs, more than ten Chinese open-source models with over 100 billion parameters—such as DeepSeek—have been released. Some of these models have matched or even surpassed OpenAI’s o1-level performance in MATH and SWE-bench, demonstrating strong technological momentum.

By contrast, the US continues to dominate the open-source ecosystem. The LLaMA series of models released by Meta have fostered a broad and robust ecosystem. The LLaMA2 series alone has amassed more downloads and citations on the Hugging Face platform than all Chinese open-source models combined [Note: the time period for this comparison is not specified]. Moreover, the US has established comprehensive mechanisms for model publication, data annotation, benchmark testing and community engagement — particularly through platforms like Hugging Face — affording its models the status of “default standard” within the open-source community.

In the field of embodied AI, China has leveraged its service robotics and smart vehicle industries to establish leading advantages in sensor technology, full-machine platforms and scenario integration. For example, the Shenzhen Institute of Artificial Intelligence and Robotics has released the AIRSHIP, AIRSPEED and AIRSTONE series of open-source projects, which together form an end-to-end open-source system connecting the model, computing and application [Note: Liu is involved in the first two of these projects]. However, given the relatively short time this ecosystem has had to develop, there remains a significant gap in core middleware and the broader development ecosystem when compared with the US-led ROS (Robot Operating System) framework.

ROS, [an open-source framework for building robot software,] was released in 2007 by the US-based Willow Garage, and its core developers remain concentrated primarily in the United States and Europe. It boasts a clear contributor structure and abundant training resources. By contrast, domestic Chinese alternatives to ROS—such as CyberRT and XBot—have achieved partial system compatibility. However, their visibility on GitHub [a code repository], global developer participation and completeness of development documentation remain low. A strong and influential open-source community around these projects has yet to emerge.

Based on this, as China promotes the upgrade and international expansion of its open-source ecosystem, [my] recommendation is to focus on the following three areas:

1. A “Chinese Hugging Face” with global influence should be established. That is, an integrated open-source platform that unifies foundational models, system middleware and application scenarios, forming a complete pipeline from model training to deployment.

2. The Chinese government could set up an “Open-Source Co-construction Fund” to support universities, research institutes and enterprises in participating in technical governance and standards-setting within international organisations such as IEEE, ACM, RFC and IETF, thereby amplifying China's voice on the global stage.

3. Finally, leveraging China’s current supply chain advantages, efforts should be made to promote “collaborative open-source” for embodied intelligence hardware and software platforms, establishing a standardised open-source stack across hardware and software to truly realise the global deployment and ecosystem integration [potential] of Chinese technologies.

The previous section concluded that China’s AI open-source ecosystem still has significant room for development. In this section, I will examine the current state of the international standardisation of China’s AI technologies, with a particular focus on whether Chinese technologies have formed internationally recognised standard systems capable of attracting the global “early majority” of users.

China’s participation in international standardisation in the fields of AI and embodied intelligence is gradually increasing, but its overall influence still lags significantly behind that of the United States. In recent years, China has actively engaged with international standards organisations such as ISO/IEC JTC 1, IEEE, and ITU, and has submitted proposals on topics including “AI Terminology”, “AI Risk Management” and “General Requirements for Intelligent Service Robots”. However, by comparison, the United States continues to hold a dominant position in international bodies such as ISO and IEEE SA, retaining crucial discourse power on core issues [核心议题中仍掌握关键话语权] such as AI ethics, model interpretability, trustworthy AI and algorithmic transparency.

Most importantly, ISO/IEC JTC 1/SC 42 is the core body for setting international standards in artificial intelligence. Its secretariat is the US National Institute of Standards and Technology (NIST), operating through ANSI (the American National Standards Institute). Since 2018, this body has led the drafting of several key standards, including the AI Management System (ISO/IEC 42001:2023) and AI Risk Management Guidelines. SC 42 comprises five working groups (WGs), each responsible for a distinct technical area. Among these, Chinese experts hold the convenorship of only one group—WG5 on “Computational Approaches and Computational Characteristics of AI Systems”. The convenors of the other groups come from Canada, the US, Ireland and Japan. As such, in terms of strategic direction, control of the secretariat, and agenda-setting, China still lags far behind the United States. This structure highlights that while China is making sustained efforts in international standardisation, there remains considerable room for it to increase its presence in organisational leadership and more generally in senior roles [within international standard-setting bodies].

Given the current situation, I recommend that China promote the internationalisation of embodied intelligence standards by moving beyond the traditional mindset of “intra-ISO competition” and shifting towards an “ecosystem-led” model driven by technology and open-source platforms. Standards should be embedded into the open-source implementations of leading platforms such as AIRSHIP, with reference implementations and toolchains published on global mainstream technical communities like GitHub and Hugging Face—thereby enabling external adoption through a “code-as-standard” model.

At the same time, standards should be tightly integrated with Chinese hardware and software systems, accompanying product deployments abroad into emerging markets, thereby establishing these standards as the de facto choice. Only when Chinese standards are widely adopted in technical applications and development can China truly build global discursive power — elevating itself from “participant” to “rule-maker”.

Due to how and when it developed [历史发展原因], China lags far behind the United States in both the maturity of its AI open-source ecosystem and its influence over international standards systems. This section will discuss the international discourse power and influence of Chinese AI researchers. In particular, it will look at how global influence of individuals can help China reach mainstream user groups worldwide.

In recent years, Chinese AI experts have significantly increased their standing within — and become distributed more widely across — the global academic community, laying a solid human capital foundation for China's technological internationalisation. According to MacroPolo’s 2023 report, 38% of top AI researchers in the US were originally from China—slightly surpassing the 37% with domestic US backgrounds. In top international AI and robotics conferences and journals, Chinese scholars are approaching US levels in terms of both their output of papers and participation in peer review. Data from Guide2Research also show a steady rise in China’s share among the world’s top-ranked computer scientists, indicating that the academic calibre of Chinese AI talent is now world-leading.

Building on this, national and local authorities in China have introduced numerous policies in recent years to encourage talent repatriation [Note: of which Liu is a beneficiary] and to support the development of research platforms—strengthening the country’s AI R&D capabilities. However, boosting the global influence of Chinese AI cannot rely solely on “talent returning home”; it also requires a systematic strategy for “talent going abroad”. More outstanding Chinese AI professionals should be encouraged to work in Belt and Road countries and other emerging markets. This means actively engaging in the development of local education systems and industry by taking up academic positions, establishing research institutes, launching joint laboratories, or founding tech companies. This will not only support the global diffusion of Chinese technologies and standards but also help build a China-centred global network of scientific and technological cooperation.

Although systematic data on the number of Chinese scholars teaching in Belt and Road countries is still lacking, the activity level of cooperation platforms such as the University Alliance of the Silk Road suggests that China has considerable potential in terms of talent export and knowledge diffusion. Accordingly, while continuing to encourage “talent repatriation”, the state should also establish institutional frameworks to support “talent going abroad”. These might include setting up dedicated overseas teaching and research cooperation funds, advancing joint training mechanisms with Belt and Road countries, and promoting localisation of technology standard certifications and curriculum development—thereby enabling the global expansion of Chinese AI and embodied intelligence and strengthening China’s international discourse power.

This article has provided a systematic examination of the current international landscape and the technological, open-source, standards-based and talent-related foundations underpinning the global expansion of Chinese AI.

In the face of a new wave of trade protectionism and technological blockades launched by the United States in 2025 — which are leading to the restructuring of the global technological order — China is entering a strategic window for technology diffusion. Historical experience shows that while technological leadership is a prerequisite, without the combined strength of early adopters, standard disseminators and consensus-builders in governance, it remains difficult to establish a dominant position within the global technology system. Therefore, the development of Chinese AI must not be confined to domestic applications and talent repatriation; instead, it must shape the global ecosystem proactively, foster international consensus and construct a China-led pathway for innovation diffusion.

To this end, the article recommends that China formulate and implement an integrated internationalisation strategy built around “technology–open-source–standards–talent”. Specifically, this means:

1. Strengthening technological leadership by focusing on critical areas in foundation models and embodied intelligence systems, and promoting continuous technological [upgrades] through iteration;

2. Building an open-source ecosystem portal, enabling the international release of systems, models and toolchains via platforms to capture greater market share;

3. While striving for greater influence [话语权] within the existing international standards system, China should simultaneously promote an “open source is the standard” approach—embedding standards into hardware–software product systems and exporting them to emerging markets as part of their global deployment;

4. Encouraging the international mobility of top-tier talent, supporting their development in Belt and Road countries and beyond, and building a global network for R&D and standards dissemination—thereby securing meaningful international discourse power and a structural advantage for Chinese AI technologies and governance frameworks.

Zilan Qian is a fellow at the Oxford China Policy Lab and an MSc student at the Oxford Internet Institute.

What conditions made DeepSeek possible? Despite widespread debate, much of this discussion remains concentrated at the macro-level of nation-states or the micro-level of tech companies. In prevailing narratives, DeepSeek is either seen as a symbol of China’s rising technological prowess or a lone disruptor challenging a top-down innovation system. Hangzhou 杭州, the city where DeepSeek is based, rarely takes center stage.

A closer examination of Hangzhou’s emerging tech ecosystem reveals that DeepSeek did not appear by chance. Hangzhou is the home of six other emerging tech companies — nicknamed Hangzhou’s “six little dragons (六小龙)”: Unitree (宇树科技) and Deep Robotics (云深处科技), two of China’s leading robotics companies; Game Science (游戏科学), which produced China’s first AAA game Black Myth: Wukong; BrianCo (强脑科技), a brain-machine interface innovator; and Manycore Tech (群核科技), the world’s largest spatial design platform as of 2023. Even earlier, in 2000, the city saw the emergence of the Alibaba Group — now the second largest e-commerce platform in the world and the developer of another leading Chinese AI model (Qwen).

So, how much does geography matter for the emergence of these companies? After DeepSeek made headlines, the media started to name Hangzhou “China’s Silicon Valley [1] [2] [3].” This convenient comparison is often used to imply a zero-sum U.S.-China rivalry, as if Hangzhou is China’s secret AI and robotics hub designed to challenge the Silicon Valley-backed U.S. The label also projects a misleading image of Hangzhou as a hyper-technical, Silicon Valley-style hotspot, which obscures the fundamentally different comparative advantages and strategies at play.

The Missing Ingredients

Silicon Valley has many essential ingredients that Hangzhou lacks. Researchers argue that Silicon Valley's model has six interconnected elements: (1) venture capital, (2) human capital, (3) university-industry ties, (4) direct and indirect government support, (5) industrial structure, and (6) support ecosystem. Compared to Silicon Valley, and even to most tier 1 cities in China, Hangzhou lacks at least four of these elements, with no clear advantages in venture capital, human capital, university-industry ties, or industrial structure.

Firstly, the venture capital system in China has always been weaker than that in the US, and the gap has significantly widened in the past few years. Overseas and domestic VC fundraising for Chinese companies has drastically fallen, with RMB-denominated funds falling from 88.42 billion USD in 2022 to 5.38 billion in 2024, and USD-denominated funds from 17.32 billion to 0.75 billion during the same period. Within China, Hangzhou did not stand out as a recipient, with venture capital investment mainly flowing into Beijing, Shenzhen, and Shanghai from 2000-2022. Although Zhejiang province (which houses Hangzhou) was the biggest recipient of venture ­capital funding in 2024, this capital poured in only after companies like Game Science and Unitree had already begun to gain national attention. In fact, Zhejiang saw 41 new corporate venture capital funds registered in 2024, the highest among 18 mainland provinces, which highlights how investment responded to, rather than catalyzed, the region’s tech momentum.

Likewise, Hangzhou lags behind not just Silicon Valley but also major Chinese cities in terms of human capital. The city does not have a strong university cluster. Although some would hype up Zhejiang University as China’s Stanford, in reality, Hangzhou’s higher education is not even competitive among other top cities in China. Zhejiang University (ZJU) is the only elite university included in the national 211 project in the whole of Zhejiang province. In comparison, Beijing has 26 such universities, Jiangsu 11, and Shanghai 10. This shortage has broader implications. Although talent can migrate, the lack of top universities also makes a Hangzhou hukou (household registration) less appealing. Due to the provincial nature of China’s university entrance exams, living in a city with prestigious institutions (like Beijing and Shanghai) offers more educational opportunities. That’s because top universities allocate a larger share of admissions quotas to local students. For example, in 2024 Peking University and Tsinghua University recruited 580 out of 68,000 students from Beijing, compared to 380 out of 405,000 from Zhejiang province. The admission rate for Beijing students (​​0.85%) was 9.5 times that of Zhejiang province (0.09%).

While university-industrial ties do exist in Hangzhou, they are far less rigorous than the “Silicon Valley model.” The few cases of cooperation between ZJU and Alibaba1 are incomparable to the numerous startup accelerators and the wide range of university-industry collaborations offered by Stanford or the University of California, Berkley. Overall, China has relatively weak university-industry ties compared to most prominent research universities in the U.S. Even within China, Tsinghua University and Shanghai Jiao Tong University precede ZJU in terms of unicorn incubation capacity.

Moreover, although almost all founders of the “six little dragons” and Alibaba have ties to universities in Zhejiang, the successful tech companies did not directly spin out from the universities themselves. Liang Wenfeng founded High-Flyer, the hedge fund behind DeepSeek, eight years after graduating from ZJU. Alibaba founder Jack Ma applied and got rejected from 30 different jobs after graduating from Hangzhou Normal University. BrainCo was a spinout from Harvard’s Innovation Lab during the CEO’s postgraduate studies at Harvard, although he completed his undergraduate degree at ZJU. After graduating from Zhejiang Science and Technology University, Wang Xingxing, the founder of Unitree, went to Shanghai for master studies and joined DJI, China’s leading drone company. The distance between these entrepreneurs and their Zhejiang academic backgrounds makes it challenging to prove a causal connection to Zhejiang’s university innovation capacity.

Lastly, unlike Silicon Valley, which developed on a base of the Cold War defense industry, the city does not have a strong industrial history. In 2023, Hangzhou’s industrial gain was 2107.4 billion RMB, 12th among major cities in China and less than half of what Shenzhen (4851.0 billion RMB) made that year. Moreover, Hangzhou’s industrial structure is largely dominated by light industries such as textiles (i.e., Silk) and food & beverages (i.e., the F&B giant Wahaha).

The Special Sauce: Market-Driven Flexible Governance

If Hangzhou can’t compete with Silicon Valley — or even Beijing and Shanghai — on venture capital, elite talent, or deep university-industry ties, then maybe we’ve been asking the wrong question. What if Hangzhou’s strength lies not in what it has, but in how it works with having less? Because Hangzhou doesn’t hold the same political, financial, or industrial significance as China’s top-tier cities, it may have more room to exercise agency at the city and provincial levels in shaping its tech industry and broader innovation ecosystem. This relative autonomy has allowed local officials more space to shape the city’s tech sector and broader innovation environment. Rather than taking a top-down or directive stance, the government has adopted a more service-oriented approach.

Hangzhou fosters a policy environment that actively supports small and micro enterprises, enabled by what Liang Chunxiao of the Pangoal Institute (盘古智库) calls “flexible governance (柔性治理).” In contrast to Silicon Valley’s boom-and-bust cycles that favor survival of the fittest, Hangzhou’s bottom-up orientation encourages experimentation and decentralization, allowing grassroots innovators to flourish without being eclipsed by state-backed giants. Rejecting the traditional image of government as a top-down authority, Hangzhou officials have openly embraced a “waiter” (店小二) and “nanny” (保姆) mentality that positions them as facilitators, not enforcers, in the city’s innovation ecosystem. Unitree once faced an urgent need for systematic protection of its intellectual property rights. In response, Hangzhou promptly rolled out a fast-track patent pre-examination service, sending officials directly to the company to explain the process and offer guidance. Similarly, during the seven years Game Science spent developing Black Myth: Wukong, the local government helped with licensing applications and connected the company with animation studios, and even arranged daily meal deliveries to their cafeteria.

Moreover, Hangzhou’s flexible governance is closely aligned with market principles, making the city especially attractive to a wide range of domestic private enterprises beyond just technology companies. This broader foundation helps explain why a company like DeepSeek, which originated as a spinout from a financial services firm, could emerge and thrive. As Qin Shuo 秦朔, a Chinese business thinker and former CEO of China Business News (CBN), notes, DeepSeek stands out for its market-driven innovation model: rather than depending on state research institutes or tech giants, local firms like DeepSeek have succeeded by identifying unmet market needs and responding swiftly.

This market-oriented environment has deep roots in Zhejiang province’s governance philosophy. As early as the 1980s, during the early phase of China’s economic reforms, Zhejiang — of which Hangzhou is the capital — pioneered efforts to encourage private enterprise and entrepreneurial experimentation. This legacy paved the way for Alibaba’s eventual rise. When Alibaba was founded in the late 1990s, Jack Ma tried but failed to headquarter in Beijing or Shanghai, due to expensive rent and bureaucratic barriers. He later moved the company back to Hangzhou and has received government support since. In 2015, Ma remarked on his decision to headquarter in Hangzhou: “Beijing favors state-owned enterprises, Shanghai prefers foreign companies, and Alibaba (as a domestic private company) was nothing in the eyes of Beijing and Shanghai. If we return to Hangzhou, we become the local only child (独生子女) (who receives all attention and support from parents).”

The success of this “only child” has in turn strengthened the city itself, creating a mutually reinforcing relationship. Beyond e-commerce, Alibaba has provided Hangzhou with vital infrastructure: digital financial services, cloud computing, an ecosystem for entrepreneurship and innovation, and a foundational employer and taxpayer. This ecosystem has laid the foundation for successive waves of tech companies to grow and thrive, reinforcing Hangzhou’s position as a leader in China’s digital economy.

Market-Driven Innovation… with Hangzhou Characteristics

Although the “market-driven” innovation model is generally regarded as a typical U.S. innovation model, Hangzhou’s “market-driven” model is different in that it still maintains a level of centralized governance, rather than relying entirely on market forces and entrepreneurial dynamism. Unlike the government support in Silicon Valley, which sometimes prioritizes tech innovation at the expense of public living conditions, Hangzhou’s government sees quality of life as a critical strategy for attracting companies and talents.

But this governmental support is also different from a stereotypical Chinese “state-driven” innovation model, as the local government positions itself as an enabler, rather than a controller. The lack of state-backed research institute clusters, traditional national champions, and strong industrial foundation contributes to the government’s humble attitude and the desire of private enterprise. If Hangzhou were more strategically important, more industrial, or more academically prominent, DeepSeek might not have had the creative space to emerge.

Is the Hangzhou model replicable in other Chinese cities? When DeepSeek first emerged, local governments and media were frantically seeking the answers to this question. On February 7, Jiangsu province’s official media platform published a three-part series: “Why Did DeepSeek Emerge in Hangzhou?”, “Why Can’t Nanjing [the capital of Jiangsu Province] Produce Its Own ‘Six Little Dragons’?”, and “Hangzhou Has DeepSeek — What Does Nanjing Have?” On February 10, the Jinan government published “What Can Jinan Learn from Hangzhou’s ‘Six Little Dragons’?” And on February 12, Hefei Daily ran a front-page editorial again asking: Hangzhou has DeepSeek — what does Hefei have?

To summarize those reports:

• The Jiangsu government concluded that Nanjing companies lack customer sensitivity, which is crucial in sectors like EVs, AI, and e-commerce, but emphasizes that Nanjing firms can benefit from their strong ties to government and industry.

• Hefei attributes Hangzhou’s success to its reconfiguration of production factors, which enabled a swift transition from the internet economy to AI. Hefei is pushing to grow sectors like EV, photovoltaics, and biopharma, in part by increasing investment in the province’s prestigious University of Science and Technology of China.

• The Jinan government notes its similarity to Hangzhou in terms of pro-innovation policies, but notes that Hangzhou invests beyond national champions and takes a pragmatic, rather than bubble-chasing, approach to tech. Citing Hangzhou’s brain drain and weak regulatory oversight, the report warns against copying innovation models wholesale: “A city's innovation ecosystem requires both passion and cool-headed reflection. When the capital frenzy subsides, only those technologies that truly cross the ‘valley of death’ will endure.”

But the Silicon Valley model is not replicable in every region. As much as one might want to create ChatGPT-like innovations using fixed recipes and ingredients, the kitchens and cookware differ everywhere. Granted, one cannot cook without rice (巧妇难为无米之炊): talent is essential to innovation, just as compute is to training LLMs. But the condiments and methods of cooking vary depending on the kitchen — that is, the political environment — and the amount of staple ingredients like compute, talent, or funding. It makes sense for Silicon Valley to thrive on venture capital and a deep industrial base, just as it makes sense for Hangzhou to develop through its distinctive governmental support. After all, innovation ecosystems aren’t like LLMs that can be ranked through benchmark tests — even benchmark tests are increasingly not reflective of the actual capability and nuances of LLMs.

Tasting the Dish

The most beneficial thing DeepSeek has brought to my life is that it makes introducing where I’m from much easier — especially when “I’m from China” doesn’t satisfy the listener.

“So, where in China?”

Over the years, I’ve tried introducing Hangzhou through various achievements I thought were important: “The city that hosted G20”, “The city with UNESCO sites West Lake, Grand Canal, and Liangzhu City,” or “The city that hosted the Asia Games”. None of these worked better than “a city near Shanghai” until January 2025. Now, I simply say, “The city where DeepSeek is from.”

“Oh,” replied many people, “so it’s like China’s Silicon Valley.”

“No. It is nothing like Silicon Valley.”

1

Yes, ZJU has an innovation center under its international business school; yes, ZJU does try to credit Deepseek and other Hangzhou tech companies’ success to its “entrepreneurial spirit and innovative thinking”; and yes, Alibaba, China’s largest e-commerce company, does have long-time and diverse collaborations with ZJU on digital healthcare, quantum physics, and other frontier technology.

Jake Newby, author of the China music Substack Concrete Avalanche, presents his official playlist of China’s best new music. It includes ADHD-inspired hip hop, experimental ambient music from rural China, and Shanghai cold wave, finishing off with a “mind-blowing” hyperpop track.

Listen to the mix on ChinaTalk’s podcast feed! Check it out on Apple Podcasts or wherever you like to listen to podcasts. Spotify Podcasts took down this episode so unfortunately it’s not available there!

1. ‘Rhyme’ — Rubey Hu

This first song is a beautiful piano piece by Rubey Hu, entitled ‘Rhyme’ and taken from his hypnotically soothing new album In Different Rooms. Although this is a solo album, Hu is best known for his work with the band Hualun 花伦, who were formed in Wuhan.

2. ‘The Last of the Mohicans’ — SMZB 生命之饼

Next, we have the Mandarin language version of ‘The Last of the Mohicans’ by SMZB, one of China’s longest-running punk bands. They were founded in Wuhan in 1996 and played a pivotal role in establishing that city as a major centre for Chinese punk. While influential frontman Wu Wei now lives in Europe (and it’s become increasingly difficult for SMZB to release their outspoken songs in China), that hasn’t stopped the band from putting out new music and booking a live date in the UK this summer.

3. ‘秋茄子之味’ — 红发少年杀人事件

We follow up SMZB with a track from a far newer band, Guangzhou’s The Red-Haired Youth Murder Case, who put out their debut EP in March on excellent Shenzhen DIY label Small Animals Records. [Ed: The title of this track translates to “The Taste of Fall Eggplant.”]

4. ‘The Wanderer of Renfengli 仁丰里的闲逛者‘ — DaYe 大叶

Next, we have a track from one of my favourite electronic records to come out of China this year. DaYe’s Road to Spring — released on brilliant Beijing label bié Records — is inspired by the canals, parks and alleyways of her home city of Yangzhou, with the album cast as a musical attempt to capture these locations based on her childhood memories of them, while she learns to navigate new surroundings up north.

5. ‘back to the pond 1’ — Jian Cui

Now it’s over to Guizhou for a piece by producer Jian Cui, whose experimental, ambient-ish album back to the pond is inspired by the time he spent in Nujiang Lisu Autonomous Prefecture 怒江傈僳族自治州, including an expedition with a group of aunties to gather some sphagnum moss (the whole back story is rather poetic).

6. ‘Chapter II’ (excerpt) — Chen Mulian 陈睦琏 et al for xuán yīn 玄音

We stay in the countryside for a while longer with a snippet from one of four 20-ish minute pieces that make up The Cranes Appeared in the Forest 林森鹤见, an album recorded on a wooded mountainside in Zhejiang province.

7. ‘月光爱人 De Luna Amour’ — 黑木 Heimu

To follow, we have something completely different: dark, electronic cold wave from Shanghai, courtesy of producer Heimu.

Subscribe now

8. ‘Ⱪorⱪetteng ⱪobeze / The Kobyz of Korkut 霍尔赫特的库布孜‘ (excerpt) — Mamer 马木尔

Next up is Mamer, a Kazakh musician from Xinjiang who found international fame when he released an album of traditional folk songs on Peter Gabriel’s Real World Records back in 2009. Since then, he’s been backing away from the world music circuit with increasingly experimental spins on Kazakh folk, but a recent release on fantastic cassette label Dusty Ballz finds the artist in a quieter, contemplative mood with a set of soft guitar pieces, such as ‘The Kobyz of Korkut’. Editor Lily’s favorite song on this playlist!

Share

9. ‘སྒྲོལ་མའི་བསྟོད་པ་། Praise to Tara 度母赞’ — Kalzang Samdrub

That excerpt from Mamer’s work is followed by a short but sweet Tibetan song performed by Kalzang Samdrub in the Amdo Ballad tradition, a form of folk music found across the Tibetan areas of Gansu, Qinghai, and Sichuan.

10. ‘我不知不觉不伦不类’ — 小老虎 J-Fever

Now we jump to J-Fever with one of the less scattergun tracks from his recent album ADHD Snake, which, as the title suggests, is inspired by ADHD and is otherwise a kind of helter-skelter, kaleidoscopic stream of consciousness kind of record — in a fascinating way. [Ed: The title of this track roughly translates to “Unconsciously Out of Place.”]

11. ‘1911 4th Mov. (live) 一九一一 第四回’ (excerpt) — Zhaoze 沼泽

Crashing in next is Zhaoze [Ed: literally, “Swamp”], a post-rock band from Guangdong who make use of a self-built electrified guqin in their music, with an excerpt from a new, string quartet-featuring version of ‘1911 4th Mov.’, taken from their live album The Triangle, which came out in early May.

12. ‘本该走神的(Should've Been Lost)’ — 张醒婵 Nono

The last few times I’ve done one of these for ChinaTalk, I’ve ended with an intense metal track. But I’m going to break with that tradition this time and round off with two pop-leaning songs — well, relatively speaking.

The first is from Zhang Xingchan, whose debut album No, No! was a breakout success in 2024. In May of this year, the Wuhan-based artist released what she called the ‘B-side’ to that album, a five-track EP entitled No, Now! — which is where the penultimate track comes from.

13. ‘失乐园’ — DJ小女孩 DJ Gurl

And then I’m going to close out with a longer track from Guizhou-based artist DJ Gurl, which kind of defies description — you really just have to listen to it. [Ed: This track’s title means something like “Losing Paradise.”]

If you like something you heard on this playlist, please support the artists by buying their music. And subscribe to Jake’s substack! Thanks for listening.

Concrete Avalanche

A newsletter about music from China

By Jake

Last year, ChinaTalk covered Operation Causeway — the WWII-era US military plan to invade Japan-occupied Taiwan and use the island as a launching base for an assault against Japan. As US Army LTC Kevin McKittrick explained, Causeway was ultimately scrapped because, “the enemy always has a vote.”

Today, we’re exploring how Imperial Japan fended off Operation Causeway — in the form of its own large-scale Taiwan-Okinawa defensive plan, dubbed Operation Sho-2Go 捷2号作戦 (“Operation Victory No. 2”). Executed along side Operation Sho-1Go (Philippines), Operation Sho-3Go (Japan mainland), and Operation Sho-4Go (Hokkaido), Sho-2Go involved rapidly fortifying Taiwan with five divisions — approximately 165,600 troops — and employing a strategy of “vertical resilience” by constructing extensive underground tunnel networks and fortifications to leverage Taiwan’s mountainous terrain. Sho-2Go’s key innovation was to thoroughly disperse and conceal air and naval forces in protected positions, preserving them from initial US attacks, and then concentrate these forces in asymmetric kamikaze strikes against US landing fleets at the critical moment of invasion. Sho-2Go’s successful deterrence of a Taiwan invasion led American troops to target Okinawa instead.

This article is the product of fantastic archival research — using Japanese-language source materials — conducted by Japan Air Self-Defense Force Colonel Hirokazu Honda, who just finished a year-long stint at the Air War College (AWC) at Maxwell Air Force Base, Alabama. Colonel Honda earned his commission upon graduating from the National Defense Academy of Japan 防衛大学校 in 2004; he also has a master’s degree in aerospace engineering. He has served as A6, A5, and A1 staff for the Air Staff Office; Commander of the 36th Air Control and Warning Squadron; and defense strategy planner at the Joint Staff Office (J5). Colonel Honda was advised by AWC professor and US Air Force Colonel Jared D. Paslay.1

Thanks to the US-Japan Foundation for sponsoring this article.

The strategic importance of Taiwan to Imperial Japan

One of the earliest Japanese leaders to focus on Taiwan’s importance to Japan was then–Minister of Education Kowashi Inoue 井上 毅. In a letter to Prime Minister Hirobumi Itō 伊藤 博文 in 1894 — just after the outbreak of the First Sino-Japanese War — Inoue observed,

By controlling the right of navigation in the Yellow Sea, the Sea of Korea, and the Sea of Japan through the possession of Taiwan, we can open and close the gates to the Orient, and by extending our communication with Okinawa and the Yaeyama Archipelago, we can control the entry and exit of others.

Inoue also posited that, should Taiwan fall into the control of another nation, “the security of the Okinawan islands would be disturbed.”

A year later, in 1895, Japan acquired the island of Taiwan as a colony, per the Treaty of Shimonoseki 下関条約, and in 1941, Japan invaded Southeast Asia in search of resources. This expansion made Taiwan indispensable to Imperial Japan as a logistics and trade-route node — and militarily, it made Taiwan indispensable as a naval base, logistics base, air base, information base, and communications relay station for deployment to the south. The colony’s strategic importance was well-summarized in the 10th Area Army Operation Record: “Taiwan, as the southernmost territory of the Empire, has long had significance in imperial policy toward the south.”

Even so, as late as 1940, Imperial Japan maintained only the minimum forces necessary to guarantee Taiwan’s security. The forces on the island consisted of two garrison regiments, one anti-aircraft gun regiment (with fortifications in Keelung 基隆, Kaohsiung 高雄, and Penghu 澎湖), and one flight squadron. Pre-WWII Taiwan was considered relatively secure from external threats — China was preoccupied in its own civil war, and neither Chiang Kai-shek’s KMT forces nor Mao Zedong’s CCP forces had the will or ability to reclaim Japan-occupied Taiwan — and as such, the island primarily served as a logistics base for Imperial Japan.

The prelude

In November 1940, in light of the intensifying Sino-Japanese War, the Imperial Japanese Army decided to expand its deployed forces in Taiwan — but only gradually, with a completion date of the expansion was set for 1943. At the Pacific War’s onset in December 1941, only a few air-defense units, a security force for Kaohsiung Harbor, a supplemental force of one division (without command staff), and an air-education unit were stationed on the island.

Three years later, and Japan’s advance had slowed significantly, while US forces began counterattacks that threatened both the Japanese mainland and its Southeast Asian territories. As of March 1944, the Imperial Japanese Army estimated that, if enemy forces were concentrated in the South Pacific in the future, a landing force of seven to ten divisions would invade Taiwan and Okinawa from the Philippines. By then, Colonel Jihei Ichikawa 市川 治平, the Taiwan Army’s Chief of Staff for Operations, was requesting seven divisions for Taiwan, while Lieutenant General Shūichi Miyazaki 宮崎周一, Director of Operations at the Imperial Japanese Army Headquarters, wanted even more: his memoirs mentioned a desire to deploy nine to ten divisions to Taiwan.

On March 15, 1944, then-Colonel Kazushi Sugita 杉田一次 led an Imperial Japanese Army field inspection team to Taiwan to research how to strengthen the defense of Taiwan and Okinawa. Meanwhile, the Commander-in-Chief of the Sasebo Naval Base submitted a report to the Commander-in-Chief of the Navy, observing that the island of Taiwan was “defenseless against air attacks and enemy landing operations,” and as such, proposed increasing air bases, installing anti-aircraft radar, and strengthening base facilities.

A week later, on March 22, 1944, the Imperial General Headquarters ordered the reinforcement of Taiwan and Okinawa by way of issuing “Outline of Preparations for Operation No. 10,” which aimed to “strengthen operational readiness and prepare for enemy surprise attacks in order to defend the mainland and secure transportation to and from the southern areas.” The initial emphasis on guarding against small-scale intermittent attacks — as opposed to defending against full-scale landing operations — focused on strengthening defense facilities, like air bases, so they could respond to enemy surprise attacks as well as secure maritime traffic and critical lines of communication. For this reason, the outline stated that preparation for air operations support would be the highest priority, and all “other matters shall be subordinate to it.”

After losing many aircraft and ships to US forces, the Japanese finally decided to abandon Saipan in June 1944. At that time, only one division and one mixed brigade were deployed to the island of Taiwan — an “overwhelmingly insufficient” force posture.

A strategy emerges

Following Japan’s defeats in Saipan as well as the Battle of the Marianas, the Taiwan-Okinawa area and the Philippines moved from the rear echelon to the front lines of national defense. Imperial General Headquarters made two initial orders in this regard:

• First, they wanted to triple the number of deployed forces on the island of Taiwan by diverting three divisions, one brigade, and one mixed brigade from mainland China.

• Second, they wanted to treat the defense of Taiwan and Okinawa as defending a single area. As Lieutenant General Miyazaki wrote in his diary,

  Taiwan is a strategic point that will serve as a base for the US invasion of Japan, and Okinawa will serve as a base for air raids on the Japanese mainland. For this reason, the area including Taiwan and Okinawa must be regarded and defended as one area.

  To achieve that end, the Taiwan Army assumed control of both the 32nd Army 第32軍 in Okinawa and the 8th Air Division; the Taiwan Army, on July 11, became a “district army”; and then on September 22, its named changed from “Taiwan Army” to “10th Area Army” 第10方面軍.

Beyond force reassignments, the Japanese leadership also needed a new strategy to defend themselves against the enemy’s overwhelming air and sea power.

At a meeting in Tokyo on June 25, 1944, analyzing their staggering losses in Saipan, the leadership made the deliberate decision to shift from a proactive engagement model to an entrapment strategy, in which, after attracting the enemy to within an ideal range, Japan would launch a concentrated counterattack (including asymmetric, self-destructive kamikaze attacks) just before an invasion-force landing. The leadership acknowledged the overwhelming gap in conventional troop strength between the invading US and defending Japanese airpower, and, rather than pursue advance attacks on enemy aircraft carriers, shifted to a “Fleet in Being” philosophy, wherein Japan would avoid being pulled into a decisive battle and instead conserve airpower in dispersed and concealed locations until a critical window of operational opportunity.

In line with Tokyo’s strategic shift, the 10th Area Army began approaching how it established air bases with a new mindset. The most distinctive feature of the 10th Area Army’s air-base setup in Taiwan was its thorough dispersal of aircraft, fuel, and other facilities. The mastermind of this formulation was Lieutenant Colonel Kan’ichi Ishikawa 石川 寛一, the chief of staff for operations of the 8th Air Division, who stated,

We will thoroughly disperse and conceal planes, fuel, and ammunition against the enemy’s advance air attack, and preserve and secure our forces until the decisive battle. When the enemy lands, we will launch an attack with the division’s frontline forces and thoroughly destroy the enemy’s landing craft.

On July 24, 1944, the Imperial General Headquarters formally issued orders to rapidly strengthen Taiwan-Okinawa defenses in preparation for a decisive battle against US forces, should they attempt to invade either island.

Operation Sho-2Go was born.

Redeployment jockeying

Fulfilling the directive to triple the number of forces on Japan-occupied Taiwan proved easier said than done. In particular, military leaders were jockeying over how to divide limited resources between engagements in the Philippines (Sho-1Go), Taiwan-Okinawa (Sho-2Go), the Japan mainland (Sho-3Go), and Manchuria (Sino-Japanese War). In which area could Imperial Japanese forces most likely win a decisive battle against US forces?

Following the Formosa Air Battle — a major aerial battle from October 12 to 16, 1944, in which Japan lost over 300 aircraft — the Imperial Japanese Army believed it could defeat US forces in a decisive battle by committing additional forces to the Philippines, ie. Operation Sho-1Go. (The basis for this belief, as it turned out, was surviving Japanese pilots’ exaggerated reports that US carrier forces had been decimated in that battle.) On October 18, 1944, following the landing of US forces on Leyte, Operation Sho-1Go was launched. Taiwan’s defending forces at that point consisted of three divisions and one brigade — less than forces of the 32nd Army (four divisions, five mixed brigades) defending Okinawa, an island approximately thirty times smaller than Taiwan. Colonel Ichikawa hoped to extract at least one division from Okinawa and assign it to Taiwan.

Tensions over force redeployments continued into November 1944. On November 4, Colonel Takujiro Hattori 服部 卓四郎 — Director of the Operations Division of the Imperial Japanese Army — traveled to Taipei to hold a meeting on the future distribution of forces between Taiwan, Okinawa, and the Philippines. Participants in that meeting included Lieutenant General Haruki Isayama 諌山春樹, Major General Kiyomizu Kitagawa 北川潔水, Colonel Hisashi Kisaki 木佐木 久, Colonel Hiromichi Yahara 八原博通, and Colonel Jihei Ichikawa. The competing arguments of the Imperial Japanese Army, which recognized the need to divert forces to the Philippines, and those of the 10th Area Army, which insisted on strengthening the defense of Taiwan and Okinawa, ended in a stalemate, with no resolution at the meeting.

Eventually, on November 10, 1944, the Imperial General Headquarters decided to dispatch one division from Manchuria to the Philippines — aiming to defeat the United States via Sho-1Go — and to incorporate one division from Okinawa into Taiwan, to balance the defensive forces between Taiwan and Okinawa.

From the standpoint of defending colonial Taiwan, then, the redeployment jockeying was ultimately successful. By December 1944, Taiwan was garrisoned by four divisions, and in February 1945, another division was transferred from Manchuria, resulting in a garrison of five divisions. One study estimates that these five divisions represented approximately 165,600 troops — an eightfold increase, from just 20,000 troops, in under one year. Though the five divisions ultimately stationed on the island of Taiwan were less than the seven divisions originally claimed to be necessary by the 10th Area Army, they were enough to deter US Admiral Nimitz from launching Operation Causeway: US war planners estimated that Imperial Japanese forces on Taiwan would increase by about 70,000 between July 1944 and March 1945 (in fact, Japanese forces increased by well over 100,000) — which meant that, using a 3:1 attacker-defender ratio, the US offensive force estimate had to be three times larger, at least 210,000 troops.

On the other hand, Japan’s decision to reduce troop strength from Okinawa by diverting one division there to Taiwan may have sealed the fate of the Okinawa defenders. General Buckner came to realize that an invasion of Okinawa — with only three defensive divisions — was more feasible than an invasion of the island of Taiwan, with its five divisions. Okinawa was invaded by the United States at the end of March 1945.

How Imperial Japan would have defended colonial Taiwan

If Admiral Nimitz had given the green light for Operation Causeway, Imperial Japan may well have been ready to defend its colony. Manipulating force ratios was only one part of Sho-2Go.

From August 12 to 16, 1944, the general staff of the Imperial General Headquarters, the commander of the 32nd Army (Okinawa), the commander of the 8th Air Division, and other commanders in Taipei convened to conduct a tabletop exercise to prepare for Operation Sho-2Go. Although many details of the discussions during this exercise are unclear, a skeleton of the Taiwan Army’s operational plan still exists. According to the skeleton, the main battle areas for Taiwan’s defense were projected to be the Kaohsiung-Pingtung 高雄-屏東 area in southern Taiwan and, depending on the situation, Taoyuan 桃園 in the north. Fortifying these defensive positions was the top priority. In addition, since weather conditions such as typhoons led the Japanese to predict the enemy might land as early as mid-September 1944, all initial operational preparations were ordered to be completed by the end of August, ie. in less than a month.

The Imperial Japanese Navy ordered the laying of underwater mines (500 ordinary mines and 4,000 small mines) in the waters surrounding Taiwan — part of an effort to prevent enemy landing operations. Should US landing forces make it through the mines, the 10th Area Army’s “Plan for the Fortification of Taiwan Island” (depicted below) detailed fortification methods along with an analysis of invasion patterns, assessed region by region.

The blue numbers in the map above correspond to the following list:

1. Yilan 宜蘭

• Area characteristics:

  • Could be used as an air base for attack on Okinawa

  • Possibility of surprise landing

• Fortification focus:

  • Airfield-defense facilities

  • Coastal-defense facilities

  • Liaison road defense to Taipei

  • Deepening of rice paddies [ie. fill them with more water, so they are harder to walk across]

2. Hualien Port and Taitung 花蓮港, 台東

• Area characteristics:

  • Isolated from other areas

• Fortification focus:

  • Facilities for independent endurance operations

  • Coordination capability between districts

3. Hengchun 恒春

• Area characteristics:

  • Isolated from other areas

• Fortification focus:

  • Preservation of Pingtung Valley

  • River flooding capability using Pingtung Valley

4. Pingtung 屏東

• Area characteristics:

  • Interception area after enemy landing

  • High east, low west topography

  • Many rice paddies in west

• Fortification focus:

  • Mountain positions in east to lure enemy

  • Tactical advantage from elevated positions

  • Deepened rice paddies as obstacles

5. Fengbitou 鳳鼻頭

• Area characteristics:

  • Interception area but not suitable for defense

• Fortification focus:

  • Extra-solid facilities to withstand bombardment

  • Thickened underground tunnel fill

  • Positions for bombarding Kaohsiung Harbor

6. Kaohsiung 高雄

• Area characteristics:

  • Critical interception point for landing forces

  • Requires coordination with Fengbitou

• Fortification focus:

  • Solid facilities connected with Fengbitou

  • All batteries placed underground

  • Reinforced rear defense

7. Takangshan/Siaogangshan 大小崗山

• Area characteristics:

  • Target for enemy to isolate Pingtung Plains

• Fortification focus:

  • Positions in mountainous areas

  • Capability for bombardment from multiple directions

8. Tainan 台南

• Area characteristics:

  • Strategic point for launching troops north to south

• Fortification focus:

  • Solid bunker near Tainan Airfield

  • Deep vertical position

  • Field obstacles through deepening

9. Taichung 台中

• Area characteristics:

  • Vulnerable to paratrooper attacks

• Fortification focus:

  • Underground facilities 30 meters below ground

  • Accommodation for one battalion

10. Houlong 後龍

• Area characteristics:

  • Possible enemy landing to block north-south traffic

• Fortification focus:

  • Facilities at each highland site

  • Maximum endurance capability

11. Hsinchu 新竹

• Area characteristics:

  • Suitable for obstructing landings in Taoyuan and other areas

• Fortification focus:

  • Solid facilities on multiple plateaus

  • Support for mutual defense

12. Taoyuan 桃園

• Area characteristics:

  • Possible landing site to capture Taoyuan Airfield

• Fortification focus:

  • Multiple reinforced positions for mobile defense

  • Numerous battery positions around airfield

13. Tamsui 淡水

• Area characteristics:

  • Possible landing site for attack on Taipei

  • Could be invaded simultaneously with Taoyuan

• Fortification focus:

  • Positions for endurance warfare

As illustrated above, Imperial Japan did not prepare for a conflict of symmetric attrition, but rather planned to fight via a tenacious strategy of “vertical fortification” that took advantage of the island’s topography. They constructed extensive underground tunnel networks, including 6.7 km near Hualien Harbor, 10 km in the Taitung area, 15 km in the Fengbitou area, 23 km in the Pingtung area, and 39 km in the Kaohsiung-Pingtung area; these tunnels were used for discreet, unexposed troop and supply movements. Imperial Japanese forces also built underground fortified positions — some as deep as 30 meters below ground (Taitung) — developed battery positions completely underground (Kaohsiung), and created positions in mountainous areas to gain height advantage (Pingtung) while reinforcing vulnerable positions in plateau areas (Hsinchu).

Making these extensive, island-wide fortifications so rapidly was a whole-of-nation effort. According to the Fortification Plan, military personnel from regular army units and flight divisions were involved in the construction work, due to the lack of dedicated construction units. Further, Taiwan Area Army records show that many Taiwanese civilians were mobilized for this construction work, alongside Imperial Japanese forces. Many Taiwanese civilians also actively cooperated in the operation of air-defense organizations, such as air-defense surveillance and communications. Assistance from the Taiwanese populace manipulated the force ratios even more to Japan’s favor.

Takeaways for defending Taiwan today

To be sure, Taiwan’s strategic environment has changed significantly from 1945 to 2025. In 1945, the United States intended to use occupied Taiwan only as a staging base for a full assault on Japan; today, China seeks the full annexation of Taiwan. In 1945, Taiwan was a Japanese colony; today, Taiwan is free to focus on its own self-defense, free from regional complexities. In 1945, the availability of mass-produced, asymmetric wartime technology was limited (hence kamikaze attacks); today, asymmetric tech is more abundant.

But the similarities which have survived the past eighty years are also striking: there is a significant power imbalance between the attacker and defender; the defenders will likely have very little advance notice of an invasion; and the defenders must fight despite the attacker’s air and sea superiority. As such, Taiwan and allied militaries have much to learn from Japan’s successful defense of Taiwan during WWII.

First, rapid force expansion deters invasion. Japan’s eightfold increase from 20,000 to 165,600 troops in under a year raised the cost of a US invasion and was a significant part of the calculus which led to the abandonment of Operation Causeway. Thus ground-force readiness is essential: Taiwan should enhance both active and reserve force readiness to raise the perceived cost of invading. One way to optimize Taiwan’s reserve system would be to implement a three-stage reserve system with a “Quick Reaction Reserve,” similar to Japan’s model today. In Japan, there are two types of reserves: a Quick Reaction Reserve, obligated to respond immediately, and the General Reserve. The Quick Reaction Reserve receives more training days per year than the General Reserve, are obliged to respond to call-ups more quickly, and are required to work in departments similar to their active-duty counterparts (and get additional pay incentives). Taiwan could adapt the Japanese model to their domestic needs by establishing a Quick Reaction Reserve quota — say, 150,000 troops.

Second, Taiwan should adopt a modern-day “vertical resilience” strategy. Taiwan should use its mountainous terrain for concealment and protection against overwhelming firepower. Underground infrastructure is vital: Taiwan should modernize and expand existing tunnel networks to protect weapons, communications, and personnel. And vertical resiliency can help establish redundancy in communication networks — from cables deep underground to space-based systems.

Third, asymmetric capabilities are critical to fighting lopsided battles. Taiwan cannot afford to invest in high-end, exquisite capabilities that will take ten years to acquire. Instead, Taiwan is now in a situation where, like WWII-era Japan, it must do everything in its power to strengthen its defenses in a compressed time. Therefore, Taiwan should focus on acquiring asymmetric military capabilities over conventional ones. Drones are an excellent example. They can be mass-produced, stored in ordinary (and concealed) cargo containers, and launched in swarms. And not all drones require expensive guidance systems: many can follow remotely controlled lead drones, while other drones can even be operated by civilians with relatively minimal training.

Fourth, just as important as having deterrence is messaging deterrence. Imperial Japan successfully telegraphed to its would-be US invaders that Taiwan’s force posture was changing rapidly, and that any battle over Taiwan would be horrific. Taiwan today can likewise clearly communicate its ability and willingness to defend. For instance, Taiwan could consider holding public competitions showcasing its military capabilities. What would China think if a world-class drone-piloting student competition were held in Taiwan, where award-winning students told media, “We would like to use our skills to support military operations in the event of a Taiwan contingency?”

And finally, if nothing else, Taiwan and allied nations should feel confident that deterrence is possible. Eighty years ago, Operation Sho-2Go worked: the United States was deterred from invading Taiwan. Deterrence can be achieved again today.

Primary sources

“台湾方面軍作戦記録  昭２１年８月,” Taiwan Area Army Operation Record August 1946, Center for Asian History, National Archives of Japan. [link]

“第八飛行師団沖縄航空作戦記録,” 沖縄作戦記録（改訂版） 昭和２４年１１月,  8th Flight Division Air Operation Record in Okinawa, Okinawa Operation Record (Revised Edition) November 1949, Center for Asian History, National Archives of Japan. [link]

“第１０方面軍態勢要図  昭２０年８月１５日,” 10th Army Situation Map, August 15, 1945, Center for Asian History, National Archives of Japan. [link]

“米機動部隊本土来襲に関する戦訓, 戦訓特報第４１号  昭和２０年２月１６、１７日,” Lessons learned for the US task force’s attack on the mainland, War Lessons Special Report, No. 41, February 16th and 17th, 1945, Center for Asian History, National Archives of Japan. [link]

“昭和19年3月以降台湾軍の実施せる築城に関する重要施策,” Important measures regarding fortifications implemented by the Taiwan Area Army since March 1944, Center for Asian History, National Archives of Japan. [link]

Are we cooked? We’re going to be revisiting this question quarterly now, it seems. To discuss, we have Peter Harrell, former Biden official and host of the excellent new Security Economics podcast, Kevin Xu, who writes the Interconnected newsletter, and Matt Klein, author of Trade Wars Are Class Wars and The Overshoot substack.

We discuss…

• Short-term positive indicators, including the stock market rebound, the continued independence of the Federal Reserve, and the administration’s compliance with court rulings,

• Why talent is the ultimate bottleneck to AI progress, and how the attacks on foreign students, universities, and scientific research will impact America long term,

• Whether Trump’s policies will cause mass divestment from dollar-denominated assets,

• The state of U.S. alliances, from the apparent thawing of relations with NATO and Zelenskyy, to signs of friction with Japan,

• Next steps for Trump’s industrial policy.

Listen now on iTunes, Spotify, or your favorite podcast app.

Dollar Depreciation and Self-Inflicted Brain Drain

Jordan Schneider: Peter, kick us off.

Peter Harrell: We did this last quarter in late April, and looking around the short term over the last three months, the indicators are decent. We’re looking less cooked. Trump TACO’d out on a lot of the tariff wars, so the stock market rebounded. He can post memes about bombing Iran and get away with it — the short-term indicators actually suggest the last couple of months have gone decently well for the United States from a big picture perspective.

However, I’m pretty concerned about the long-term trajectories. Looking at some of the core things that have made the US what it is today — substantial investment in scientific R&D, attracting talent from around the world to do science and innovation, business leadership here in the United States, and the social fabric we have — I remain worried about whether we’re cooked long term, even though I have to give some credit to the last three months.

Kevin Xu: To add onto what you just said, Peter, in preparation for this podcast, I looked at the value of the dollar chart because I remember we brought it up the last time we talked about this. It went down by another 5 to 6 percent from our last recording to today. That was one of the biggest conundrums at the time. Right after Liberation Day, everyone who thinks about tariffs in perhaps the most conventional way would expect the dollar to go up, everything else being equal.

But the erosion of the dollar, no matter how you want to slice it, continues slowly but surely going down. It’s no longer a big story because there have been so many other attention-grabbing geopolitical and economic stories since then. But for me, that’s something I’m going to watch from time to time as this collective “how does the world vote with their feet” question regarding how much credibility they place in the United States government and leadership through how they treat the dollar.

Matt Klein: The overall level of “cookedness” has not actually changed much one way or another. Partly because I discounted the damage from the tariffs. Not that the tariffs announced on April 2nd were good, but that was never the main issue. The attacks on scientific research, the attacks on foreigners coming into the country — and we’re not just talking about changes in Southern border policy, but the way European tourists are treated — across the board restrictions on foreign students coming into universities, the attacks on the rule of law — these things have not abated in any way.

There has been ongoing pushback from courts, but many things remain problematic. The changes to scientific research policy, the changes to approaches on medical research funding in particular and vaccinations — that has not improved in the past few months. If anything, it’s been entrenched.

If I wanted to be more pessimistic, I would say the fact that immediate concerns about things such as tariffs have subsided and the stock market has recovered might actually reduce the pressure to unwind those longer-term, more damaging policy changes that have occurred over the past six months, making them more likely to persist. Even if the immediate impact is not necessarily substantial, it could be a very long-term drag, not just on GDP growth, but on the quality of life, power, and well-being in this country.

The analogy is Brexit. When the UK was voting about whether to leave or remain in the European Union, there were concerns — people called this “project fear” — about an immediate catastrophe if they voted to leave. That didn’t happen, so people said the case was overstated. But the more realistic and correct argument was that everything would be worse — not catastrophically worse, but persistently worse over time. That has been borne out. I fear we’re ending up in a similar, if not worse, situation as a result of all this.

Jordan Schneider: I’m going to take the other side of this argument. Many of the worst-case scenarios we were contemplating in April are much less likely to manifest now. I’m taking Brexit as a win, if that is our analogy, as opposed to something even more dramatic.

Elon is done. DOGE’s energy is abating for the most part, which I think is positive for governance and civil service. We have Congress — maybe they’ll pass a bill — it doesn’t look great, but it’s also not republic-wrecking. There’s going to be an election not that long from now.

On the immigration side, yes, the direction of travel is not great, but we have not banned all foreign students. The country bans that we’ve seen and will continue to see are directionally problematic, but we could have had much more aggressive paths that would have presaged an even more aggressive four years, which have not happened. You’ve even seen Trump at one point tweeting about protecting the farmers, saying we can’t kick out all of our undocumented immigrants.

It’s not just on trade where this administration has blinked. It seems there are other topics — science accepted, we’re in a pretty poor timeline on that front — but at least on the immigration front, it seems we could be in a worse place than where we are now.

Subscribe now

Peter Harrell: I’ll add two points, one on each side of the ledger, building on what you’ve both said.

On the positive side, Trump appears to be stepping back from the brink, at least somewhat, on whether he’s going to outright defy the courts. That felt more real to me a couple of months ago. He did bring home the individual who was sent to El Salvador and charged with crimes — a complicated issue. But he does seem to be stepping back from his showdown with the judiciary, though time will tell. I’d put that in the positive ledger.

On the other side are things that are eroding U.S. centrality in the international economic system, at least long term. I want to pick up where Kevin left off on the decline in the dollar’s value. We’re seeing several other data points where foreign governments and private parties have quadrupled or quintupled their efforts to create financial networks substantially outside U.S. jurisdiction.

You see this in the drop in the dollar’s value — investors are clearly beginning to pull back from the United States. You also see it in the large run-up in gold prices. This 3,000-year-old store of value has had an explosive couple of months, building on a strong year or two, driven very much by central bank purchases of gold as a way of hedging and reducing their exposure to the dollar.

But perhaps moving to gold isn’t even enough. There’s now a debate in Europe, in Germany and Italy, about whether they need to repatriate their gold reserves — which are currently stored deep in the rock underneath Manhattan, in the custody of the New York Fed — because they don’t even trust us as custodians of their gold anymore.

This is a significant data point. Now, it’s not easy to develop large-scale financial networks outside the U.S. — it’s very hard and doesn’t happen overnight. We’re still seeing foreign investment here, and we’ll still attract the capital levels we need. But this foreign government distrust about the long-term stability of the United States, whether right or wrong, is driving something that you’re beginning to see in the numbers.

Matt Klein: I might push back on that a little because while that intuitively makes sense, the numbers aren’t necessarily showing something as extreme as one might have expected.

The dollar depreciation was sharp around April 2nd, but if we zoom out, the dollar is still quite expensive on a trade-weighted basis relative to the rest of the world. It’s still up something like 20 percent compared to early 2021, up 30 percent compared to 2014. It’s still quite expensive — maybe not quite as expensive as before.

Foreign investors had been massively overweight in U.S. assets. On a market cap basis, they were equally weighted, but 70 percent of stock market capitalization being American feels odd given that the U.S. economy is a quarter of the world economy. Yes, American companies may be more profitable, but still, there were plenty of reasons why, with more benign catalysts, you would have seen some of these market moves, particularly on exchange rates or foreign allocations.

In some ways, the story is that there hasn’t been as much of a move yet. I agree that if the Germans and Italians move their gold, that’s symbolically interesting, although what that means in practice is debatable. But it’s striking what hasn’t happened, which is honestly surprising.

The same thing applies on the U.S. side — have Americans been moving a lot of money out of the U.S. into foreign assets? Maybe some, but it’s not super clear in the data that it’s been enormous volumes of the kind you would have expected.

We have other examples of countries with sharp moves in response to political instability — Turkey in March, for example, after the arrest of the mayor of Istanbul. There was a very sharp move, actually a similar-sized move in the exchange rate for the Turkish lira as for the dollar. But in terms of the quantities of what was actually happening, it was on a much different scale. Turkey’s central bank only has $40 billion in reserves to keep the currency stable. We didn’t see that here.

As much as it would be logical to say that’s what’s happening, I don’t think we’ve seen it yet. Maybe we will; maybe it’ll be more gradual. The gold price move is certainly interesting, although that really dates back to the sanctions on Russia’s central bank when that trend began. Otherwise, it’s hard to see it as a dollar-specific story.

Jordan Schneider: Matt, we’re doing better on the Fed, right? That was a news story that completely burned itself out. I think we’ll have a Fed come 2028.

Matt Klein: Maybe. This is the other thing — I talk about the independent Fed, the independent judiciary, but these are Senate-confirmed positions. It’s not clear to me what kind of pushback the Senate would provide if you have a Trump-supporting majority there.

The judiciary we have now is not necessarily the judiciary we’ll have four years from now. That could change. The Fed situation could change. If I remember correctly, there’s a passage in the big, beautiful bill they’re putting through that basically says administration officials would be immune from contempt of Congress. I’ll defer to the lawyers on the exact terminology, but there are things being done by the other branch of government to support the potential for further executive overreach, which should be concerning even if it’s not necessarily generating the same kind of headlines and market responses as tariffs.

Peter Harrell: In fairness, I said a couple of minutes ago that Trump is stepping back from his confrontation with the judiciary in a noticeable way, though the long-term impact remains to be determined. Part of that is because he’s winning. He’s making the assessment that he can win at least a large enough percentage of his cases in front of the judiciary.

We saw that today, at least temporarily, with what the Supreme Court did on nationwide injunctions and birthright citizenship. He’s essentially saying, “If I’m winning two-thirds of my big cases in the courts, maybe I don’t need to outright confront or defy them over the third that I’m losing."

As a lawyer, I’ll take compliance with judicial rulings. That’s all I’m saying.

Jordan Schneider: Peter, give us the “tariffs are actually illegal” update.

Peter Harrell: This is one near and dear to my heart. Most of Trump’s tariffs — certainly all of the Liberation Day tariffs from April that caused significant market turmoil — weren’t imposed under a trade law statute. He imposed these tariffs under a 1977 emergency powers statute that previous presidents had used for sanctions on Iran and Russia. It had never previously been used for tariffs.

This is progressing through the court system. There’s a specialty district court in New York called the Court of International Trade. Plaintiffs challenging these tariffs won at the specialty district court, but that victory is on hold while the government and plaintiffs deal with an appeal before the Federal Circuit, one of the federal appellate courts.

The Federal Circuit will have its hearings on this case on July 31st. We’ll likely get an opinion from the Federal Circuit by the end of August, and then this will go up to the Supreme Court in the fall.

If I were betting, I would bet Trump loses on the merits here. At the end of the day, these tariffs are going to be found unlawful. That doesn’t mean the end of all tariffs — it means a mad scramble across the U.S. government to see what tariffs they can recreate under other legal authorities.

Matt Klein: They’ve been doing that with Section 232. The 50 percent tariff on steel isn’t being challenged. The pharma tariffs in the pipeline are now separate.

Kevin Xu: There’s your alpha for the episode.

Peter Harrell: For those thinking about this from a market perspective, one thing that’s going to happen is that companies paying the tariffs this year will get a rebate. They’ll be entitled to a rebate of paid unlawful tariffs, which would probably happen next year. From a market perspective, you’ll see an earnings shift from this year to next year as they get rebates of tariffs paid this year, which they’ll then have to pay going forward again. Retroactively, 2025 might look better on earnings than it does right now.

Matt Klein: Are the tariffs tax-deductible? Are tariff payments deductible against income tax? Is the rebate going to be taxed?

Peter Harrell: I’m not a tax lawyer — that’s a good question. It’s certainly a business expense, so you can claim it as such.

Kevin Xu: They’re not going to tax tips.

Peter Harrell: How to tax this is full employment for lawyers. That’s Trump trade policy 101 — full employment for lawyers.

Talent > Transistors

Jordan Schneider: Kevin, how about an AI update?

Kevin Xu: One thing in the AI world — whether it’s investment or technological progress — speaks to how equipped the U.S. is. I’m not necessarily saying China will lead or run the table — I’m analyzing our country specifically.

CapEx is growing for at least another one or two years. The AI trade is on. But everyone’s realizing that GPUs are no longer scarce. That was more of a temporary thing. If you have the money, you can get whatever latest NVIDIA GPU you want for a data center.

The pace of energy coming online in the U.S. is still relatively slow compared to other places. That’s why the Middle East swing by the president and all the AI executives is so consequential. There’s an executive order coming from the White House to make that easier in the U.S. — we’ll see what happens in real life.

All of this really comes down to people. I’m talking about both the supply of general contractors to build and dig data centers and wire all the power sources to make data centers come online as quickly as possible, as well as the top-end research talent that’s becoming more of a bottleneck — not just for which country could advance, but which company within which country will advance.

The most interesting position right now is probably Mark Zuckerberg’s spending spree to hire whoever from OpenAI is willing to jump ship. He’s willing to pay whatever he needs to pay. It’s like a classic “buy the free agent market” way to win the championship.

It’s an interesting comparison, keeping the sports analogy going, to how DeepSeek got to where it is — a pure scouting and drafting way of building their team. Their team got built within China (maybe they got some secret offer from Meta too, we don’t know), but these are very different ways of accumulating AI talent, which we know is in real scarcity — not the hardware anymore.

How do we extrapolate that to which country or company will be first to reach the promised land? This makes me think about the immigration issues we discussed at the beginning. Our country never had a great immigration policy for any kind of immigrant, to be perfectly honest. We had enough prestige and resources to attract the brightest in the world. But when that chilling effect sets in, it could be a problem for the long haul, while the scouting ecosystem or drafting source on the China side seems sufficient to put together a DeepSeek, if nothing else.

Peter Harrell: Can I ask you a question, Kevin? It’s not an area I know much about. Is Europe anywhere on the map with AI these days, or is this really remaining a U.S.-China competition?

Kevin Xu: This is very much a U.S.-China competition, particularly because of the talent question — not so much about the money. Jensen went through a Europe swing just a couple of weeks ago and announced a bunch of sovereign AI projects in the UK, France, Germany, Slovenia, and some smaller countries. The governments are generally pretty good at building hard infrastructure from the ground up and signing really big checks when everyone comes together politically. That’s what governments are good at.

But when it comes to organic talent development, most European top-end talent either moves to the U.S. if they can — because they find the European system generally stifling if you want to move fast — or they work locally at the labs of American companies that are building locally to recruit there. The three most recent successful Zuckerberg recruits from OpenAI were from the Sora Lab of OpenAI. Is that European talent or American talent? You tell me.

Matt Klein: DeepMind is UK, right? They’re owned by Google, but —

Kevin Xu: That’s the one shiny example.

Jordan Schneider: Does it count anymore? They were a subsidiary as of 10 years ago. It’s an interesting question, all this sovereign AI buildout. It almost reminds me of Chinese provinces that spin up their own data centers and proudly buy some hardware chips. What are we doing here?

Another aspect I wanted your take on, Kevin: there are rumors that Meta was going to bow out of the open source game. In that case, we’d have no 10-figure companies in the U.S. really pushing on open source, basically seeding the entire ground to Alibaba and DeepSeek. What does that mean from a 10-year perspective?

Kevin Xu: First, Meta’s version of open source — LLaMA — since we do have a lawyer in the room here, is actually from a legal perspective less open than all the Chinese open source models currently, from a legal licensing perspective. Meta came up with its own version of an open source license that was hotly debated within the tiny community of open source licensing lawyers that I’m part of.

Part of those restrictions is that once you use LLaMA’s model up to, I believe, 700 or 750 million monthly active users, then you need to engage with LLaMA’s parent company to have a commercial licensing agreement. That’s an important nuance to start with.

From that perspective, and I’ve heard this from other sources too, there was at least an expectation from LLaMA at some point to still make money directly from its open source models. If it gets competitive enough — whether it embeds into another coding agent like OpenAI and Anthropic, or some other B2B or enterprise product — they had some expectation of that. They weren’t just going to throw a ton of money and talent into this theoretical financial dark hole and do something good for the world, even though they were falling behind.

The fact that this was even leaked as an internal discussion point — I don’t know if they’re really going to do it — but the fact that serious people inside Meta have considered abandoning the open source part of LLaMA and actually abandoning LLaMA entirely in favor of just the best model in the market has so many layers of irony. We have the open society here in America, the open this and that, freedom and whatnot, but we chose this most singularly transformative technology direction to be as closed as possible among the companies.

You can imagine a lot of siloing going on. A lot of things that probably could have been shared but aren’t shared for one reason or another. That actually blunts progress on the long-term horizon, as well as the diffusion pace of certain technology, where open source is just the best way to diffuse a lot of technology — even if it’s the worst way to make money off that technology. That might be the reason why a lot of U.S. companies are choosing the closed source path, while the Chinese side is right now leading in all the open source model benchmarks.

Peter Harrell: Jordan, it’s a question for you. When we think about “Is America cooked,” we often think about this in the U.S.-China context.

Peter Harrell: Has China done anything right over the last three months?

Kevin Xu: I’ll jump in here real quick. This is actually related to the dollar again — speculative because it’s very recent. They did have a recent high-level finance summit in Shanghai, the Lujiazui Forum 陆家嘴论坛, where the governor of the People’s Bank of China and all the relevant high-level financial regulators came and gave speeches.

What came out of this specific forum this year is a concerted willingness to really promote the RMB in ways and places where the dollar is falling short for one reason or another — not to necessarily replace the dollar as the global reserve currency, because that’s a whole other can of worms that I don’t think anybody in China really wants to be. But in countries or trade directions where having the RMB be the settlement currency in lieu of the dollar, or being able to buy global commodities in large chunks using the RMB versus the dollar, is advantageous, there is now a policy appetite.

This is very much a reaction to Liberation Day, to the way that the U.S. instigated global trade war has unfolded. That prompted this. How far will this particular policy direction go? Nobody knows at this point — it’s way too early. But the appetite is there now to be a little more proactive about making whatever reforms necessary to make the RMB more attractive in certain situational settings when the dollar is becoming a little less appealing.

On Allies and Immigration

Jordan Schneider: How bad is Trump going to end up being to the allies? Q1 was probably pretty close to a worst-case scenario with those really obnoxious speeches in Munich. But since then, we had Haley do a trip to Asia where she basically said things that a Nikki Haley administration would have said.

What does the Iran bombing mean for America’s commitment to Asian allies? On one hand, it shows that Trump’s willing to bomb stuff, but on the other, it’s bringing us back to the Middle East. Maybe that’s a bit of a wash.

There was also talk of some crazy DOGE cuts to the Pentagon, but there are basically going to be no defense funding cuts. Whether they’re actually going to reform defense acquisitions and buy stuff we need to buy is an open question.

Even though we’ve had some serious U.S.-Japan weirdness — with Japan being sick enough of the Americans to cancel a two-plus-two dialogue — I still think we could be on a worse timeline than the one we’re on now with the way Trump has treated Asian allies.

Matt Klein: If you really want to be optimistic, you just say the way that the administration is now treating European allies — that’s where you’re seeing the biggest shift, maybe. We’re friends with Zelenskyy all of a sudden. Who knows how long that will last, but that is notable. The European NATO countries seem to have managed to figure out how — with the exception of Spain — to put together a deal that looks like it’s getting everyone on the same page. Who knows how that will play out longer term, but if you want to be encouraged, that’s definitely somewhere to look.

Kevin Xu: Isn’t that just giving Trump what he wanted?

Peter Harrell: That is part of it.

Matt Klein: You reframe it. He said he wanted five percent, and they said three and a half percent plus one and a half percent other, which is fine. Everyone agreed on it. Arguably, it’s good for them too. But the shift with Zelenskyy is interesting, as Jordan said. Trump met him at the NATO summit and said they had a great conversation. Now he’s saying, “Supposedly, Putin called and I need help with mediating with Israel, and I need help with you.” Who knows?

Jordan Schneider: It’s hard to know what’s really going on there. But if you want to be optimistic, the thing with this stuff is that it was unclear if Trump was going to be doctrinaire about these things or just vibrate to weirder places than any American president had gone to, then vibrate back to a more mainstream place and keep swinging. But him staying at “No, we’re leaving NATO, this is a terrible thing" — that’s not something that’s happened. He has shown the capability to get pissed off and fed up by people like Putin, which is something you may not have entirely priced in April of this year. I consider that positive.

Although I’m the Trump optimist here.

Matt Klein: To be fair, it’s also not as if we’ve actually adjusted our Russia policy in a way to imply that he’s actually gotten upset with them. So far, it’s just words. But we’ll see.

Jordan Schneider: I’ll take words.

Matt Klein: Better than the other words.

Jordan Schneider: They’re a start.

Peter Harrell: I very much agree with the point Matt made. Obviously, the words coming out of the NATO summit and the words around Zelenskyy are much better. The lived experience over the last six months is that words change. The thing I’d note in the other direction: two or three weeks ago, Trump was saying nice things about Canadian Prime Minister Carney. Right before we got on this call, he’s now back to saying Canada’s the worst and we’re going to tariff them all over again. Nothing really stays stuck with our current president.

Jordan Schneider: But if things don’t stay stuck on the dumbest timeline, then that is — again, it’s more the Brexit arc, less the actual America self-owned catastrophe one, which is okay.

Peter Harrell: Maybe I should be worried here. The other thing I’m watching to see how it plays out — conceptually related to immigration but also distinct — is where we’re going with our universities. To pick up on a point Kevin made earlier, it’s not that we always had the best immigration policies. American immigration policy has been totally dysfunctional for my entire life — I’m 45 — but we have had, against a backdrop of dysfunctional immigration policy, some very important bright spots. The ability to attract truly great talent to some of the world’s best universities who then stay here and found and run businesses and contribute to scientific innovation.

Trump’s war on the universities has now spread, at least with Harvard, beyond just cutting off funding to trying to cut off their ability to enroll foreign students. It appears that he may be trying to get the University of Virginia to fire its president. I’m not going to defend all kinds of crazy things universities have done over the last five to eight years — there have been plenty of challenges with American universities — but they are a huge source of long-term American strength from a scientific innovation perspective, from a talent attraction perspective, from educating people who then go out and do great things.

It’s looking like a fairly full-out war and not just something that he’s looking for a détente on. There’s a ton of institutional strength in American higher education. I have bias here because both my parents were full-time academics all my life growing up, so I’m biased toward American educational institutions. I worry that if we structurally harm our higher education system, that would be another thing — cutting scientific R&D — that helps us be cooked longer term.

Matt Klein: That’s part of how we’re hurting it. They get a lot of that scientific R&D funding, and if that’s cut off — and also, this is not just a Trump obsession — in the bill that currently passed was this punitive tax on endowment income for universities. For universities that have enough endowment income for that to matter, that pays a very large chunk of their operating expenses. Most of the rest comes from things like government grants for research.

If you cut those things at the same time, they’re going to go out of business or they’ll have to drastically change what they do. Again, this is not something that’s really getting a lot of pushback because people pay a lot more attention to other things — the tariff on this is going up by this amount or whatever — but these other things are still steadily going on despite the fact that people’s attention isn’t focused on that. As you said, it could be a relatively benign Brexit outcome, but it adds up and would be very harmful.

Kevin Xu: That will never show up in these more short-term twitchy indicators, whether it’s the market, the dollar, or any of that stuff. Two points I want to add to what Peter said — which probably makes the best case for long-term cookedness if there’s no really forceful reversal of what’s happening right now, but hopefully some actual tangible improvement on immigration — not just keeping together whatever we had left prior to this presidency.

First, the way in which a lot of these scientific research grants are cut isn’t just a problem because they’re cut from an absolute value perspective, but they’re cut from a “Control-F” perspective. You go through the grants, you Control-F for words that the administration doesn’t like for entirely ideological reasons, and you cut them. That is the reason a lot of the top academics and researchers who came to the United States from other countries specifically came to escape from.

Matt Klein: Right, like “neurotransmitters” because it has “trans” in there. Now you’re saying you’re not looking at Alzheimer’s research.

Kevin Xu: Exactly. The word “diversity” just trips up a bunch of alerts from this Control-F way of thinking about scientific research when the word “variance” could have replaced “diversity” and that grant could have been safe. All these very blunt instrument ways defeat not just the fact that we get to have this innovation happen in our country, but that the people who come to this country to do the research and want to stay — they didn’t just come for the money. Nobody in academia, even at Harvard or Princeton, really celebrates the amount of salary they get to make, even at the highest level. They came here because they don’t have to be worried about being Control-F’d in their own country of origin.

The second point that builds off of that — which is really the long-term cook scenario — every time we have this American declinism conversation (this is not the first time we’ve had this conversation as broader public discourse) the most common way that we patted ourselves on the back or were able to go back to sleep after the conversation was immigration. We still had immigration. All the smartest people in the world are going to save our ass by coming here. That may not be an answer we can just default to anymore. That’s probably the one thing that would put you on the cook ledger in the 10-year timeframe — certainly not a quarter-by-quarter timeframe, because who knows what the market is going to do.

To come back to immigration today — we’re actually selling Trump cards. We’re selling a spot in the best country in the world, quote-unquote, to the highest bidder, not the smartest people. What kind of people do you attract from that? I’ll let the listeners have their own imagination over what those people could be.

Matt Klein: It’s worth noting that the U.S. university system has been so dominant and so good for so long that we assume it has always been that way. But that’s not actually true. Before 1933 — which is a significant date to think about — the U.S. university system was not particularly competitive globally. After that, it was. English wasn’t even the dominant language among academics — it was German. It’s not inherent that it has to always be this way. You’d have to work hard at it, but if you’re committed enough to self-harm for ideological reasons, then you can get rid of these long-term advantages.

Peter Harrell: Another topic that the Trump administration needs to prove it can handle — and it’s an open question — is supporting U.S. manufacturing and some of these strategic industries and products it talks about. There’s also a lot of bipartisan support for the idea that we should have at least some onshore or ally-shored critical minerals production, shipbuilding, and these strategic lower-tech industries.

The administration has certainly talked a good game about some of this. They’ve put out executive orders, they’ve said they want to reduce permitting hurdles to make it easier. It seems entirely up in the air whether they’re actually going to be able to achieve any of this manufacturing renaissance that they’re talking about.

This is going to require a more integrated approach than what we’ve seen from them to date. That’s a big X factor for me: Are they going to be able to actually put this together or not? Will it be in the industries that actually matter to us? All due respect to Howard Lutnick, I’m not sure “screwing tiny, tiny screws into iPhones” is really the industry we want to be focusing on in terms of our manufacturing.

Jordan Schneider: Are there any data points that are encouraging on this line? We’ve seen various efforts to gut the DOE loans program. The CHIPS Act has been exploded.

Peter Harrell: If you look at what they’ve actually done with the money, the DOE loans program was focused on clean energy, and they don’t want clean energy. That seems to be fizzling. There’s anecdotal evidence — Matt, you may have plugged into the data more — that some of the big announced manufacturing investments in those sectors have been put on pause. There’s a slowdown there.

While Trump’s talking about things like shipbuilding, we’re not actually seeing shipbuilding. Let’s put aside the climate impacts of oil and gas and current energy prices — I’m not sure you’re going to see a huge oil and gas boom.

Matt Klein: We haven’t. The joke about “now we’re going to unleash drilling” — you should talk to the shale drillers. They don’t want to do that. Briefly, when Brent went close to $80 because of fears of Iranian activity in the Strait of Hormuz, if it had stayed there long enough, maybe you would have seen more shale. But between the fact that their costs have gone up tremendously because of tariffs on steel — what do you think the pipes are made of? — and the fact that the price of the product they sell has gone down, plus labor costs are still going up over time, it’s not super compelling.

Investment in manufacturing peaked in the middle of 2024. A lot of the growth in investment and manufacturing capacity in this country had been directly attributable to CHIPS and IRA — semiconductors, batteries, things like that. If that’s what the incremental investment is coming from, and then those sources of incentives are being gutted as quickly as possible, it’s not surprising that you’re going to see that move in reverse.

Some of the money may already be out the door enough that it won’t be pulled back on projects that have already started. But it’s not super encouraging. The Loan Programs Office was talking about clean energy, but they were also talking about doing it for critical minerals. Those minerals have green energy-related uses, so you could say, “Oh, we’re going to get rid of it.” But to the extent that they say those things are strategic — whether it’s magnets or whatnot — if you cut the funding, it doesn’t matter whether you call it green energy or not. You still are not going to have the resources.

It’s going to be very difficult to have a domestic industry in those sectors be competitive without some kind of government support because they’re so cyclically volatile on price and because the main producers right now are in China, where market-based constraints are not as severe. Because the capacity is there, they can basically flood the market when they want to. If any seemingly viable competitor shows up — which is why you’ve had such problems with lithium in the U.S. in the past — you need some kind of government backstop to prevent that. That was being done out of DOE.

It could change, but it’s not clear to me there’s strong interest in doing that. I don’t know why a manufacturing renaissance would occur. On top of which, if part of the rationale for a manufacturing renaissance is that these tariffs are going to make U.S. production for the U.S. market more competitive, but then no one knows what the tariffs are going to be — either because they’re illegal or because they’re going to get negotiated away for unclear gains — why would you make a commitment now? It doesn’t pencil out. You want to use a pencil, not a pen, because you can just change.

I don’t see it. It’s not showing up in the data that we’re having any kind of manufacturing renaissance. Far, it’s the opposite. It’s not like it crashed or anything, but it’s not clear to me why we would see a big change given what we know so far.

Kevin Xu: This is a question for Matt, but obviously anybody can jump in. What are the things that got some play but did not get nearly enough play during the DOGE stuff? The cutting of staffing or investment in places like the Bureau of Labor Statistics and BEA. Basically the only neutral arbiter of actually accounting for how cooked we are or not as a country — hard data, not soft data, not live data, not sentiment data.

That’s another thing that just confounds me. If we were to actually answer this question from a statistics perspective or in a rigorous way a year from now, two years from now, whatever — I don’t even know if I could really trust any of the GDP numbers going forward, any of the non-farm payrolls going forward. The government pumps out so much good data and it takes so much time to build up that infrastructure, but you need one bad cut to completely get rid of that discipline.

Subscribe now

Matt Klein: One thing I would say is some of that actually precedes even the election. It had to do with some of the budget deals that were being done in 2023-24. I don’t know why these were agreed to, but basically — I’ve talked to the BEA people and seen they announced tables have been discontinued because they’re not in the budget for it. This was over a year ago. That’s unfortunately been an ongoing problem.

The BLS, which compiles things on inflation and labor market, has been having issues for a while — not necessarily because of underfunding, though that’s probably related to it, but just because people don’t respond to surveys the way they used to, for who knows why. You take a survey of employers every month, figure out how many people are working, and then aggregate it up to give an estimate. But if the employers are not answering your call for whatever reason, or fewer of them are proportionally, that creates potential problems. That was already an issue.

It’s compounded by cutting people arbitrarily. There was an announcement in the past few months — the Commerce Department, which runs the BEA and the census (not BLS, that’s in the Labor Department), had this outside body of advisors, mostly academics, to judge the quality of federal statistics. They just decided to cancel the meetings. Who knows? Maybe they just decided we don’t need the meeting. But it’s the kind of thing where if you were to say, “Okay, this would be the first step if you wanted to fudge the numbers."

I don’t see any evidence that’s what’s going on so far. I don’t want to impugn any of the numbers or any of the work that’s being done now. I don’t see any evidence of that yet. I don’t even know what it would look like, but I agree it’s something to be worried about. Just the fact that we’re talking about it is indicative of the kinds of people who are in charge and what they’ve generally shown in their attitude toward things — that this is a possibility, that it could be happening, even if it hasn’t actually happened yet.

Jordan Schneider: It would be pretty remarkable if we’re in a world where we start treating American numbers like Chinese numbers. There’s the intel community stuff of fudging judgments and people leaking out what the real numbers are. I can imagine that happening also in BLS pretty quickly, where there’s enough of an institutional allergy to this that you hear pretty fast within that org of any of these types of shenanigans happening.

Matt Klein: I had two things. One is that far there are independent ways of verifying inflation stuff. The Billion Prices Project was invented by an Argentine working in the United States as an academic economist — very appropriately — to figure out alternative measures of tracking inflation. They just look at online listings. Far for the U.S., you say, “Okay, what are the prices of a basket of goods that matches the CPI basket of goods based on online listings?” And it’s tracked perfectly so far. That’s encouraging.

On the Chinese side, I guess you might know more than I do. My understanding is the headline GDP numbers may or may not be massaged. The underlying stuff is probably the good faith best estimate. But if they don’t like something, they just stop publishing it. It’s not even that they make it up — they just stop publishing the number. That seems like that actually could be an issue, especially given you have budget cuts anyway.

Jordan Schneider: There’s enough money in getting good numbers on U.S. employment that people will just invest in getting this right.

Matt Klein: It might not be public. There’s value in it being publicly available.

Kevin Xu: It’s locked up in Jane Street’s data center somewhere.

Peter Harrell: Bloomberg will add it to the terminal for an extra $500 a month.

Jordan Schneider: Does anyone have recommendations? Anything you’re reading or enjoying?

Peter Harrell: I just finished reading a book from 15 to 20 years ago by a Wall Street Journal reporter about the rise of the Texas oil families from the 1930s to the 1950s — the wildcatters like the Hunt family and similar figures. It’s fascinating because there are so many parallels, both culturally and psychologically, to the people who have made fortunes in tech over the last 20 years.

These were individuals doing innovative things with new technology who made tremendous amounts of money and then started converting that wealth into political influence. They began buying up media — literally purchasing radio stations at the time. It was quite striking to see how these people started out drilling oil wells in the 1920s and eventually came to build or at least fund the modern conservative political movement in the 1950s. It’s a very interesting piece of history.

Kevin Xu: I started reading Inside the House of Money, which is a collection of interviews with macro hedge fund managers, conducted by another macro hedge fund professional. The interviews are very intimate and revealing.

One of the interviews features our current Treasury Secretary, Scott Bessent, from when he was running his own fund right after leaving Soros. Apparently, going into this interview, he was extremely sleep-deprived — he hadn’t slept for days — and was worried about something that might happen in China. He didn’t specify what, but that was the context for his interview.

This book was written between 2004 and 2005, and I still haven’t figured out what was happening in China at that time that kept our future Treasury Secretary awake at night. If anyone can help me solve that mystery, I’d appreciate it. Regardless, the book is a really engaging read if you enjoy understanding the mindset of global macro traders.

Matt Klein: I don’t have anything quite as sophisticated, but I recently started watching Your Friends & Neighbors, and I’ve been enjoying it tremendously. I also bought The Party’s Interests Come First on your recommendation, though I haven’t started reading it yet. It sounds excellent.

Jordan Schneider: I have another recommendation that will preview a future China Talk episode — To the Success of Our Hopeless Cause: The Many Lives of the Soviet Dissident Movement. It just won the Pulitzer Prize.

It was actually amusing because I booked the author before he won the Pulitzer, and then we had to postpone our interview because he suddenly had major media obligations to attend to. The book is beautiful and literary, with incredible sources. You have all these transcripts from court cases where these dissidents are essentially mocking the prosecutors and judges, taking them to task at every turn.

While it doesn’t necessarily tell you much about this moment in American politics, it provides a fascinating window into the world of these very brave and frankly eccentric people who made remarkable choices in the 1950s, 1960s, and 1970s in the Soviet Union. This was after the Stalin era, so they weren’t going to be executed, but they were still putting an enormous amount on the line to confront the regime.

Our “cooked level” remains undetermined. See you all in September.

Mood Music

How does Russia prevent uprisings, and what can other authoritarians learn from Moscow’s methods of coup control?

For the second anniversary of the Wagner uprising, ChinaTalk interviewed London-based historian Kamil Galeev, who was also a classmate of Jordan’s at Peking University.

We discuss…

• Why the Wagner Group rebelled in 2023, and why the coup attempt ultimately failed,

• How Wagner shifted the Kremlin’s assessment of internal political challengers,

• Similarities between post-Soviet doomerism and the American right,

• Historical examples of foreign policy influenced by a victimhood mentality,

• Barriers to Chinese hegemony.

Listen on your favorite podcast app.

Today’s post is brought to you by 80,000 Hours, a nonprofit that helps people find fulfilling careers that do good. 80,000 Hours — named for the average length of a career — has been doing in-depth research on AI issues for over a decade, producing reports on how the US and China can manage existential risk, scenarios for potential AI catastrophe, and examining the concrete steps you can take to help ensure AI development goes well.

Their research suggests that working to reduce risks from advanced AI could be one of the most impactful ways to make a positive difference in the world.

They provide free resources to help you contribute, including:

• Detailed career reviews for paths like AI safety technical research, AI governance, information security, and AI hardware,

• A job board with hundreds of high-impact opportunities,

• A podcast featuring deep conversations with experts like Carl Shulman, Ajeya Cotra, and Tom Davidson,

• Free, one-on-one career advising to help you find your best fit.

To learn more and access their research-backed career guides, visit 80000hours.org/ChinaTalk.

Why Russians Don’t Rebel

Jordan Schneider: Kamil, can you explain the Wagner insurrection that happened two years ago in Russia?

Kamil Galeev: The Wagner coup attempt in summer 2023 was perhaps the lowest point for Russia during the course of this war — the biggest moment of internal instability. It made a profound impression internally and changed the priorities and policies of the ruling regime.

How did the coup attempt become possible in the first place? Before discussing that, we need to examine why coups haven’t been more common in Russian history prior to the Wagner insurrection.

The primary reason is that since 1917, the modern Russian state has carried direct, uninterrupted institutional continuity. Since the October Revolution, the Russian Armed Forces have been heavily co-opted, which is easy to explain.

Imagine you’re Lenin, Trotsky, and others taking power in 1917. You’re constantly comparing yourselves with revolutionaries of the past, most importantly the French Revolution, but also the English Revolution. What pattern do you see?

You see English parliamentarians taking power from the King and waging war. As they wage war, they create an army, a military force that eventually overthrows them. The same happens with the French Revolution — National Assembly deputies take power and wage war, but the powerful armed force they create ultimately overthrows them.

The pattern they observed from history is that the English Revolution ends with Cromwell and the French Revolution ends with Bonaparte. This became their primary fear for their own revolution.

They took massive, systematic precautions against this outcome. In Trotsky’s memoirs, My Life, he describes discussing an officer of the Red Army called Blagonravov, who performed admirably. Lenin had a strong reaction when Trotsky remarked that Blagonravov, if translated into Italian, sounds like “Buonaparte.” This may seem paranoid, but it demonstrates their deep concerns.

“‘Out of such a lieutenant,’ I had once said jokingly to Lenin, ‘even a Napoleon may come some day. He even has the right name for it: Blago-nravov, [In Russian this means “good-natured” or “good-mannered.” - Trans.] almost like Bona-parte.’ Lenin laughed at this unexpected comparison, then he grew thoughtful, and, with his cheek-bones bulging even more, said very seriously, almost threateningly, ‘Well, I think we’ll manage the Bonapartes, don’t you?’”

~ Leon Trotsky, My Life p. 317

When building the Red Army, they were constantly obsessed with the possibility of Bonapartism, scanning for potential Bonapartes and weeding them out. From the beginning, the Red Forces, from which the Russian army naturally inherits, were extremely optimized to prevent any coup attempt.

This explains what the KGB was all about. You don’t build an enormous, all-permeating, all-powerful state security apparatus merely to fight dissidents — that would be overkill. You don’t even build it just to suppress rebellions. You build the KGB to infiltrate the army, control it, and if necessary, suppress armed insurrections. It’s essentially a second internal army to control the first.

Since 1917, Russian military development has been heavily optimized for this specific purpose. All armed forces, either in the Soviet Union or the Russian Federation, were heavily infiltrated by state security agents with double and triple layers of control. The system was designed so that, first, you cannot prepare an insurrection in secrecy, and second, you cannot execute one because these layers of control would be activated.

State security agents were embedded in every unit, regiment, and army. They would simply eliminate a general who attempted anything subversive. This system worked effectively in preventing military coups, which may explain why the Russian state remained internally stable through the 1990s, despite poverty, lack of resources, and extreme discontent within the military. Many officers were upset, but taking action was impossible.

The interesting thing about Wagner is that it may have been the first military unit since 1917 that avoided this state security infiltration, escaping these double and triple layers of control.

Wagner famously recruited people with various backgrounds — former military officers, police officers, customs officials, drug control agents, and prison guards. Interestingly, they even recruited ethnic Russians from the French Foreign Legion. However, they didn’t recruit state security personnel for obvious reasons — if you start recruiting former state security officers, you’ll quickly be completely infiltrated by current state security officers.

Why was this allowed to happen? It seems like a completely irrational decision for the political leadership. To answer this question, we need to view it not as a single decision but as a sequence of consecutive decisions.

The creation of this large, non-infiltrated force without checks or control layers didn’t happen instantly. It evolved over time, iteratively. Originally, Wagner was a small unit of mercenaries, primarily for special operations far from Russia’s borders — in Syria, Libya, and eastern Ukraine.

A member of the French Foreign Legion displays Wagner-affiliated patches, December 2022. Source.

Early Wagner was small — dozens, then hundreds of fighters who were usually far away, sometimes thousands of miles from Russian borders, with some veterans in reserve at home. These reservists lived normal lives in Russia until called upon. At any given moment, the force was small and mostly distant, which allowed it to largely bypass direct control, partially because it wasn’t considered important.

The enormous expansion occurred only in 2022. Before the Ukraine war, Wagner was essentially an assassin group or hit squad. When the initial invasion failed massively, the political leadership tried to improvise new tools, mobilizing whatever resources they had. At this point, Wagner was authorized for massive expansion.

This wasn’t part of the original plan. When Russian troops crossed the Ukrainian border, Wagner wasn’t there. After the initial invasion failed, they mobilized whatever they could — they called reservists, recruited other people, and soon began mobilizing prisoners.

Before long, they went from having a few hundred fighters mostly in Africa or Syria to tens of thousands relatively close to Russian borders. As this happened rapidly, normal control mechanisms weren’t established. The most basic control would have been oversight by state security officers and possibly the Federal Protective Service, who are essentially Putin’s bodyguards.

When the political leadership eventually tried to impose stricter control by integrating Wagner into the normal military structure, the uprising began.

My point is that although creating a large, uncontrollable structure seems irrational for the leadership, that’s not how it happened historically. First, they created a small structure where additional control layers seemed unnecessary. Then, when the situation deteriorated dramatically, they had to improvise with tactical responses to tactical problems, dealing with consequences later. This happened through many iterations, not just one.

Wagner’s rebellion represents perhaps the greatest internal threat to the regime in decades. Not only during Putin’s rule, but including through Yeltsin’s era as well. There hasn’t been a military rebellion of this scale probably since the October Revolution. It was the most serious internal political threat. The question is, why did it evaporate and turn into nothing?

There could be several valid answers to that question. One possibility is that the Wagner group never actually intended to take power. It was most probably a negotiation attempt rather than an effort to overthrow the political leadership and replace them. It appears to have been an attempt to renegotiate their position, perhaps as a reaction to being taken over. They tried to avoid the imposition of additional control through a show of force. It didn’t work out, but they likely never aimed to overthrow the leadership from the beginning.

This may actually be a common occurrence. Many rebellions and revolts don’t initially aim to overthrow leadership but rather to make specific demands. However, in many cases, these still bring disastrous consequences. Why didn’t this one?

Another valid explanation is that they were positioned too far away. Were there any Wagner garrisons in the Moscow region? Had there been, events might have unfolded differently.

The main reason, however, is that the dirtiest secret about military coups is that successful ones are typically invited. In most cases, it’s extraordinarily difficult to execute one without invitation. The creation of Wagner as an independent and uncontrollable armed force didn’t happen overnight but resulted from an accumulation of many tactical decisions over time.

The common image of military officers secretly preparing and then executing a lightning strike to seize power is largely fictional. In most functional regimes, this approach simply wouldn’t work. For the military to successfully take power, they typically need to be invited or solicited.

Jordan Schneider: Kamil, would you apply a Nazi Operation Valkyrie analogy here? How would you classify that within the solicited versus unsolicited framework?

Kamil Galeev: Consider the Eighteenth Brumaire — probably the most famous and iconic military coup in history, when Napoleon Bonaparte came to power. The important understanding here is that it didn’t happen in a vacuum but culminated a long process.

Since at least the Thermidorian Reaction, the political leadership in France had been using the military to control political processes. They deployed troops to deal with armed rebellions in Paris or popular mutinies. When afraid of losing elections, they used troops to “correct” electoral procedures and break the opposition. When wanting to purge political opponents from parliament and the National Convention, they used troops.

This happened repeatedly. Early in the French Revolution, the use of force was disguised as popular rebellion. Eventually, it became undisguised military force. With each passing year, the government ruled increasingly through explicit military force.

The problem was that as this progressed, the leadership became more dependent on the military. When Bonaparte took power, the coup wasn’t even his idea. It originated with Abbé Sieyès, a political leader who wanted to achieve personal political goals using the military — just as the political leadership had done many times before. This last time, however, it backfired.

What we see isn’t a single coup but many “mini-coups” where political leadership repeatedly corrected political outcomes through military means. In doing so, they gradually invited the military to take power. Examining history, we could find many more examples of this pattern.

Jordan Schneider: The Nazi analogy might be more appropriate than the 18th century Brumaire comparison. There was a war going poorly, with generals realizing their leader was unstable and deciding action was necessary. Many senior Nazi leaders supported the plan, but it failed simply because the bomb exploded on the wrong side of the table.

Kamil Galeev: That’s an interesting comparison with Wagner, noteworthy for both similarities and differences. One key distinction is that the officers who attempted to assassinate Hitler were pursuing what they perceived as a common interest, not personal gain. Many of these officers actually owed their careers to Hitler, so they weren’t motivated by professional setbacks.

The Wagner coup attempt, however, clearly pursued particularist interests. They were essentially protesting against being taken over and demonstrating their force. The first case involved belief in a common good — the second lacked this motivation.

Another obvious difference is that the Nazi coup plotters actively tried to eliminate Hitler and arrest Goebbels, attempting to seize political leadership directly. This wasn’t really Wagner’s approach. Their action resembled a negotiation more than a takeover. The officers who tried to kill Hitler weren’t seeking concessions or privileges, while Wagner was.

The third difference is that Nazi officers were part of the regular military structure, while Wagner operated as an irregular mercenary force. Many Wehrmacht officers could be seen as part of established elites predating Hitler, whereas Prigozhin and his organization were entirely Putin’s creation.

This last point is particularly important because, unlike the Nazi officers within the normal military hierarchy, the Wagner phenomenon only became possible through Yevgeny Prigozhin’s personal connections. Though a junior member, he belonged to the same network as Putin. The creation of this force didn’t happen overnight but through several iterations, succeeding because Prigozhin was integrated into the same contract network.

In Russia, there exists an informal personal network of Putin’s longstanding friends from St. Petersburg, including oligarchs who rose to prominence under his rule, such as the Rotenbergs. Prigozhin appears to have been a subordinate of the Rotenbergs, placing him just one step removed from Putin — a trusted man of Putin’s trusted man. This position gave him access to the sovereign’s ear and allowed him to be entrusted with matters outside normal processes.

One thesis I want to develop later is that power is often misunderstood. Power largely represents one’s social metric status within the contract network. In highly centralized systems, politics revolves around gaining the ruler’s ear at any cost. I believe we’ll increasingly see elements of this in the United States as well, where informal contact with the supreme ruler will matter more and cost more.

Infected by Victimhood

Jordan Schneider: Kamil, how did this experience change Putin and the regime’s calculus going forward?

Kamil Galeev: Somewhat paradoxically, the regime appears to have become relatively more relaxed about civilian threats. After the coup, there seems to have been a recalibration of what constitutes a threat. Dissident groups were effectively downgraded in threat assessment, while state security and intelligence services focused on dealing with Wagner’s aftermath and monitoring potential opposition from armed entities, including those within the regular military.

Essentially, following the coup, significantly greater emphasis was placed on monitoring potential threats from individuals with weapons, regardless of their agency affiliation.

Jordan Schneider: Kamil, what impressions do you have of what’s happening in America right now?

Kamil Galeev: One impression — and this is not so much an analysis as a genuine observation — is that it feels oddly familiar. I grew up in Russia, born in 1992, and much of the Russian atmosphere of the 1990s and 2000s was characterized by a pervasive feeling of resentment. There was a widespread belief that we had been screwed over, ripped off, and victimized. This mentality fueled a desire for revenge in some form.

This feeling of resentment and victimhood has been the major cause of what outsiders might perceive as irrationality in Russian foreign policy. When you feel like a victim, when you believe you’ve been abused and treated unfairly, you often declare a crusade against a cruel world that has mistreated you. This is a common occurrence both among individuals who feel victimized and start attacking everyone around them, and among nations.

The deep resentment and sense of defeat in Russia and post-Soviet countries made them behave in counterproductive ways. This was somewhat explainable because Russia lost the Cold War and consequently acted as a sore loser, which typically means acting irrationally.

Now I sense the same feeling of resentment and a similar atmosphere in America. There’s this notion that “we have been mistreated, we have been ripped off, and now we’ll take revenge.” This is much less comprehensible to me because America didn’t lose — it actually won. I can understand how this sentiment emerged in Russia, but it’s much more difficult for me to construct a functional model of how it developed in the United States.

Consider the trade war, for example. You could argue that decoupling from China serves American interests, which may indeed be the case. However, it doesn’t appear that America is simply decoupling from China — it seems to be attempting to decouple from the entire world simultaneously. This doesn’t resemble a reasonable or rational plan; it looks more like an emotional reaction, an attack against everyone at once.

The only explanation I have is that this behavior stems from a sense of victimhood, which leads to attacking indiscriminately. Within the American conservative movement particularly, I observe an incredible feeling of resentment. Yet it remains puzzling to me how this emerged as a mass phenomenon in the first place. America is the richest large country in the world, yet a big portion of the population sees themselves as victims, abused by everyone.

Jordan Schneider: That’s the world we get to live in now.

Kamil Galeev: Yes, but again, it’s a puzzle I don’t quite understand. In the UK, where I live now, there’s a sense of decline in the standard of living — not necessarily in absolute terms, but in relative terms. Not so long ago, the UK was much wealthier than almost any continental European country, particularly before World War I or even World War II. That’s no longer the case.

But the American situation feels especially puzzling. Perhaps one takeaway is that it’s very difficult to monitor the sense of victimhood within yourself, because once you become infected by it, you begin undermining yourself through your own actions.

Jordan Schneider: Are there any historical examples of victimhood mentality unwinding itself and a new conception or national purpose taking its place?

Kamil Galeev: I’ll avoid modern examples as they would generate too much controversy. Instead, I can offer some more distant historical ones.

We often discuss the problems and madness of Nazism and Nazi decisions, but it’s difficult to ignore that much of German foreign policy was established long before the Nazis rose to power. Consider how hostilities with the United States developed during World War I with the Zimmermann Telegram. Germany sent a telegram offering Mexico a portion of the United States. Even though the British intercepted and published it, few believed it because it seemed so absurd. Then the German ambassador confirmed it and repeated the threats, which led to hostilities with America and ultimately contributed to Germany’s defeat.

This may be the most illustrative example, but when examining German foreign policy before and during World War I, it appears they were actively collecting enemies worldwide. Finding a rational explanation for this behavior is challenging, but this pattern was evident in early 20th century Germany long before the Nazis came to power. This mentality certainly made their policies, particularly foreign policy, much less effective.

To condense it into one formula, if you perceive yourself as a victim, ripped off by a cruel world, then the entire world becomes your tormentor. Launching a crusade against the whole world might sound like a perfect plan, but it inevitably leads to conflicts you cannot possibly win.

Jordan Schneider: Kamil, would you like to share your reflections on China?

Kamil Galeev: If we were living in a conventional world, comparable to 100 or 200 years ago, I would say China appears to be the next world hegemon. This is partly because China has taken over much of global material production and is rapidly improving and expanding into other sectors. In almost every sphere of the material economy, and increasingly the digital economy, Chinese companies are outperforming nearly all competition. Typically, this is how a future global hegemon behaves.

However, we now live in a fundamentally different world from the past, most notably because it is aging and will soon be shrinking quantitatively. When the Netherlands became the economic hegemon of Western Europe, they were booming and expanding. It was a young, rapidly growing country. England during the Industrial Revolution and throughout its 19th-century hegemony was also young and quickly growing. The small manufacturing cities of England were booming with people, particularly young people. The same applies to America in the 20th century.

Now we’re entering an interesting phase where most industrial countries will be shrinking quantitatively and, more importantly, aging. How this will unfold in this new era is unclear. Perhaps the most serious problem facing China is that it appears to be aging faster than other nations.

The most logical solution for them would be to ease their stance on immigration, bringing in more people — and to be cynical and candid — particularly facilitating the inflow of women. The fact that they haven’t taken steps in this direction leads us into unknown territory. I don’t fully understand what will happen when the number of young people falls below a certain minimum threshold.

Jordan Schneider: To close, what are you researching right now?

Kamil Galeev: Currently, I’m primarily interested in coups and power dynamics — how political power is gained and lost. My main topics of interest are the October Revolution and its consequences, as well as the French Revolution and its aftermath. In the coming months, I’ll be writing about Lenin, Stalin, and the Bolshevik rise to power, as well as revolutionary Napoleonic France. Both these historical periods significantly shaped the world we live in today.

is the author of the Concurrent Substack and host of the CyberPink podcast. Concurrent explores the parallel and colliding tech and cultural currents shaping Silicon Valley, China, and beyond. Her show 疲惫娇娃 CyberPink is a Chinese-language podcast about popular culture. From screens to the cosmos, it explores culture in its broadest sense—using women’s voices to expand the ways we engage with and imagine the world. It is very good I highly recommend!

I recently hosted a 2-session reading club for Abundance, bringing together paid subscribers of my Chinese-language podcast CyberPink and those of our sibling podcast American Roulette. Ours was conducted entirely in Mandarin Chinese, which meant that the participants, by default, were like me: born and raised in China, educated in the West, and now navigating the cultural and ideological fault lines between the two. Among the attendees were academics, lawyers, consultants, AI investors in Silicon Valley, and engineers in big tech.

When you discuss American politics in a group like this, perspectives shift constantly. As we dissected Abundance, we toggled between the imagined readers in Brooklyn, the Bay Area, and D.C., and a more distant, more foreign vantage point — one grounded in the trajectories of China itself. In short, we examine the U.S. as foreigners, as immigrants, as Chinese.

Somewhere along the way, our Signal group of book club organizers — my podcast co-hosts and friends — jokingly named the group *Ezra Thought Study Group* (Ezra思想学习小组), a parody of Maoist-era “Marxism Study Group” or more recent “Xi Jinping Thought Study Group.” It was tongue-in-cheek, of course, but it also captured something real: we all sensed that this book was both a techno-optimistic manifesto from central-left, and a piece of strong political persuasion to the reconfiguring Democratic Party.

In this article, I’ve selected fragments from the 100-page (!) transcript of two book clubs, each lasting 1.5 hours — but not the ones you’ve probably heard before. I’ve deliberately cut the standard U.S.-centric book analysis and policy talks; this article is some curated moments that I believe offer some fresh takes. Consider this: Reading Abundance from China.

Note:

The book club followed the Chatham House Rule: participants may use information from the meeting but cannot reveal speakers' identities or information sources. However, I received permission to include some speakers' names and occupations. I've tried to maintain the conversational rawness and haven't done extensive editing, so the transcript may contain factual mistakes. If you find any, please leave a comment. The transcript has been translated and made readable by Claude.

The poverty of American imagination:

Lokin (Lawyer, lives in NYC): I think there's this incredibly limited imagination about what the "good life" looks like in American culture—both among ordinary people and political elites. And this limitation becomes a huge obstacle to building a more abundant, more public, more sustainable future.

There's this perfect example: a Democratic delegation visited Japan recently, and several congresspeople toured the Shinkansen—which, by the way, was built in the 1960s with American funding.1 What's almost laughable is that even today, this public transportation system—which isn't exactly cutting-edge anymore—still managed to "shock" American congresspeople. This tells you that even the highest-ranking political elites don't really have opportunities to experience or understand the kinds of lifestyles other countries have built.

More broadly, American society's vision of lifestyle is still heavily dependent on this outdated American Dream: everyone should own a detached house with a front and back yard, white picket fence, and one or even multiple cars of their own. This individualized, anti-public understanding of the "good life" makes it hard for people to imagine, let alone accept, lifestyles based on public transit, shared spaces, and urban density. So we can understand why American rail infrastructure always struggles to move forward—it's not just technical and budget issues, it's cultural and imaginative barriers.

I was talking to a professor at Columbia—one who’s knowledgeable, well-traveled, worldly person. He told me how impressed he was by Beijing's public safety: at 1 AM, he saw a woman in fur and jewelry walking alone on the street, eventually taking the subway home without any worry. He cited this as positive evidence of modern urban life.

I politely reminded him: if you want to use an example to illustrate good urban safety, you should really mention Tokyo or Seoul instead. Because these cities are equally safe, but they're less likely to be misunderstood as depending on "authoritarian order" for maintenance. When you use Beijing as an example, in the American context, it easily activates this "authoritarian scratch"— people who are culturally inclined to believe order can only be achieved through strongman rule will instinctively equate urban safety with authoritarian governance. They'll think only highly centralized power systems can achieve clean, safe, orderly urban life. This imagination further reinforces their pessimism about democratic countries' inability to govern cities well, providing psychological support for rationalizing some kind of authoritarian governance logic.

But this is actually a very dangerous misreading. We have to dismantle this binary thinking: cities can be both safe and free; public life can be both efficient and democratic. This kind of life exists not only in Tokyo, Seoul, Amsterdam, Zurich, but could absolutely be realized in America, provided we first culturally change our assumptions about what "ideal life" looks like.

What's even more concerning is that even these elite intellectuals still lack basic concepts of what "modern urban life" should look like. They may have never truly internalized the daily experience of "stepping out and taking clean, safe subway, walking freely in dense urban neighborhoods." They still view "cities" as dirty, dangerous, anxiety-inducing places, while treating "suburbs" as safe, clean, ideal residential areas.

This deeply rooted cultural cognitive structure is the biggest resistance we face when promoting public lifestyle transformation. Under this cultural logic, even with sufficient resources and mature technology, it's hard to push for truly progressive infrastructure transformation.

Afra: The poverty of American elite imagination about "happy life" is a key factor preventing visions like Abundance from being realized. Even if we set aside structural obstacles like racial discrimination and economic interests, just looking at popular culture, America’s deeply influential soft power, the society has already fallen into a kind of imaginative local optimization.

As the world's most powerful popular culture exporter, Hollywood has long been continuously and repeatedly producing a specific kind of life picture, and this picture's singularity is actually quietly limiting public understanding and imagination of the possibilities for abundance and "good life."

For example, when Hollywood wants to show a big city's bustling scene, the template is almost always New York: broken-down subways full of rats, Manhattan's towering skyscrapers, a city that's vibrant yet dirty and chaotic. This "city equals anxiety" narrative has almost become American culture's default setting.

And when film and TV want to imagine "ideal life," the camera often turns to suburbs: spacious detached houses, white fences around front yards, two cars, a few dogs, quiet tree-lined neighborhoods, and a typical nuclear family. This repeatedly reinforced template creates a ceiling for cultural imagination, making it hard for people to conceive of a lifestyle that's both urban and livable, both public and high-quality.

It's precisely this imaginative pathway that's been constantly reproduced by the cultural industry for years that makes "abundance" futures largely misunderstood as extensions of consumerism or "upgraded versions of suburban dreams," rather than optimization of public spaces, infrastructure innovation, or reorganization of human relationships.

On environmental assessment in China and the US

T, (Economist): I'm an economist, and my main research area is causal inference. Right now, most of our policy evaluations are built on causal inference methods. The most common example would be A/B testing or randomized controlled experiments. Many policy evaluation paradigms, including things like environmental assessment, basically follow this same logic and process—they're all built on this scientific framework about what makes policy "rational."

But the problem is, this approach might be missing a crucial point: when we do A/B testing, we often can't capture general equilibrium effects. We don't have the ability to properly measure the chain reactions that policies create at the system level. What we get is often just a local impact—a primary feedback observed under specific settings. But this local effect might not be the most important first-order effect; the deeper structural impacts might be exactly what gets excluded because our evaluation methods are too reductive.

I think environmental assessment is a perfect example of being shaped by this paradigm. Many countries do environmental assessments—China has similar systems too. But China's environmental assessment often feels more like formalistic copying: seeing that the U.S. and other countries have environmental assessment mechanisms, so we "should" have them too. But for a long time, China's environmental assessment wasn't treated as a serious, scientific problem, so it didn't adopt the kind of rigorous processes based on scientific methods like America does.

The result is that in China, environmental assessment plays a very limited actual role in infrastructure development. And in America, while the environmental assessment system itself is more influential, the methods it relies on still tend to only capture partial effects while ignoring broader general effects. For example, opportunity cost might be a crucial variable in policy choices, but it's not something that's easily captured in our current mainstream operational inference frameworks.

So I think there's this really interesting, even ironic paradox here: our methodology is indeed getting more and more advanced, allowing us to make increasingly "scientific" policy evaluations. But at the same time, it's precisely these methods themselves that are making our understanding of the complex, multi-layered impacts after policy implementation more narrow. Maybe in trying to make policy evaluation more "falsifiable" and more "rigorous," we're also losing our sense of its holistic and long-term dimensions.

The "tech OS gap" between China and the US

Du Lei (tech investor, lives in the Bay Area): From my background—I originally did AI research and now I'm in tech investing—so I instinctively tend to look at problems from a "system design" perspective. My very intuitive feeling right now is: America is using an outdated institutional framework, this old governance "software," to deal with a real-world social system that has higher bandwidth, more complexity, and faster change. The result is—the whole system is starting to fall behind.

The deregulation that Abundance mentions frequently, in the short term, might indeed be an emergency measure to improve efficiency. But if it's just deregulation without actually rebuilding this institutional infrastructure, it's probably just "treating the head when the head hurts, treating the foot when the foot hurts."

It's like a programmer who just joined a company that's been using a ten-year-old system, saying: "This legacy code is too messy, just delete it." Deletion might feel good, but three months later the whole system could just crash.

Let me talk about the differences between China and America. I think China's execution advantage in certain areas doesn't necessarily come from so-called systemic superiority, but is more like a manifestation of a technological generation gap.

At the end of the day, China's bureaucratic system is more updated than America's. So even if both countries are equally bureaucratic, Chinese government departments use WeChat to communicate directives and share spreadsheets—at least at the tool level, it's faster and more efficient than America's approach which still relies on email and paper processes.

Of course, China has its own systemic problems too. Like this "layer-by-layer responsibility implementation" approach we've seen since SARS, all these "red line" administrative mechanisms often cause inefficiency and even suppress genuine local feedback. But from an operating system perspective, China is indeed faster at "hardware updates," even if the "software logic" still has plenty of problems.

I also want to respond to this topic about "industrial planning vs. basic research." My personal view is: I don't really believe the government can actually push forward much substantive breakthrough in "basic scientific research."

In another book our CyberPink listener’s community is co-reading, Nexus, it talks about how many real scientific breakthroughs actually come from a broad peer network—they need free, open exchange and sufficient resources. So a lot of times, the technological explosions we see are actually the result of long-term, multi-path, slow accumulation.

For example, America's AI explosion didn't just appear one day with a "lightbulb moment," but is the result of twenty years of big data accumulation, gamers pushing GPU performance, miners driving computational power—all these things stacking up over a long time.

It's the same on China's side. A lot of the technological progress we see isn't some policy suddenly deciding something, but the entire manufacturing system, craftsmanship system accumulating through massive iteration processes.

This is a kind of compound advantage. I think we can't misread this compound progress as achievements brought by "government systems." It's more the result of an ecosystem.

The economic-political split: ground-level contradictions in the U.S.

Amber (Consultant, lives in New York): Hi everyone, I'm Amber. When Afra mentioned America "wanting to catch up" or "wanting to lead" in areas like batteries and solar energy, I really resonated with that—there's this complex emotion behind it, both envy and a deep sense of being torn apart.

Let me give you some background. My day-to-day work involves directly dealing with local governments across America, investment promotion departments, including some global investment projects landing in the U.S. Many of the projects I handle involve building factories, offices, job creation, and so on. And in this process, I've observed a very obvious contradiction:

On one hand, local governments desperately want these investment projects to land because they bring tax revenue, jobs, can revitalize their communities and drive regional development. But on the other hand, they're extremely cautious, especially when facing investments from China—political agenda almost always gets put first.

Here's a simple example: it's not just Texas or those Southern states everyone's familiar with—even states like Southern conservative states and Midwest states that seem like swing states are now very sensitive about "Chinese property ownership." Many local governments explicitly write in their RFP documents that they won't accept investments from China—they just completely won't entertain it. For many of the projects we're working on, this is like a blow to the head.

What's more ironic is that many local officials are actually very conflicted privately: they really do want these projects to land, they know these projects are good for the local economy, but they can't help it because the governor, state legislature, and voter pressure require them to "stay aligned with Trump" politically, they have to put on a tough stance. So the whole system is full of internal tension.

And it's the same on China's side. Many Chinese investors really want to enter the American market, they see opportunities, but they just can't get in.

So what we're seeing now is this bidirectional tear: America and China are mutually wary at the policy level, mutually hostile in public opinion, but economically, they both want to get a piece of the pie from each other. America is worried about leaks on one side and unwilling to give benefits, but on the other side wants to take China's money; China is the same—worried about technology blockades on one side, but hoping to get American subsidies and market access on the other.

So on the surface, both sides seem "calm," but privately, the ground-level exchanges are full of struggle and distrust, like a relationship being pulled in two different directions.

This is also the most real state I've experienced when dealing with American local governments, especially some small-town industrial parks: they really are desperate for development on one hand, but on the other hand they're completely constrained by the upper-level political environment, stuck between a rock and a hard place.

The US and China spend too much time doomscrolling on each other’s social feeds

Yiting (tech worker, lives in London): I think China and America right now are exactly like two doomscrollers brainrotting on each other's social media feeds.

It's like America suddenly scrolls through China's feed, sees some superficially shiny stuff, and goes: "Oh? I want that too!" But they don't really understand what this "person" China is actually like—what reality they're facing, what their situation is, what their resources and challenges are. They don't care, and they don't understand. They just see a filtered photo and start envying, imitating, even getting anxious.

Then America's own actions start getting distorted, wanting to "become another person," without figuring out why that other person is the way they are.

Actually, China does the same thing to America.

Many Chinese people, maybe even including some people at the policy level, don't really understand America's political ecology, social structure, or how their institutions actually work. But they scroll through America's social feed—like some very free, very advanced, very prosperous moments—and think: "I want to be like that too."

So the end result is: both sides are looking at each other's highlight reels while ignoring each other's complex realities, and they both fall into this illusion of "everyone else is living better than me."

At the end of the day, I think both China and America should probably spend less time brainrot.

The blind spot of China envy

Luke (tech worker, lives in NYC): I want to ask a direct question—have you guys listened to the latest episode of Bumingbai podcast?2 They mentioned some huge problems behind China's prosperity. I'm particularly curious: if it's people like Derek Thompson or those centrist liberals, in their process of "rediscovering China," after learning about these problems, what kind of reaction would they have? How do they view these realities?

H (media professional, lives in NYC): Personally, I feel like they don't really understand China—many of them haven't even been to China. When they mention China in articles or podcasts, it's not because they really care about China's history, policies, or the situation of its people. They're using China as a mirror or reference point—not exactly a cautionary tale, and not a positive example either, but a contrast that can inspire Americans' imagination, fighting spirit, and policy action.

What you just mentioned, like solar panels—I'm not an expert in this field, but from some reporting and observations, China's rise in this industry did go through a complex process: from early government subsidies and factory expansion to gradually establishing a globally leading position. But in this process, many Western observers ignored the real costs behind it, like compressed labor rights, serious resource waste, and even corruption and benefit transfers. These realities, as people with Chinese backgrounds, we should of course pay attention to, but in the discourse of people like Ezra Klein or Derek Thompson, these are hardly mentioned. What they care more about is how to use "China's success" to inspire competitive consciousness within America.

When they talk about China-U.S. relations, they easily apply the Cold War framework, like comparing it to the Soviet Union's Sputnik moment: the Soviets launched the first satellite, which inspired America's systematic investment in space, education, research, and other fields. This kind of "being inspired" is the process they hope to replicate from China again. But whether this path is correct is itself a controversial political judgment.

Afra: I agree. Ezra Klein and Derek Thompson are, after all, American political commentators working within that context. When Ezra Klein mentioned in his podcast that he went to China once about a decade ago, he mentioned that he was talking about having a lot of meetings with officials, with economists, with some business people. So naturally, his China experience draws from that particular slice of the country.

When H mentioned the lithium battery industry earlier, I also remembered chatting with a British scholar a few days ago. She said she attended an academic conference where a young female scholar shared her research on the environmental impact of a lithium battery factory in Sichuan or Anhui (I couldn’t recall). After this young scholar finished speaking, she told all the attendees: "I really hope this paper can be published in China, but I know it's almost impossible." This exposes a core problem: many environmental and industrial costs in China cannot be openly discussed.

A few months ago, I also read an article in Rest of World by Viola Zhou about Chinese lithium battery company Gotion wanting to invest in building a factory in a small Michigan town, bringing huge funding and job opportunities. But the project was ultimately blocked by small-town politics and strong opposition from local residents, where the main resistance was exactly concerns about environmental pollution. In contrast, in China we can hardly truly see these local costs of EV manufacturing or lithium battery factories—exactly how much wastewater was discharged, how much land was occupied, how many people's lives were affected. Much of this data is unknowable because of media lack of transparency and the absence of civil society, meaning these voices have no channels to be heard at all.

H (media professional, lives in NYC): Right, these American "China-envy" people only see the results, like Chinese electric car companies like BYD rising in the global market. But they often ignore the real problems behind these companies, like product quality issues, defaulting on supplier payments, and so on. These more detailed, more complex layers—they don't really care about them.

Debate about the secret sauce and manufacturing mobility

Afra: Ezra Klein mentioned, “the re-emergence of industrial policy in America is 100% about China. Take China out of the equation, and there is no re-emergence of American industrial policy.” Actually, going back to this American reindustrialization thing. Manufacturing: When you really interrogate what technology in manufacturing actually is, of course, you can say part of it is about patents, about things that can be written down, about hard knowledge in manufacturing.

But there's actually another huge part that's tacit knowledge.3 How do you manufacture an iPhone? You need many skilled workers going through many complete assembly lines to put this iPhone together. It's not like you can write the iPhone assembly steps in a piece of paper, and then have new American workers read that paper and immediately go into the factory and start working.

If America needs to reindustrialize, you might really still need to bring Chinese workers back to continue training American workers before you'd have a relatively effective production process. Is America's imagination about reindustrializing very arrogant? Massively ignoring this kind of tacit knowledge in manufacturing and China's accumulated experienced workforce.

X (ML engineer, lives in NY): I have a pretty different, pretty opposite view, because manufacturing itself has extremely high mobility. The reason it exists, the biggest reason it can scale, is that it can reproduce very quickly—the entire factory, the entire process, assembly lines can reproduce very quickly following a template. This is the essence of manufacturing.

At least we might see individual cases like Huawei in the news being a bit less common, or things like that Cao Dewang factory in America having trouble getting started being less common, but including—I don't know if that CyberPink listener is here today, he works in chip industry landing in America—this stuff being able to flow back is the norm.

Manufacturing being able to flow between countries, building a factory and being able to make stuff—this being possible in most cases is actually the norm. If this wasn't the norm, globalization wouldn't have happened. Shenzhen, I think, does have a very unique ecosystem, including its concentration of talent and concentration of knowledge and the existence of the entire ecosystem. I think Shenzhen is a unicorn, just like Silicon Valley itself is a unicorn. But I don't really agree with what you just said about treating manufacturing as something that needs secret sauce, because the essence of manufacturing itself is that it can land anywhere and you can follow the recipe to make it, because if this wasn't possible, this industry wouldn't exist.

So including why everyone in manufacturing, all people in the manufacturing industry, always have this strong sense of crisis about manufacturing mobility—it's still because its mobility is too strong. Including that article I posted before about India, about that piece Viola wrote about India, I found that article very familiar because all the problems they were discussing—Indians feeling like our Indian manufacturing will never get off the ground—but the problems they cited are exactly the same as what I think Chinese people said 20 years ago about how our Chinese high-end manufacturing will never get off the ground.

So I think many things aren't inevitable—it's the current state produced by globalization at this moment. And including my overall view—because I might have followed a few cases at work where a chip factory was moved over from start to finish—I think everyone's degree of exaggerating the secret sauce of many industries is still a bit much.

Afra: What I actually want to say isn't that China now has high-end manufacturing secret sauce that other countries don't have, but that I think this secret sauce cultivation is a very long process. And the reason China was able to cultivate this secret sauce is because China was in a situation at that time—early 2000s, late 90s—where everyone was sharing secret sauce without reservation in a very radical and idealistic state of globalization. But now in 2025, it's no longer that kind of state. Of course, I feel like secret sauce is a very inappropriate term—I'd rather use the term tacit knowledge.

Back in the early 2000s, if China's Foxconn built factories in China, Apple could send large teams of technical bureaucrats to help China build factories and educate the workers. But now if India builds factories saying they want to replace China in manufacturing iPhones, if they want to ask China for some high-end manufacturing help, China definitely say no, because globalization now is not the same as globalization back then. Everyone knows this tacit knowledge is very precious, everyone thinks this is to some degree a huge asset.

X (ML engineer, lives in NY): I think whether they give it or not depends on who's paying—that's also part of the... Back then I think Koreans weren't that willing to give us (China) technology either, but this thing... I think we're all old enough to remember China's manufacturing going from nothing to something, so I'm also very worried that it going from something to nothing is also something we could witness in our not-very-long lifetimes.

Afra: I'm not saying this transition won't happen, but what I quite agree with in Abundance is "America forgot how to build." I think that's a very precise statement, because in manufacturing, knowing how to build and deploy is a "practice makes perfect" technical skill. It's a state that needs more than a lab, more than a book. Just like if you don't exercise for a long time, your muscles gradually atrophy—America's manufacturing muscle has atrophied for decades.

The Deploy chapter in Abundance talks about the lithium battery case, which is really exactly about Americans inventing lithium battery technology and then forgetting how to deploy and scale. Now Chinese lithium battery manufacturers can compress prices in every single link to the lowest, optimize every aspect and every raw material of lithium batteries to the optimal state—this deployment optimization has gone through twenty years. If suddenly tomorrow a Silicon Valley company appeared saying: we're going to build a lithium battery factory from scratch without relying on Shenzhen to teach us, I don't believe it would immediately succeed.

Bay Area’s military-industrial startup renaissance, and China’s "crossing the river by feeling the American stones" innovation style

Du Lei (tech investor, lives in the Bay Area): When we think about San Francisco startups, we traditionally picture a software-dominated entrepreneurial ecosystem. But if you head down to the South Bay, there's still this atmosphere tied to traditional industry—especially over the past decade, we've seen the emergence of a whole batch of defense-related startups in America.

Companies like Anduril represent this trend. People are starting to realize that in this new phase of industrialization, defense startups can actually attract venture capital funding, and their returns and growth potential look pretty damn good. So we're seeing that in America, entrepreneurship isn't just limited to crypto or AI—even defense is becoming a new direction that VCs are paying attention to.

I just shared an image in our Zoom chat, it's the business card of a CEO from a startup that makes anti-drone defense systems. You can see a perfect example here: Reddit-style geek culture meeting the defense industry. This seemingly awkward fusion actually shows that America isn't completely hopeless when it comes to rebuilding its manufacturing and industrial base.

At the same time, fields like blockchain and robotics are gradually recovering too.

This highlights a fundamental difference between the innovation ecosystems on both sides. In America, the entire industrial structure and institutional environment is more encouraging of disruptive innovation. Entrepreneurs are more inclined to do things that are completely different, trying to solve old problems with entirely new approaches. Once this kind of innovation gets market validation, it can quickly attract massive capital, and America's institutional and capital mechanisms can give successful players decent returns.

In contrast, truly disruptive innovation is relatively more difficult in China. The path dependency is stronger—a lot of times it's still "crossing the river by feeling America"4, so incremental innovation is more common. This incremental approach actually has strong advantages, especially in large-scale industrial production industries. As mentioned in Abundance, much of real technological progress often doesn't come from some genius's "eureka moment," but from countless small iterations on the production floor, on the assembly line.

America definitely has advantages in excavating "innovation points" that have clear commercial value and market acceptance, but when it comes to those "1.1 improvements"—the continuous polishing and optimization—the gap compared to China is obvious. This actually reflects deep institutional design differences between the two countries: America's system encourages high-risk, high-reward winner-takes-all, while China's structure tends more toward stable returns and inclusive distribution. These institutional and cultural differences in risk appetite ultimately show up in the industrial realities and innovation models we see today.

X (ML engineer, lives in NY): I've been thinking about the past few years, and maybe as someone in the healthcare industry—at least from my own perspective—I actually feel like the reasons why these two countries can't achieve disruptive innovation in certain areas are quite similar: whoever's figuring out how to make money off the government, whoever's trying to optimize the bureaucracy, they can't do disruptive innovation.

The ones who can leapfrog across barriers, especially recently I've been reading Careless People. I've found that whoever can cleverly circumvent regulation, taking action before regulators even react, they're more likely to succeed. This viewpoint might sound a bit cliché, but that's really the situation now.

I think many industries in both China and the US are the same—they're all trying to figure out how to capture more government funding. These industries, as we just touched on, have fallen into a state of false innovation. Just like we discussed in our last book club, they're addicted to surface-level innovation, with all costs going into the bureaucracy. The people who can actually focus on innovation, focus on coding, are really few and far between, and these programmers spend every day in meetings.

The hidden costs of re-industrialization: the Factory Girls nobody talks about;

Afra: Before the book club ends, let me add one more thing about American re-industrialization. Remember during that period when Trump and J.D. Vance were heavily promoting "American re-industrialization"—there were tons of memes on X. Professor Huang Yasheng from MIT posted a trolling tweet.

Huang Yasheng was basically suggesting that this book tells you some reality about manufacturing and re-industrialization. that you wouldn’t learn in the schools in the U.S. I found that tweet fascinating, so I tracked down the book, which I had heard about for many years, and read it. Many scenes in the book are shocking. Not just because they made me remember that institutionally chaotic China from over 20 years ago, but because they made me rethink the lived reality of ordinary workers under the factory system. The book spends extensive pages describing the real conditions of young female workers on assembly lines.

For instance, in the factories, many female workers (almost all come from rural areas) wouldn't tell their coworkers their real names. Because once you reveal your name, if you happen to run into someone from your hometown, information might get back home, and relatives—even parents—might find out about their income, leading to economic exploitation, with wages being demanded and controlled. These female workers were doing shift work day after day on assembly lines, twelve hours a day, while also facing "exploitation" from their families of origin. The book also describes how they struggled to establish their own identity in extremely compressed living spaces—in factory dormitories with no privacy, living in rooms with six or eight other female workers, where even just trying to make yourself "different from others" was difficult. There was no mobility, no prosperity. It was this state of people struggling to survive within the system, under the assembly line, within silent norms.

After reading this book, I went online to look up what Foxconn factories look like today. I was surprised to find that photos, videos, and even content shared by workers on platforms like Xiaohongshu that circulate on the Chinese internet really aren't that different from the lives of that generation of Shenzhen female workers that Leslie Chang wrote about in the early 2000s.

During my doomscrolling, and some of them really saddened me. If you just search keywords like "Foxconn," "assembly line," "factory girls", "factory boys" (厂妹，厂弟）on Xiaohongshu, you can see tons of real documentation: what workers eat, where they sleep, their shift conditions, what they do after work. There are also many factory veterans"giving advice to vocational school students and high school graduates preparing to enter factories.

Behind this content is actually a complete factory culture and factory logic; a grim way of organizing people, controlling time, and stripping away identity.

This made me think—if US wantS to replicate a massive, complex, comprehensive manufacturing ecosystem like Shenzhen, then the social costs and structural prices behind it are far beyond what American politicians are actually thinking about when they're currently discussing "re-industrialization."

Another related example: director Yu Xinyan made a documentary called Made in Ethiopia, about Chinese companies setting up textile factories in Ethiopia. They established what's called the "Eastern Industrial Park"—just the name alone carries this heavy self-orientalist flavor, plus some Belt and Road style developmentalism logic.

The first part of the documentary tells the success story—thousands of Ethiopian female workers employed, smooth cooperation with the local government. But the focus of the documentary is on how to expand the factory to the second phase, they had to requisition villagers' farmland and promote village-scale migration. It was about trying to "system-switch" an agricultural society into one that accepts factory labor logic.

The film is full of conflict and tension: farmers' resistance, government buck-passing, cultural misalignment, and how Chinese-style production logic, such as progress determinism, gets forcibly transplanted into a completely different social soil. These scenes are extremely similar to the pain points in China's early manufacturing development.

Speaking of transnational factory landing, there's another example really worth paying attention to (which was previously mentioned): NPR's Planet Money recently did a follow-up episode about the Gotion story with more details, including how several consecutive town hall meetings in this small town gradually evolved into strong opposition to "Chinese capital entry." The whole thing essentially became a collision between American grassroots democratic politics and global industrial expansion.

Through these reports, I really saw the complexity of "industrial projects" in the American local context: how small-town political culture operates, how the public expresses opposition, how policymakers struggle to compromise.

Subscribe to Afra’s great substack and check out her podcast!

Concurrent

Understand the parallel and colliding tech and cultural currents shaping Silicon Valley, China, and beyond.

By afra

Today’s post is brought to you by 80,000 Hours, a nonprofit that helps people find fulfilling careers that do good. 80,000 Hours — named for the average length of a career — has been doing in-depth research on AI issues for over a decade, producing reports on how the US and China can manage existential risk, scenarios for potential AI catastrophe, and examining the concrete steps you can take to help ensure AI development goes well.

Their research suggests that working to reduce risks from advanced AI could be one of the most impactful ways to make a positive difference in the world.

They provide free resources to help you contribute, including:

• Detailed career reviews for paths like AI safety technical research, AI governance, information security, and AI hardware,

• A job board with hundreds of high-impact opportunities,

• A podcast featuring deep conversations with experts like Carl Shulman, Ajeya Cotra, and Tom Davidson,

• Free, one-on-one career advising to help you find your best fit.

To learn more and access their research-backed career guides, visit 80000hours.org/ChinaTalk.

is a Professor of Urban Studies at Singapore Management University, and research fellow at MIT. His research examines the political economy of digital infrastructure in China and Southeast Asia, and maintains the excellent substack Sinocities.

U.S.-China competition in AI is seen as the defining technological rivalry of our time. Much of the analysis in this competition is premised on a zero-sum logic — we cannot let China get ahead of us in AI, so the logic goes, as this will inevitably mean the forfeiture of American technological and military primacy.

There is another possibility–China and the U.S. may develop different “varieties” of AI. For example, the U.S. advantages in cloud computing, software development, and openness to talent (tbd…) give it an edge in development of enterprise software and large language models (LLMs). However, China has clear advantages in manufacturing and infrastructure, which could offer an edge in what experts term “embodied AI”, or in Chinese jù shēn réngōng zhìnéng, 具身人工智能. Embodied AI systems interact with the physical environment through sensors (like cameras, microphones, touch sensors) and actuators (motors, limbs, wheels, etc.). Embodied Intelligence is shaped by a real-time, physical engagement with the world.

The central government recently included “embodied intelligence” in its work report, indicating the area as a key priority. Zhongguancun, Beijing’s hi-tech area, recently released its plan for embodied intelligence. A recent report by Georgetown’s Center for Security and Emerging Technology (CSET) focused on efforts in the city of Wuhan to embody “AI algorithms in real environments. Imbued with the Chinese Communist Party’s predefined values, the AI interacts with its natural surroundings, learning as it proceeds.” In this post, I explore some of the origins and implications of China’s interest in “embodied AI”, and then present a few examples of this approach in sectors such as autonomous vehicles, personal mobility, robotics, and the“brain-style” AI models being used for smart city operations. While the current push for embodied AI is part of a prevailing fever for AI globally, China’s priorities reflect long-standing beliefs in the use of technology for solving governance problems, and the need for physical infrastructure and manufacturing as key priorities for the country’s development.

Internet+, the Fourth IR: the Party’s Embrace of Embedded AI

In 2015, China’s Made in 2025 Plan drew the ire of many in the West for its ambitious goals of replacing foreign companies in China and dominating frontier industries like aerospace, biotech, and smart manufacturing. In the same year, Vice President Li Keqiang unveiled a vision for Internet Plus (互联网+), which called for applying data and AI to manufacturing, smart homes, intelligent vehicles, robotics, wearable devices, and more. The Internet+ Vision is seen as one of China’s earliest national strategies on the application of AI to industry, and underscored the degree to which policymakers in China saw digital data not merely as a sector in itself, but as an input for many other sectors. Li Keqiang’s Internet Plus was also inspired by Industry 4.0, a concept initially proposed in Germany in 2011 (Huda 2023), and which influenced World Economic Forum founder Schwab’s idea of the “Fourth Industrial Revolution” (Schwab 2017). Schwab popularized the idea of the Fourth Industrial Revolution, but the term itself emerged around 2011 at the Hanover Fair as part of Germany’s strategy to use digital technologies to maintain and deepen its edge in manufacturing. The 4^th IR predicted the application of computing technology and the internet to a wider range of everyday life and objects. This encompassed “artificial intelligence, robotics, the Internet of Things, autonomous vehicles, 3D printing, nanotechnology, biotechnology, materials science, energy storage, and quantum computing.” The concept found resonance in China, a country that, like Germany, was arriving on the world stage as a major manufacturing power.

The idea of the fourth industrial revolution was embraced by top party leaders, including Xi, who saw the coming technological revolution as part of a critical period of changes in world history that China had the opportunity to lead. In 2017, Xi used the term “profound changes unseen in a century”, or bǎinián wèi yǒu zhī dà biànjú 百年未有之大变局, encompassing technological transformations but also a shift away from a U.S.-led global order. As Jin Canrong 金灿荣, Dean of the School of International Relations at Renmin University, put it in 2019, “After the fourth industrial revolution, the productivity of the East is likely to be ahead of the West, or at least a balance between East and the West will be achieved. This is the most important change among the three changes unseen in a century.” Jin goes on to note that, “If the 4th industrial revolution is, as Schwab described, ‘5G + the internet of things,’ then China is already leading this revolution, but I tend to be in the camp seeing this as more of a deepening of the 3rd (internet) revolution.” Nevertheless, Jin still saw China as having an advantage over the U.S. due to its production capacity. “The U.S. still has the best innovation capabilities, but the hollowing out of industry is a big problem. If you cannot turn innovation into products, it’s the same as a piece of waste paper.” Thus, Jin viewed manufacturing as a crucial component of China’s strength that would allow it to outcompete the U.S. Recently, Tsinghua Professor Tang Jie 唐 杰 repeated a view on China’s advantages as comprising “super-large population, rich application scenarios, and rapid iteration of the end-side ecosystem provide fertile soil for the rapid development of AI big models.”

Of course, the “4th industrial revolution” hasn’t played out entirely as Schwab or China’s leadership assumed. Around 2015, 5G and the “internet of things” were predicted to revolutionize everything. So far, 5G and the Internet of Things have proven less transformational than originally anticipated, but we are still early innings of what could be a longer story in the application of AI to other areas.

Governance: City Brains and other Brain-inspired AI

In addition to being home to Alibaba, the city of Hangzhou helped spawn the so-called “six little dragons”, 六小龙 which include DeepSeek. Hangzhou is also the site of one of China’s first successful “City Brains,” 城市大脑 — a smart city platform developed by a partnership of the city government with Alibaba’s Ali Cloud back in 2016. The platform was first deployed to help reduce traffic jams. A network of AI-enabled cameras, produced mostly by local firms Hikvision and Dahua, feeds information into a cloud system that gathers real-time information on traffic flows and accidents, and processes this information to adjust and optimize traffic lights across the city. The City Brain 3.0 integrates data from traffic lights, surveillance cameras, vehicle GPS, municipal records, public services, and IoT sensors. Alibaba has gone on to market its City Brain platform to cities around China and the world, and China has promoted the construction of City Brains around the country as a core component of the 2023 Digital China Plan. China’s 14th Five-Year Plan (2021-2025) calls for using “digitalization to boost urban and rural development and governance model innovation, build an urban data resource system, and promote the construction of urban data brains" (14th Five-Year Plan). Subsequent policies have followed the 14th Plan, including the 14th Five-Year Plan for National Informatization, which includes plans for a “ubiquitous, intelligent, connected digital infrastructure system”, comprising 5G, big data centers, and smart networks, for example.

The “city brain” concept dovetails with the established interest in using AI to improve governance. The PKU-Wuhan AI institute has apparently been developing a “large-scale social simulator” (大型社会模拟器). While the focus of these initiatives is on AI, they follow from a longstanding faith in systems thinking and technology to solve government administration problems. In the 1980s, the concept of cybernetics was embraced across a wide swath of scientists and leaders as the country turned away from the fervent ideology of the Cultural Revolution towards science and technology. The crux of cybernetics as developed by Norbert Wiener in the 1950s is information processing: “Cybernetics takes the view that the structure of any system engaged in communication or control must be analyzed in terms of the transfer and processing of information, whether the system be an electronic brain, a living organism, or a social institution.” Qian Xuesen 钱学森, a Chinese student of Wiener’s, helped build the PRC’s rocket force and other technologies in the 1950s after being forced to leave his professor position at Caltech in the wake of the Red Scare. In a 1978 article, Qian described cybernetics as a “technology of organizational management” (zǔzhī guǎnlǐ de jìshù 组织管理的技术). One of his protegees in cybernetic thinking, Song Jian 宋健, would go on to become a leading advocate for the one-child policy.

There are a number of research centers and projects working on various dimensions of “brain-inspired AI.” (类脑 Lèi nǎo AI) which aim to develop artificial systems based on models of human cognition and learning. Launched in 2016, the China Brain project focuses on the neural basis of cognitive function, brain-inspired AI, and the development of intelligent technologies that mimic the brain’s architecture and processes. The project is supported by the Chinese Academy of Science (CAS) Centre for Excellence in Brain Science and Intelligence, a consortium of laboratories at over twenty CAS institutes and universities, focusing on understanding neural mechanisms of cognition, developing brain-inspired intelligence technologies, and advancing brain disorder diagnostics and treatments. Beijing Institute for General Artificial Intelligence” (北京通用人工智能 研究院, BIGAI) was set up in 2020 in a joint effort of the city government, MIIT, and Peking and Tsinghua Universities. Peking University also helped launch the PKU-Wuhan Institute for Artificial Intelligence in 2022.

Training the Terrain: Autonomous Vehicles

After an hour-long subway journey southeast from Beijing’s city center, I arrive at a new gleaming glass-walled mall in the Yizhuang district. On first glance, the area is unremarkable and virtually identical to the myriad other “new areas” and new districts that have sprouted up on the edge of Chinese cities over the past few decades. But Yizhuang is not merely a typical new area, it’s also home to one of China’s most advanced autonomous vehicle testbeds, Beijing High-level Autonomous Driving Demonstration Zone (BJHAD). In 2020, China’s NDRC released its ’ “Intelligent Vehicle Innovation and Development Strategy, which called for “building a big data cloud control infrastructure platform for intelligent networked vehicles.” This oriented China’s approach to AV development around a principle of “vehicle-road-cloud integration” or 车路云一体化 chē lù yún yītǐ huà. In this model, pilot projects have been developed to deploy intelligent sensing and camera equipment on intersections within testing zones. The approach has been championed by an influential Tsinghua engineer, Li Keqiang 李克强 (same name as the former Premier), and endorsed by several of the country’s ministries and vehicle engineering associations. The idea has been around for a while — Andrew Ng, the Stanford-trained AI expert and former head of Baidu’s AI division, had championed what he called “training the terrain” back in 2016, just before he left Baidu in 2017 to found his own autonomous vehicle startup.

The crux of this approach is that, in addition to vehicles having onboard cameras and other equipment to navigate, intersections are outfitted with smart devices to communicate with vehicles in the area. As an autonomous car outfitted with sensing equipment approaches within 200 meters of such an intersection, the millimeter wave radio (MWR) detects incoming vehicles. Now, say there is a long truck in the left-hand lane. As the car approaches, the MWR and the low-latency cameras on the traffic pole detect the incoming car. Feeding the information rapidly into the mobile edge computing unit, the computer system processes the information. The latency, a measure of the time it takes for data to travel from one point to another, is within 200 milliseconds — around 70 milliseconds is how fast a human can react, so it’s not much slower than a human. Another car approaching the intersection from the perpendicular direction is also detected by the MWR and cameras on the other traffic pole, which also feed into the system. Using the information of incoming traffic, the mobile edge computing unit then communicates with the incoming car, which automatically begins to apply braking, slowing the car down before it reaches the intersection and averting a collision in the intersection.

There are problems with this approach. Intelligent intersections are costly to build and expensive to maintain. The safety benefits are relatively marginal for these “edge cases” that don’t occur often. Autonomous vehicle companies want their systems to work anywhere, not just where intelligent intersections are constructed. So why has China continued to pursue this infrastructure-centric approach? In the view of a former planner who worked for Baidu in Yizhuang, it's more of a political arrangement. “Think of it this way,” he said. “It’s a win-win for the government and Baidu. The local district government wants to attract automakers to invest here. They need to figure out how to provide a subsidy to Baidu. But they can’t buy the cars from them. So they purchase the software and computing infrastructure. Baidu has built an autonomous car factory nearby, so that counts for the local government’s economic development numbers.” The data collected in these “pilot zones” can also be used by the local government, whereas in infrastructure-independent scenarios, most of the data on autonomous vehicle operations remains with the company, such as Tesla or Waymo in the U.S. Foreign companies operating in China, such as Tesla, have been required by regulators to store all data from their operations in China, and refrain exporting sensitive data out of the country to comply with China’s strict cybersecurity laws.

A former Vice President of Baidu’s AV division, Wu Shulin 吴书林, recently founded a startup called Freego (正奇智能技术公司), focused on autonomous personal micro mobility. When bikeshare startups arose in China, they scaled quickly but also led to problems such as piles of misplaced or broken bikes cluttering city sidewalks. These systems also have problems in matching supply and demand, for example, how to get bikes to areas when they are needed most. If personal mobility devices were equipped with autonomous driving features, they could automatically return themselves to areas of higher demand after a user finishes a trip. The company is starting with automated wheelchairs, which are regulated as health devices in China and therefore require less permitting than vehicles. But they also have a larger ambition of expanding into a range of personal micro mobility devices such as scooters, e-bikes, or micro vehicles.

In their office in a startup park on the southwestern outskirts of Beijing, the founder animatedly showed me a diagram of an iterative loop in which devices are trained in real environments, continuously collecting new data on a variety of terrains, environments, and situations they encounter. Each type of device would collect different sorts of data, but collectively, this helps train a general model for navigating a variety of environments, such as climbing up staircases, navigating on sidewalks, or through crowded interior spaces.When I asked what the greatest obstacle to implementing their vision is, the founder replied, “It’s not a lack of government support or policy, but it's really the lack of high-quality AI software engineers.” (Interview, Beijing, 2024).

Robotics

Robots are seemingly everywhere in China today — running marathons, boxing each other. Unitree (宇树科技), China’s leading humanoid robotics company, released robotic dogs that became a social media sensation. Robotics is one of the most promising areas for deployment of “embodied AI” and one in which China already has a commanding position. China operates the world’s largest stock of industrial robots, accounting for over 50% of the global total. Robotic cooking machines have been seen in China for some time, tossing bowls of noodles. So what’s different now? According to Grace Shao, “embodied AI robots are trained on real-world data that uses reinforcement learning and they have developed the ability to “think. Recently, the Huisi Kaiwu platform was unveiled in Beijing as the world’s “first general embodied intelligence platform” for interfacing with robotics and other devices. The idea is that robots could be programmed and customized through this platform to handle specific tasks. The project is overseen at the Beijing Innovation Center of Humanoid Robots, jointly funded by the Ministry of Industry and Information Technology (MIIT), the Beijing municipal government, and other private firms and research institutions. In Shenzhen, Huawei (Shenzhen) Global Embodied Intelligence Industry Innovation Center began operation in 2024 and has cooperation with Shenzhen-based robotics firms Leju Robotics, Zhaowei Electromechanical, and Daju Robotics. Leju’s humanoid robot Kuafu (夸父) apparently offers 5G compatibility, which would facilitate data collection and training of robots for industrial applications.

Implications of Embodied AI with Chinese Characteristics

1. Diverging Models of AI Development, and bifurcation in global AI trajectories: The U.S. centers on abstract, cloud-based intelligence (e.g., LLMs), while China develops AI tightly integrated with physical systems and infrastructure. This divergence has implications for technological standards, ethics, and the global diffusion of AI technologies.

2. Infrastructure as an Advantage/Differentiator: China's infrastructural focus enables deployment of AI in real-world contexts at scale, especially in mobility, urban governance/management, and robotics. This presents a comparative advantage not often captured in typical U.S.-China tech rivalry narratives.

3. Legacy of Cybernetics and Systems Thinking: The enduring influence of cybernetic thought — from Qian Xuesen to current smart city projects—reveals a unique continuity in how China views technology as a tool for social coordination and governance.

4. Human Capital Bottlenecks: Despite policy support, China faces challenges in high-end software and algorithmic talent. The article notes concerns about the availability of top-tier AI engineers, which may limit scalability of embodied AI initiatives.

What can the U.S. learn from China’s approach to AI? So far, U.S. application of AI has focused largely on LLMs and application of AI in enterprise software. This means AI could be monopolized by tech platforms like Google, Facebook, Microsoft, and OpenAI. AI is being embraced across a wide array of sectors. Diffusion of embodied AI could even make efforts to reshore manufacturing more feasible, with the caveat that it would employ fewer workers and offer fewer well-paying jobs. Conversely, a failure to incorporate embodied intelligence could mean American manufacturing becomes even less competitive with China’s increasingly advanced factories.

Subscribe to Sinocities below!

Sinocities

Cities, China, Geopolitics, Technology, and everything in between. By: Andrew Stokols ; Phd, MIT Department of Urban Studies & Planning, Masters in Urban Planning, Harvard.

By Andrew Stokols

Patrick McGee is the author of Apple in China: The Capture of the World’s Greatest Company. Our discussion led us through a detailed history of Apple’s relationship with China, where iPhone manufacturing became a project of nation-building.

Cohosting today is Kyle Chan of the High Capacity Substack.

Today, our conversation covers:

• Why Apple moved production to China in the 1990s, and why it struggles to leave,

• How Apple’s obsession with perfection catalyzed China’s industrial upgrading and why it bought every CNC machine in the world

• The political side of production in China, including how Apple’s relationship with the Chinese authorities has evolved over time,

• The rise of Foxconn and other partners in Apple’s network,

• A peer into Apple’s management style, including the “Divorce Avoidance Program.”

Listen now on iTunes, Spotify, or your favorite podcast app.

Job Opportunity to Research AI + China

The Centre for the Governance of AI is looking for folks to research China and AI! Check out their three-month Winter Fellowship, one-year Research Scholar position, and Research Fellow position. I like the people there and they’re looking specifically for ChinaTalk listeners—take a look at the positions!

“Real Men Own Fabs”

Jordan Schneider: I want to start with the near-bankruptcy moment and the challenges of manufacturing in America in the 1990s.

As you write, Patrick, there was very much a “real men own fabs” energy to hardware manufacturing, where most companies, Apple included, decided that if they wanted to compete, they had to build all the stuff themselves. But as you write in your book, that line of thinking reached its limits in the mid to late 1990s. Why don’t you start our narrative there and tell us that story?

Patrick McGee: I talked to the earliest employees at Apple, and it was part of the ethos that you built your own computers. This is how Steve Jobs and Steve Wozniak met at the Homebrew Computer Club. They’d be taking apart computers, presenting to other people, and Steve Wozniak’s genius was to take apart a computer, improve some of the circuitry himself, and then put it back together in a faster way.

Apple wasn’t doing anything different by assembling computers themselves. That’s what you had to do. I don’t think the electronics manufacturing supply chain existed, certainly not in a globalized sense of taking things from Taiwan or elsewhere. That’s all later.

The computer I give credit to for changing everything was the IBM PC in 1981. Steve Jobs largely misunderstood it because when he disassembled it, he looked inside and said, “There’s nothing special here.” He’s not wrong about that. But what he missed was that IBM had launched a whole new battleground that Apple wasn’t prepared to fight against.

The IBM PC wasn’t about the graphical user interface. It wasn’t about user design or UX. It was about making these things and building them at a massive scale. To do that, they relied on third-party assembly giants. One that I highlight is called SCI Systems, which you could think of as the Foxconn of its day. It was founded in Alabama and played a role in satellites and rockets — it played a role in the Saturn V rocket that went to the moon.

Once they had that military precision, IBM scouted them out and they ended up doing the circuit board for the first PC and then assembling computers writ large. Then they played this massive role, doing that for HP and Dell and all these other companies, along with the rise of a whole new industry of Solectron, Celestica, and Jabil— all these North American companies who outsourced on behalf of the PC giants.

They gave birth to electronics globalization because once you have standardized parts in this booming industry, then the likes of Taiwan can outmaneuver everybody else and start building them in great quality and at great scale. Then they vertically integrated. They went back upstream and did more of the computer until by the mid-1990s, you had Taiwanese companies that were putting together everything and badging it with different names, and then they were even badging it with their own names.

There’s this fascinating 20-year history where Steve Jobs in particular has this DNA that he leaves to the company, which is something like, “We will build it ourselves because we’re going to design it ourselves. That’s how we distinguish ourselves, even as the PC companies go to this outsourcing boom.” When he goes to NeXT Computer — the other company he founded after he’s ousted from Apple in 1985 — he says of the computer, “My favorite thing about it is that it wasn’t built in Osaka.” Obviously Japan was playing the role that China plays now.

There’s something fascinating about that. Even when the narrative after 1997, when he came back, was that Apple had to outsource, it took Tim Cook years to convince Steve Jobs to give up manufacturing. I write about how in 2000, he handed out t-shirts to everybody that said “Mac-tories” on them. He wanted these Mac factories to play a role.

Without the dot-com crisis, you would have had Apple trying to make a push in building things themselves, which would have included building in America. But Apple stock, as people sometimes forget, lost 50% in a single day in September 2000.

They’re almost purchased — well, maybe I shouldn’t say they’re almost purchased, but Gateway thought they were such a contender that they could purchase them. That’s how bad things were at Apple — that Gateway of all companies was going to acquire them.

Anyway, my point is I have this manufacturing-focused history of Apple, that’s a new angle to report on the company, and yet I think it shouldn’t be a new angle. This is a company that manufactures — now they orchestrate the manufacturing — but their DNA was doing it all themselves. This shouldn’t be an obscure angle to have brought a new lens to, but somehow it is. I’m uncovering more than I should have been able to because I think we should have been reporting on this the last 25 years.

Jordan Schneider: It’s fascinating, Patrick, because when people think of Apple’s moat and what distinguishes them from the competition, a lot of folks first go to iOS. We have this America arc, and then we have this “go abroad to lots of different countries and try to find one or more that works” arc. What were some of the starts and stops in that arc in the 1990s and early 2000s before they ended up going to China?

Patrick McGee: This will sound less strange if the listener understands that Apple had a tri-continental strategy for building locally but at the continental level. Apple built its own computers in California and Colorado — that’s obviously for North America. For Europe, they built them out of Ireland and for Asia, they built them out of Singapore.

When they get involved with contractors in the late 1990s, they have them replicate this strategy. LG makes the translucent computer — the Bondi blue color that then becomes Life Savers colors — and LG does it in Korea. But when they expand to meet demand, they do so in Wales and in Mexico.

When Foxconn comes on as a second supplier, this is the meeting of the minds of Foxconn founder Terry Gou and Tim Cook, who’s a pretty new person at the time but is running operations. Foxconn does it in China, which is significant because Foxconn’s Taiwanese and some of the other Taiwanese companies are not in China at that point. I’m thinking of Inventec, for instance, which made the iPod in 2001, and that’s out of Taiwan.

But then Foxconn expands, and they expand to the Czech Republic for Europe and California for America. Once you are comparing costs and efficiency between these three regions, China is winning out time and time again for multiple years, such that by 2003, Apple has effectively given up everywhere else and consolidated into China.

I suppose technically the Czech operations are in effect until 2010 or so, but everything is moving to China because China is tailor-making policies for someone like Apple. Industrial clusters are forming. They’re putting up factories, not in great quality, but at an amazing scale.

I have this funny anecdote with this engineer from Apple in the early 2000s, and he’s literally counting the stairs between floors because it’s so obvious to him that it’s not the same distance between each floor. He’s even measuring the stairs and he’s finding that this is so slap-dash — everything’s put together so quickly. This does, if we’re jumping ahead, have implications for us, because of course, we would never build something that quickly because we would emphasize the quality, the safety, and everything else that we should emphasize.

We would still be doing the environmental paperwork by the time an entire factory in Shenzhen would be built.

John Rubinstein’s anecdote for the iPod is that it’s steel girders when he goes there and sees the iPod factory actually up and running because you get to the second floor and it’s all finished, but the rest of the building is not, and yet they’re still churning out iPods.

There’s this amazing sense that when Apple went to China in the early 2000s, it was very early days. China was not known for quality at the time. They are known as a manufacturer, of course, but they’re not building quality electronics. That was something that Apple played an instrumental role in.

Kyle Chan: One reason I found this book so fascinating — it’s the story of global electronics manufacturing told through one company across a number of different countries. You chart the movement of Apple’s efforts to figure out the best place to manufacture — the one place, the company, the country — that might suit their quality needs. They’re extremely demanding with rapidly changing specifications.

The way I think of it, they’re dating a bunch of countries over time, and you know that they’re going to marry China in the end. They’re going to marry Foxconn. It’s like a rom-com — “When Harry Met Sally” — and you’re going to follow the dating along the way.

Could you tell the story of searching for the right partner? In the background I have in my head the whole “flying geese” model of development in East Asia, conceived by a Japanese theorist in the 1930s about how Japan would take the lead technologically. Then, over time, different sectors would move on to South Korea, Taiwan, and then eventually to China.

The Apple story, the Apple manufacturing story, matches that so well in terms of the search. By the time you go through the process, you’re like, “Okay, come on. I know you’re going to end up in China.” But what were some of the steps along the way? Why did they end up settling on China as the place to be?

Patrick McGee: A saying I love about journalism is that you should always show, never tell. When I was submitting the book, my biggest black swan worry was that my editor would say, “I don’t understand why this book begins in 1996. China’s not in the book for the first 90 or 95 pages. Let’s get there.” But the whole point was, I needed to tell you without actually ever explicitly saying so that Apple had this totally obsessive, maniacal, perfection-oriented, no-tolerance-for-defects attitude.

You need to understand that when they were building something like the iMac G4 — the thing that people might remember because it’s anthropomorphic and looks a little bit like a Pixar lamp — I joke that it looks like it’s ready to break dance. That’s actually the first commercial that Apple used for it, grooving to the beat, nodding its head.

Nobody knew how to build that thing. Apple upended tools and machine makers all across Taiwan to get it done. Without me explicitly saying so, the reader begins to understand that Taiwan is too small to get this done. They’re literally having to get people from the Philippines to fill the factory lines, and they’re using a VCR machine maker in Malaysia. They’re relying on Singapore Airlines for a blade fabrication facility.

You realize that for them to have a meeting about various supply chain conundrums that they’re having, you’re having to get on flights and hand-carry things from one country to the next, which by the way is illegal. They’re doing so with five or six countries as far apart as Singapore and Japan.

When China begins to offer all of this in one industrial cluster around Shenzhen, it is mind-bogglingly good for anyone involved. One thing that I say in interviews — I don’t think I explicitly say so in the book — is that Tim Cook is often credited as the architect of the China strategy. He’s not the architect, and there is no architect. It was the suppliers themselves — and there are hundreds, if not thousands of them — are all choosing China because if they’re comparing Singapore versus China, the labor rates, the abundance of labor, the flexibility of that labor, the hardworking determination of that labor is all off the charts in China versus any other place.

The suppliers themselves are all moving there and Apple is able to take advantage of that movement more than anybody else. If you still want to give Tim Cook a title, I would say he was more like an admiral navigating the macroeconomic seas of supply chains. But he’s not the architect because he didn’t have to strategize anything. Terry Gou was strategizing. Tim Cook was following, but then doing more with what Terry Gou and others were coming up with than anybody else.

Jordan Schneider: Yeah, you have this line: “Nobody architected the move to China, but in one opportunity after another, Apple operations were lured into the country.” I think having a tri-continental strategy sounds ridiculous sitting here in 2025. But it’s also the case that you were scaling hardware manufacturing at an exponential rate for 20 years. If you’re making 10,000, 50,000 computers, it’s not insane.

But as you write about in your book, the number of actual items you’re making is increasing by a factor of 10 or by 50 or even by 100 on a year-on-year basis. It is getting more difficult and complex and the labor required is also rising at an exponential curve. You’re not going to find that many people in Wales and Singapore to do all this stuff.

Patrick McGee: I would emphasize another thing, which is that Moore’s law is decreasing the size of the products. Apple was not a big volume player in anything until 2003. I think in 2003, they made more iPods in a year than they made iMacs in the prior five years.

If you’re shipping things either by air or by sea, the fact that it’s the size of an iPod rather than a desktop computer is a major distinction. When SCI was building stuff in Huntsville, Alabama, it wouldn’t have made sense for them to move over to China and be early on this trend. That would make no sense at all. The computers were too big and bulky. Especially if you went back 10 years when it was a mainframe computer the size of a garage.

Once you’re into laptops, which is where Taiwan first makes its mark, it begins to make a lot of sense. Especially when you get into iPods and then dumb phones and especially smartphones, which then scale — that becomes a dramatic increase in efficiency. The cost per item — the logistics becomes de minimis once you’re doing it in the millions and let alone the tens and hundreds of millions.

Kyle Chan: Speaking of the logistics, I love the parts about United Airlines and the Boeing 747 playing a role in both directions. You’re talking about the product getting smaller and smaller to the point where you can have a Boeing full of iPhones that’s totally worth a flight, more than worth a flight. They’re helping to export this product around the world. At the same time, they’re providing the inputs in terms of the Apple engineers and the managers flying in.

It’s ridiculous — entire flights, I think you had documented, were for Apple staff to go back and forth between Cupertino and Zhengzhou or Shenzhen or other parts of China. It was astounding.

Patrick McGee: You’re finding this poetic symmetry that I didn’t make myself. Well done, Kyle. But to give that anecdote, because that’s pretty funny. I forget which cities it was and in what order, but in 2014 it’s probably Zhengzhou and in 2016 it’s probably Hangzhou, where the same thing happens where Apple convinces United Airlines: “Look, we need you flying nonstop to this place three times a week. It doesn’t matter if the rest of the plane is empty, we will pre-purchase so many of the first-class tickets that you will still make money.”

I think that’s hilarious. People who went to Shanghai and then took the bullet train to Hangzhou beforehand said it wasn’t exactly a schlep — it was pretty easy. But if you were an American engineer who had to go there regularly and you didn’t want to have to deal with the Chinese signs and all this kind of stuff, it was a nice comfort.. Because Apple had so many people going, it was worth them doing it. That was amazing.

The thing that was already out, but maybe people don’t know it, is in 2019, United accidentally leaked that 50 engineers from Apple were flying first class from San Francisco every day. They were the largest corporate client on the planet. When you talk about Apple’s size, there are many things you can point to that show they upend entire industries. But who knew that the airline industry was one of them?

Manufacturing the “Un-manufacturable”

Jordan Schneider: I want to talk about the design of the items themselves. You have this refrain over and over again where people say the contractors who don’t see the future ask, “Are you crazy? We can’t do this.” Then Apple comes up with a new manufacturing methodology that’s both labor-intensive and requires a level of sophistication that they have to teach to their contract manufacturers in order to execute.

Let’s do a case study. Pick your favorite item — which one haven’t you talked about on the other podcast that you want to discuss?

Patrick McGee: The translucent iMac is the one where the chapter is called “Unmanufacturable,” and that’s where I devote most attention. It’s also a computer. Everybody knows there’s one thing I hadn’t banked on — people say there’s a sense of nostalgia when they’re reading the early chapters because they’re going through some of the products they loved as high school or university students.

Jordan Schneider: Totally. I felt the same way. It was very tangible. I remember being 12 years old, going to my uncle’s house, seeing his candy-colored iMac and thinking, “What the hell?” It’s this object from space — but it’s not from space, it’s from these crazy engineers who were battling with Steve Jobs and the manufacturing people. Thank you for that.

These things are in design museums now and in your older cousins’ closets if they were smart enough to keep them.

Anyway, please continue, Patrick.

Patrick McGee: The computer was considered un-manufacturable by the product design team at Apple. The tooling engineers at Apple looked at what Jony Ive had come up with and said, “This can’t be built. I’m not saying it can’t be built at scale — we can’t build one of these in the lab.”

It’s interesting because they had created this thing that’s impossible, yet this is where you get the Apple mentality of “even though it’s impossible, we’re going to figure it out.” In the end, they don’t figure it out and the design has to be changed. But nevertheless, in hindsight, it looks like Jony Ive had tasked them with something that was impossible just for the sake of seeing who would stay on the project to get it done.

I’m not saying that was actually intentional — Jony Ive had to go back to the drawing board and make major changes to the product. But by the time they have a product that can be built, the only people left on the team are those who are willing to try and experiment with all these new things.

If people remember the computer, it has these translucent pinstripes on the front. Those used to be horizontal. What the engineers told me is that they couldn’t be horizontal because of how plastic injection molding works — the mold has to be parallel with the lines. Jony Ive had done something where they were perpendicular to the lines, and that couldn’t work.

They experimented for months. They had specialists come in from the outside to work on the product. What’s fascinating is that Steve Jobs had just come back. This was the product that, in my telling — because I’ve got all these Steve Jobs notes from the summer of 1997 — this is an interesting period where he’s given the role of CEO if he wants it, but he won’t take it. What his notes reveal is that he thinks Apple’s about to die, and he doesn’t want to oversee its demise. He is trying to help and see if it’s possible.

It’s a meeting of the minds with Jony Ive. Steve Jobs goes to the meeting expecting to fire him. Jony Ive goes to the meeting with his resignation letter in his pocket. Instead, they hit it off. What they come up with is the translucent iMac.

If you’ve seen the eMate 300, this translucent product they’d already made — it was a Newton and laptop mixed together — that was the first translucent product they made. You can see how Steve Jobs would’ve gotten excited by that and began to envision what he thought at the time was a network computer. That was their big idea.

They came up with this design. Steve Jobs got all excited about the idea of Apple making a comeback and redefining the aesthetics of the computer. But the first thing that Jony Ive comes up with literally couldn’t be made. The product design team goes through multiple iterations of who the leader is because nobody can do this, but Steve Jobs had inherited this team. He honestly didn’t know if they’re any good.

When they told him it was unmanufacturable and he goes on this tirade where he threatens to sell his “one last fucking remaining share of Apple stock,” he goes to his favorite design consultancy called Lunar. These two guys had actually designed the NeXT computer — that black magnesium cube, if you’re familiar with it. They go through all the blueprints and they agree with Apple’s team: “This cannot be built. This is not a quality design.”

I found this interesting because when I’ve read biographies of Jony Ive, even in grade seven, 12-year-old Jony is all about minimalist designs and elegant sketches.

For me, this was a real turning point where Jony Ive realized, “I need to know more about manufacturing to actually be able to give something that can truly be built.” He does have to change the lines. They do have to make the computer foggier than they wanted it to be because the inside was kind of ugly — the circuit boards were sticking out. That’s something they eventually changed by 1999 or 2000.

He had idealized this computer that couldn’t be built. But the changes that they made — I quote somebody else saying — they’re not major changes. The essence of the idea still lived. It’s still this egg-shaped thing rather than this angular square.

But it is the product that absolutely saves Apple. If they couldn’t figure out how to build it and how to change the design so that it was manufacturable — the term is DFA, designed for manufacturing — Apple wouldn’t exist. It was absolutely clear — the company was going to be bankrupt if this product wasn’t a hit, and it became America’s bestselling computer. It’s quite the dramatic narrative.

Jordan Schneider: I want to tie this to Ren Zhengfei 任正非. We’re running this episode after our series on Huawei, and this story is another one of these “leadership matters” anecdotes. The other thing that struck me about that is how far Steve Jobs was willing to go with the gimmick of, “Okay, this impossible thing — figure it out, figure it out, figure it out.”

There’s a lot of human pressure when all these people are saying, “No, you can’t do this.” Kyle, is there a Ren Zhengfei angle here?

Kyle Chan: Yeah, I think there is. Ren Zhengfei is famous for pushing his team to the limit and then some. There are all sorts of issues related to the work culture — people being driven too hard — but then that culture is carried with those personnel later on when they leave Huawei and go out to other parts of the Chinese ecosystem.

I see this parallel with a maniacal focus on realizing this concept, whatever it is — whether it’s going to be a cutting-edge Ascend AI chip or whether it’s going to be this candy-colored, perfect device that everyone wants to buy immediately when they see it in the store. That kind of drive is very reminiscent.

Terry Gou also plays a huge role in this story, and I see a similar parallel with his personality, although complementary in some ways, he’s so cheap, for lack of a better term. He’s focused on efficiency — he eschews the glitzy corporate HQ of Apple. But they share ultimately the same goal of producing at scale with ruthless efficiency these goods that will make all of them together a lot of money.

Patrick McGee: I’ve got to throw in the DAP here — the Divorce Avoidance Program — because so many marriages are broken up in the first five years of Steve Jobs’s comeback that the engineers are given these different policies to save their marriages. One of them is an understanding that, for example, there might be days when Jordan and Kyle aren’t going to be in to work because their marriages are on the line, and we have to give them a break.

That works well for a time, but we need Jordan and Kyle on the factory line. Instead, it becomes, “If we need to send them to Korea on a random Sunday even though they only got back two days earlier and their spouse is going to be upset, let’s give her $10,000. We need to assuage her because we don’t want to be losing engineers through attrition.”

That was happening to such a degree that all these policies needed to come out. It’s funny — well, I say it’s funny, it’s tragic as well. I’m getting the stories 20 years after the fact. People talk about the dollars being called “Danny Bucks” or “Dan Bucks” in reference to VP of Product Design Dan Riccio, because he was the person who fought for these things.

Sometimes our narrative is overly emphasizing how difficult the Foxconn workers worked, and that’s necessary. Obviously it gets to the point of suicides and suicide nets to prevent that from happening. But the Apple engineers were doing a similar number of hours while also having to fly back and forth all the time. It was causing so much strain that Apple had to institute these policies that engineers talked about. If you talk to anyone on the iMac project in the late 1990s and you mention the DAP, they haven’t thought about it for 20 years, but they immediately know what you’re talking about. Then they start telling stories.

I should say that one guy said, “Never mind the divorces, you need to look at the deaths.” That shook me. I was finding the DAP funny at times, and this person was saying there was nothing funny about it. He could rattle off names of people who had died on the production line or come back and died. I didn’t want to overemphasize that because as a journalist, I can’t determine, when someone died 15 years ago or 20 years ago, that overwork was the cause of death.

Steve Jobs himself says that the reason he thinks he got cancer is he was working so hard as CEO of both Pixar and Apple in 1997 that his immune system was weak and it allowed cancer to creep into his system. Whether that’s the medical diagnosis, I don’t know, but Steve Jobs understood that about himself or believed that about himself.

Jordan Schneider: Another interesting parallel is that the exit opportunities for these people existed. The compensation they were getting was not so different from what they could find at other firms from the Apple side that they couldn’t find other jobs. But there was something to this work that, similar to early Huawei, was not just the money. There is an aspect of team and mission and excitement that was around in this era of Apple that drove the Apple employees to give themselves and give all they had and more to this company and these products.

Patrick McGee: Jordan, let me agree on one part. I do think there’s a band of brothers, bunker mentality that would happen among the Apple engineers putting in the 18-hour days.

On the other hand, Apple later gets sued for collusion because there was an agreement among Silicon Valley giants not to hire from each other’s firms. One of the engineers I spoke with is someone who sued Apple later on because he didn’t understand at the time how being such a high-profile product design engineer at Apple wasn’t getting him job offers. He said, “I’m doing such good work. Why aren’t I being hired by Motorola or Google or whoever?”

Much later, he found out that was taboo. Steve Jobs had yelled at people and the emails came out later. I forget exactly what happened — I’m pretty sure they settled. The evidence was pretty good that there was collusion among the tech giants.

Subscribe now

Kyle Chan: Let’s talk about the example of the CNC machines. One of the parts that struck me — I had heard a little bit of this before, but reading it in the book, I’ve got to read this short snippet:

“Apple dumbfounded the industry when it purchased more than 10,000 CNC machines in a single year, enabling a form of mass production that Steve Jobs called ‘a whole new way of building notebooks.’ Apple even made a deal with Fanuc Automation Group from Japan to purchase its entire pipeline of CNC machines for years to come, shutting out all of its competitors from access, and then scoured the globe for more. ‘There were not enough CNC machines in the world to do the machining that we needed to do,’ one person said.”

I was struck by the scale and also by the — again, this is reminiscent of Elon Musk — “We need this part now. Whatever it takes, fly it out personally, make it happen.” For this Apple story, for the aluminum single-body MacBooks, it’s what it’s going to be. If it takes 10,000 CNC machines, then so be it. We’ll fly them over. Those stories capture the “whatever it takes” mentality.

Patrick McGee: We’ve jumped ahead to 2008 here. This was an achievement of imagination. There’s nothing novel about a CNC machine — I forget how many decades they go back. But the thing is, they build prototypes. Apple had the wherewithal and the imagination to say, “Well, wait a minute. If these things are so good, why don’t we build all of our products using these things?”

My focus is on the MacBook — they got the unibody MacBook — but they’re used in all sorts of products, including the iPhone, and have been ever since. These are machines that individually cost between $500,000 and a million dollars. Who knows what Apple is paying when they’re able to buy them at scale and they have someone like Tony Blevins negotiating for them.

This is so successful that this is the first time that MD (manufacturing design) is put on the map. In other words, it’s the first time Steve Jobs, in a keynote presentation, talks about their manufacturing prowess. This is probably the time — I don’t know this for a fact — that MD becomes MD. Before then, they were a part of product design, and it’s called Supply Base Engineering. Now they’re a part of ops.

It’s interesting that MD doesn’t have their own senior vice president. I don’t know exactly why it’s a part of ops — they should be their own thing — but it’s possible that Apple doesn’t love the narrative of more people knowing about this division because they are the ones where, when my book talks about the geopolitical influence that Apple has, it’s because they’re flying MD engineers who are brilliant, usually based in California, over to Asia to train and audit and supervise and equip these third-party manufacturers with what becomes billions upon billions of dollars of machinery.

This is where I get into the argument that Apple’s influence is like that of a nation-building program.

iPhones and Nation Building

Jordan Schneider: I have a smaller question — more on the design stuff. As the design process is being incredibly creative with the CNC machines, one of the questions I have is to what extent they could have optimized for design features which would have given them more geographic flexibility?

The degrees of freedom that Jony Ive and company were able to operate under are unique in the industry. As we've discussed with the CNC Apple Unibody story and the translucent iMac, they were willing and excited to push the physical design features of their products to the point where the answer at the end of the equation had to be China and Chinese suppliers.

I'm curious whether there were other pathways where they could have introduced different constraints at the design phase. This approach might have allowed them to manufacture more products in different places around the world without losing as much capability. The alternative would be trying to manufacture something like the unibody MacBook—as it was designed to be constructed in 2008—somewhere else outside of China, which would result in significant compromises.

Patrick McGee: That’s a great question. One question that you get now is: why doesn’t Apple automate these processes so that it’s not dependent on China, where you’ve got 400,000 people building an iPhone? That sounds like a rational thing to do.

But what you would be doing is upending how Apple designs its products, which is to say that Jony Ive’s team — and I’m aware that Jony Ive has left, but we’re talking about the golden era here — they could operate with a “no constraints” mentality. In other words, other companies will say from the get-go, “This is going to be automated, so we’re going to design for automation.”

The degrees of freedom under which Jony Ive and company were able to operate is unique in the industry. As we’ve talked through with the CNC Apple unibody story, as we’ve talked through with the translucent iMac, they were willing and excited to push the physical design features of their products such that the answer at the end of the equation had to be China and Chinese suppliers.

But I’m curious if there were other pathways where they could have introduced more or different constraints at the design phase such that you could have at the end of the equation been able to manufacture more stuff in different places around the world without losing as much — without losing as much as you would if you were trying to manufacture the unibody MacBook as it was designed to be constructed in 2008 somewhere else outside of China.

I like to ask people what their favorite Dell computer from the early 2000s was. Of course, none of us has a favorite Dell computer from the early 2000s. It’s not only the money that Apple’s making and investing in China that distinguishes them from others, because they were a small player in the early 2000s.

What distinguishes Apple is what they are doing and the fact that nobody knows how to build it. It’s not a knock against China to say that Apple taught them a lot. When you think of Apple products — whether it’s multi-touch glass or the anthropomorphic metal tubing between the base and the computer for the Sunflower iMac — nobody knows how to build that stuff. I’m not saying the Chinese didn’t know how; nobody knew. The marriage of Apple and China is skill and scale coming together.

The line I use is that Jony Ive and Steve Jobs made Apple products unique. Terry Gou and Tim Cook made them ubiquitous.

Now, if you had a different mentality where you said, “Let’s make these things easier to manufacture, more automation-friendly, to give us geopolitical safety,” that absolutely could have happened. You could have had more boring products, fewer design changes year to year, and you would’ve been able to have a more resilient supply chain based on production in Mexico, for example.

In that world, the first iPhone, if it were built at all, absolutely would have a plastic screen rather than a glass screen. It was only by working with the Chinese factories at their scale and industriousness that Steve Jobs was able to make dramatic changes. Literally after he has presented the iPhone — when he famously announced the iPhone in January 2007 — he was holding a phone that has a plastic screen. It was in the period between him announcing it and it going on sale that Apple upended the entire thing.

They work with a Taiwanese supplier called TPK, which builds the invisible circuitry within the glass so that your finger actually causes an effect. You’ll notice that if you do that on your window, it doesn’t do anything. You need to have a bunch of technology in there. Then they worked with Lens Technology to shape and temper the Corning glass that everyone’s familiar with.

The glass is American-made, but then it has to be shaped, tempered and cut. There are tens of thousands of people working at Shenzhen Lens to do that. You have people like Steve Zadesky, who’s on a whole bunch of iPod and iPhone patents, literally sleeping on the factory floor to get that all done.

China offered something that nobody else has offered. The reason why the epigraph to the book is a “Made in China 2025” document from 2015 that says:

“Without manufacturing there is no country and there is no nation.” Find me another country that has that as their motto, their mantra. You don’t get that just anywhere.

Sometimes I find it amusing that I have a 90-second sound bite on X and someone will say, “This guy’s saying China couldn’t be anything without Apple.” It’s the total opposite. I’m saying Apple is screwed without China because nobody will offer the things that they offer. They were a once-in-a-century partner to respond to the demands of Jony Ive.

If I’m fast-forwarding a bit, the only thing I would say is that Apple thought they were wearing the pants in this relationship for the first two decades — from 2000 to 2020. It’s only afterwards that there’s this realization that, “Wait a minute, we were lured in by the siren call of an emerging superpower who had all their companies willing to do all these things so that they would learn all the technology transfer and put America in a position, and all other industrialized nations in a position, where none of them can compete.” Everything has gone into China.

Kyle Chan: Can we talk about learning and this process of learning and technology transfer? To me, this is such a big deal — not only for Apple, but Apple exemplifies a lot. All these foreign firms that China in many cases very deliberately tried to bring into the country, tried to attract — maybe it’s by the central government, maybe it’s by local government, maybe it’s certain incentives, maybe it’s a whole concerted effort.

The point was to bring in these foreign firms, have them manufacture in China, share know-how, and train a whole generation of not only workers, but engineers and managers. The story of Apple and China captures that fantastically.

To give you a contrast, I love the quote: When Apple was working with Sony in Japan, you had an operations executive from Apple leaving Tokyo saying, “I was in Japan about five minutes, and Apple can teach the Japanese nothing.” This was earlier in the story, and this was a difficult product, but not as difficult maybe as some of the later ones. That story was not one of technology transfer.

Whereas later on, especially for the case of China and Foxconn, there was a very deliberate effort to use Apple as a source of upgrading your entire industrial ecosystem — upgrading your suppliers, upgrading your machinery, your equipment, all of the tacit knowledge, not only the blueprints and stuff that you can license and patent. Some of this stuff was not anywhere in existence. It was pioneered — they were all operating at the forefront.

Can you say more about this process of learning? I read a lot and talk a lot about, from the Chinese side, their very active process of trying to get as much as possible from every kind of foreign firm in high-tech industries. This process of technology transfer and learning.

Patrick McGee: In the late 1990s, Apple actually struggled to find contract manufacturers for its products because they were not seen as worth the effort. We have to remember that Apple was creating things in pretty small volumes, and they were very demanding. The workers at these Chinese factories often did not like working with Apple because you’ve got these overbearing engineers who are dictating all these things. Usually they’re dictating it through translators because these are American engineers who don’t speak the language.

I have these funny anecdotes that aren’t in the book where some engineers have a free weekend, so they decide to go to this factory that they’ve been training up. They cause such hell that when they’re trying to leave the following day, there’s a problem with the car — the wheel is flat or something. As the car drove, it was causing sparks to fly. They’re thinking, “I guess we’ll have to stay here another day.” Then they get a call from the leader of the factory who demands that the taxi driver drive even with the sparks flying and everything, because God forbid these Apple engineers come back because they’re such a pain in the neck.

Subscribe now

This is the sort of level of training that Apple is providing — the sort of stuff that makes you go mad. Yet obviously they’re very, very effective.

My narrative about Terry Gou is that he is the first person to understand: “This is not about making big volumes necessarily, and this is certainly not about making margins, but we’re getting tuition-free, on-the-ground training from these overbearing engineers who are willing to teach our cohorts all sorts of stuff.”

Apple’s orderly meets its match in Terry Gou and vice versa. This on-the-record anecdote from Tony Fadell is that he and his team would come in to train the engineers they’ve been working with and find not a face that they recognize because Terry has taken these graduates and put them on the Dell line or whatever, where they can use their newfound skill sets to actual money-making effect. He’s, in a sense, without Apple’s permission, started a new semester for these other engineers.

I don’t think the Chinese understood that until 2016. That is a key reason why you’ve had Foxconn being squeezed out the last couple of years and they’re having to go to India because Apple’s essentially telling them, “If you want to keep your market share, you need to go to India for us.”

The learning is absolutely fundamental to the relationship between Foxconn and Apple and to why all the Chinese factories are able to do so much. If you want, we can go into the wonky differences between ODM and OEM and why Foxconn does so much better than Quanta or Inventec or anybody else.

Jordan Schneider: Sure, let’s tell that story.

Patrick McGee: In the late 1990s, none of the Taiwanese wanted to be doing assembly. Assembly is how you get the orders, but it’s very low margin, it’s totally cutthroat. The likes of Inventec and Quanta and the predecessors to Pegatron are doing something called ODM work. “D” is for design, and design means more margin, and it also means more investment. It’s an investment that typically pays off.

What it means is that if you’re a Western manufacturer of a computer, you’re already offloading your manufacturing to a third party in Taiwan. But what this will allow you, if you work with an ODM, is that they’ll also do the research and development, and they’ll also do the design. You have less and less to do. It gets to the point where the Taiwanese can literally show a catalog of computer designs, and you choose one. Imagine how much that takes off your balance sheet.

It works very well. The problem, of course, is that the Taiwanese begin branding their own computers and competing against you. If there are times of scarcity of a certain product, they’re going to source it to their own companies rather than to you as the third party.

Apple never falls for this trap, in part because they’re wanting to do so much design themselves. It’s never appealing for someone to say, “We’ll do the design for you.” Foxconn maintains itself as an OEM, a manufacturer — the manufacturing-as-a-service model. They were condescended to because it’s not sexy, high-margin work. Yet Terry Gou has an understanding of Chinese politics that nobody else seems to have.

Kyle mentioned that this is what we get wrong about Chinese communism: it’s not only about the relationships with the federal officials, it’s the provincial officials because the cadres in various districts are incentivized for factory growth. They will offer tax exemptions and bonded zones and tailor-made policies and all sorts of stuff — getting in the labor from the hinterlands or whatever — so that your investment goes to that district and not some other cadre’s.

It’s worth knowing that that is one of the key distinctions between Soviet and Chinese communism. The Soviet system was top-down.

I don’t know that you’d call Chinese capitalism grassroots per se, but I compare it to federalism on steroids. It’s something that we miss all the time.

The reason why being an OEM is important is the labor intensity and the vertical integration that Foxconn introduces. Instead of doing design and R&D, they are just building. What makes them so good at building is having many clients and then allocating those resources. That allows Terry Gou to expand — literally in a real estate sense — building dormitories, building entertainment venues, and having these migrants set up shop in Shenzhen all the time. If you’re doing that, you are making the local cadres in Shenzhen look good.

He’s able to parlay his brilliant investment strategy, his labor intensity, into free machinery, free tools, more labor migrants, etc., in a way that nobody else is able to take advantage of. Apple didn’t understand Chinese politics until 2013 when they were made to, but Foxconn is their biggest partner and Foxconn understands it very, very well.

What’s tragic is Foxconn was never rewarded very much for this. Their margins fell the closer they get with Apple, but the political connections that Foxconn made were instrumental to Apple’s rise.

Made in China 2025

Jordan Schneider: I want to talk about that — the getting rewarded thing. I asked ChatGPT for the market cap of Apple and all of its near-ish competitors. We’ve got Apple at $3.25 trillion. We have Samsung at $300 billion, Xiaomi at $175 billion. Huawei, no one knows, but maybe $100-200 billion. Oppo and Vivo are rounding errors.

We have this big argument. The central thesis of your book is that this was a dangerous trade for America, but we did get a $3 trillion company out of it, even though there was all of this industrial upgrading that Apple helped China do. Setting aside the getting Chinese people out of poverty angle to this, there’s got to be something to be said for the fact that America is now home to one of the largest companies that the world has ever seen.

Patrick McGee: I’ve tried to answer this. If for whatever reason we were living in a simulation and the game ended now, then obviously Apple got more out of this relationship and it was a great deal for everybody, and we lived through a golden age of Silicon Valley focusing on software and China doing the hardware.

The problem is it’s probably not a simulation. The world continues and we’re now at a point where Apple has no Plan B. China for whatever reason has become more belligerent and wants to be self-sufficient. “Made in China 2025 中国制造2025” is a grand master plan to sever itself from the West in terms of automation, robotics, electronics, and other things including pharmaceuticals.

It seems pretty crazy to me that the world’s most valuable company doesn’t have any sort of Plan B. The Tim Cook doctrine, named in 2010, maybe 2011, was to own and control as much as possible and only enter markets where they could be number one or number two. As Ben Thompson from Stratechery constantly points out, the fact that everything’s manufactured in a fairly hostile country is the biggest violation of the Tim Cook strategy that could possibly exist.

Kyle Chan: Another parallel is the way that all of these U.S. chip designers depend on TSMC. I tweeted earlier — and it’s funny, the numbers still hold a year later — but Nvidia, Apple, Microsoft, Google, Amazon, and Meta have a combined market cap of $15 trillion, give or take some trillions. Yet they all rely on TSMC for advanced semiconductors, for AI chips, for smartphones.

It’s incredible that there’s this amount of value that’s created on the American side from these companies, purely from design. All those companies I’m talking about, maybe with the exception of Tesla, outsource all the manufacturing. For the high-end chips, that goes to TSMC. TSMC itself is probably one of the most valuable companies in Asia, but itself is an order of magnitude less.

There are two big questions. One is, how did we get to this point where we were allowing global risk to be concentrated so much on a single firm or a single country? Same thing with Apple and China. But then on the other side too, do you lose something from separating design and manufacturing?

Your stories about Apple engineers flying back and forth, deeply embedding themselves with Chinese suppliers, bringing the machinery, doing this — you have this phrase, “man-on-mirror” structure that I love. That was very interesting. That was actually for Korea, but a similar model later on for China, maybe 10x. How much can the two be separated and how much can you still be innovating at the cutting edge when you know your manufacturers are somewhere else, maybe in a different country altogether, maybe 6,000 miles away?

Jordan Schneider: I want to push back on that characterization. It’s a misnomer to say NVIDIA is a design firm. They’re constantly flying back and forth to TSMC. The innovation required to create not only the chips but these entire racks is incredibly manufacturing and hard tech intensive. It’s not only sitting with EDA tools and rearranging where circuits go. A similar story can be told from their perspective.

Patrick McGee: Andy Grove, Intel’s co-founder, gave a great interview to Bloomberg Businessweek in 2010, where he lamented that everything had left America’s shores. He said we’re going to miss out on shop floor innovation — that if you’re not in the factory working on the next product, you’re going to lose out. He was concerned about the shortsightedness of it, and that was quite a profound lesson.

I come back to this later in the book to highlight how any number of Western companies completely outsource without knowing how to build things anymore. It’s not clear to me that Facebook has any idea how to build the Portal. The Portal isn’t a device that’s out there anymore, but as I understand it, it was a product that was completely outsourced to Asian companies who knew how to do it.

Apple is in a different boat — they do know how to build things. They have the experiential know-how built on proprietary processes. We know about them doing iOS, but they also do software and operating systems for machines. We don’t know about that because we’re not on the factory lines.

The problem, as I quote from engineer Michael Hillman who has 16 years of experience, is that to execute any of those plans, they need China. The line I often use is that an iPhone has roughly 1,000 components. If you’re building a million iPhones a day, that means you’re managing the logistics, manufacturing, and production of one billion components daily. There’s one country on earth that’s able to execute that. Anyone else will take at least 10 to 15 years to reach that stage. Because China is an increasingly belligerent country, they won’t let that happen.

If you follow high-quality engineering on a map, America had total dominance at the end of World War II. They began teaching the Japanese to build radar and electronics industries. When they hit supply constraints and the yen rose in value, Japanese entrepreneurs went to their former colonies — Taiwan and South Korea — to do the same thing. When Taiwan hit its own constraints and had better relations with mainland China, the Taiwanese entrepreneurs 台商 went to China and built immense manufacturing capacity over the last four decades.

The next natural place to go would be India. Vietnam to some degree, but Vietnam doesn’t have the size. India is the next logical place if you play this game. But no one knows that better than Beijing, and they want technology transfer to be a one-way gate. The information comes in, but it certainly doesn’t go to Karnataka.

That’s where Apple has this big predicament right now. It makes perfect sense to set up production lines in India, but the machinery is now produced in India and blocked by Beijing. The experiential know-how is in the minds of Chinese engineers — from low-skilled workers to PhDs — and they can’t get visas to India. That’s why Apple is captured. I’m not using that as a line to sell books. I don’t know what strategic decision Apple can make that would untie the knot with China. The more plausible scenario is that they retie another knot and end up with a bifurcated supply chain. But that’s not in China’s interest, and if it’s not in China’s interest, then good luck to Cupertino — I don’t know how they’re going to execute that.

Subscribe now

Kyle Chan: This is a very big topic in the scholarly and development world about whether this whole “flying geese” model — Japan, South Korea, Taiwan, China’s moment — whether that ends with China. Apple is perhaps one of the best examples, but there are many other areas where maybe the auto industry will end up in China and there’s no next flying geese destination. Maybe a little bit to Vietnam and Thailand for transshipment to get around tariffs.

India is always the question. India wants to be next and has been waiting for a long time, watching China’s rise in anguish, thinking “that should be us.” They’re making a big push with manufacturing and have their own major industrial policy, especially targeting iPhone and consumer electronics manufacturing. To some extent, there seems to be success on paper — the oft-quoted statistic that 20% of iPhones are now assembled in India.

But there are big questions about how much of the real supply chain can be brought into the country the way China had done, whether that’s bringing in foreign firms to invest in manufacturing or cultivating domestic suppliers.

Jordan Schneider: The case study I’m looking forward to seeing play out over the next few years is wearables. Meta made all of the Oculus devices in China, and a few months ago, they said they’re going to try to get GoerTek, their Chinese contract manufacturer, to do half of it in Vietnam. Good luck with that.

Google’s now getting into the game. They had a big announcement a few months ago and did a big talent acquisition of HTC, which was a Taiwanese-based VR company, and are obviously trying to price all this in and diversify.

But you have six or seven Chinese manufacturers who are ahead of the game when it comes to what’s currently on the market in the West — when you look at Meta Ray-Bans as the main thing people have been exposed to. There’s that dynamic, but beyond creating redundancies around the world, it’s clear that doubling down on China will get you to the frontier of technology faster than trying to be cute in different countries.

The Apple Vision Pro — not a huge commercial success, but an incredible technical accomplishment — give me a break if you think you can try to make that at scale or with anywhere near the level of policy and precision that Chinese contract manufacturers were able to deliver.

Patrick McGee: It’s worth noting that’s Luxshare, not Foxconn — the rise of the red supply chain in action. The other example I’d point out is Sam Altman and Jony Ive coming up with their next device. Where’s that going to be built? If you’re Jony Ive, you might demand the best, and the easy thing would be to say, “Of course we’re going to work with the Chinese partners that he knows from three decades at Apple.”

On the other hand, he must be more awake to the issues I point to in the book than anybody. He might very well know the political agenda of Trump and all the ways that could be used in his favor to build out a nascent supply chain here.

It’s much easier to do when you’re beginning to build units literally in the tens because you’re doing prototypes, rather than in the quarter billion that Apple deals with — Apple’s scale is so big that sometimes you get off by an order of magnitude.

Apple in China, Future Edition 養虎為患

Kyle Chan: The phrase I kept thinking about was “training your replacement.” Apple is different, but there’s this pattern we notice repeatedly where China brings in foreign firms attracted by China’s efficient, nimble manufacturing base. They set up shop, and then — maybe it takes decades — you end up with a whole crop of Chinese competitors that first nibble away at your market share within China, then start expanding and going global.

You see this happening right now with the EV industry and Tesla. There’s a whole debate about how much Tesla turbocharged China’s EV industry or how much was already there, with Tesla playing a catfish role.

Reading your book, Apple had some of this problem. Samsung is the most famous long-term global competitor, but also Huawei and then a whole fleet of Chinese smartphone competitors — Xiaomi, Vivo, Oppo. You talk about how Apple kept trying to innovate on both design and manufacturing to stay a step ahead of the competition, at least up until recently. They were able to maintain their position in the global market, even within China, even with all the pressures of domestic competitors, in the high-end smartphone market.

Could you talk about this problem and how Apple dealt with this feeling of potentially creating the very competitors that become existential threats in other industries?

Patrick McGee: This is a great segue into one of my favorite chapters, Chapter 36, called “Five Alarm Fire”, because this is what happens when Apple realizes the competitors in China have caught up with them.

What’s amazing about the chapter is that it’s exclusively based on court documents that were made public but were never found by another journalist. I found them 16 months ago, but I’m the only person who’s written about them, even since I published the book nobody has gone to do their own deep dive. We’re talking about more than a thousand pages — depositions of Tim Cook and others, internal emails between Tim Cook and the board of directors at Apple, between him and production staff and his underlings.

The entire issue is about how the iPhone XR (pronounced “ten R”) is a dud, and they know it’s a dud. It’s geared toward the Chinese market, yet the Chinese are all deciding to buy a competitor from Huawei. Apple knew this for several weeks and then told investors that everything was fine.

The chapter is called “Five Alarm Fire” because that’s what a VP of Sales or Finance described the situation as — fire department jargon for when at least a hundred people are needed on the scene. The quote from Tim Cook, at least a week before he talks to investors, is: “This is a disaster. We need all hands on deck now.” There are people within Apple who are panicking about this.

What happened is that the tenth-anniversary iPhone is the first one with that infinity pool design — you get rid of the thumbprint and have Face ID. That didn’t ship until November 2017. Then, by April 2018, four or five months later, all four of the major Chinese brands have an infinity pool-like rival. Cupertino was expecting to bask in the glow of this amazing device, and literally everybody caught up.

You realize the Chinese went from mimicry to — maybe they’re still in the stage of mimicry there — but they’re surprising Apple with how quickly they’re able to keep up. Now, fast-forwarding, they’re in a stage where they’re doing things that Apple isn’t doing. The example I always point to is the Mate XT from Huawei, which unfolds twice. There are also silicon carbide batteries in the latest Vivo phones that have better density, faster recharge time, and longer duration.

Apple is no longer the company to pay attention to if you’re looking for the most cutting-edge phone. There’s something disquieting about that, especially because not only are the Chinese manufacturing them, but they’re doing the industrial design and product design. They’ve taken that pyramid structure I described in the book, put it organically in their own companies, and are now outmaneuvering Apple on multiple levels of the pyramid.

Kyle Chan: Now we’re seeing this especially with AI features coming out. Huawei is doing a lot with its Harmony operating system, trying to have AI everywhere you want it — and maybe in many places where you don’t, like summaries where you don’t want summaries, maybe mimicking what iPhone users are getting used to now with Apple Intelligence.

There’s this big question mark about whether Apple will be able to catch up and be on the cutting edge in that space. There are recent issues with linking up with Alibaba, for example, for an AI partner in China. If you don’t have that kind of partnership, whether as Apple, you can still be competitive in the Chinese domestic market.

Patrick McGee: You brought this up because this is mostly post-manuscript for me — it’s not in the book, yet it’s current and dynamic as a discussion. First, Apple has done poorly with Siri. Siri might have been amazing 14 years ago, but it hasn’t kept up. It’s not even a large language model, so it’s not doing the most basic stuff.

Because of Siri’s failure, Apple has to work with a partner. Because ChatGPT or anyone else, such as Perplexity AI isn’t available in China, they have to work with a Chinese partner. In a sense, that’s the closest thing we’re going to see on the product side to Apple having a joint venture — maybe it’s not technically a joint venture, but my book is about how Apple consolidated all hardware engineering into China over the last 25 years.

Are you telling me that, over the next 25 years, they’re going to work hand-in-glove with Baidu or Alibaba and essentially do the same thing with AI?

This is already the stuff of movies regarding existential risk to humanity. We’re talking about the most complex, consequential thing in tech, if not in business or anything writ large. This is up there with fire and the computer. I don’t know how much sense it makes for Apple to be training one of these companies to make their product better, which they would inevitably have to do. They’re not going to use an off-the-shelf AI tool.

Jordan Schneider: I don’t know if they can train themselves to make their product better when it comes to AI.

Patrick McGee: That’s a great point. We are at the stage where you’ll buy your next phone based on the AI features it has, and Apple is very much failing there. We are recording this during WWDC, so who knows? Maybe we’re way wrong by the time this comes out, but I feel like that’s not going to be the case. They’re behind in AI in an embarrassing way.

Jordan Schneider: One more detailed question before we do a closing one or two. The $55 billion annual investment in China — can you deconstruct that number? Where did it come from and what is included?

Patrick McGee: Apple was on the back foot in 2013 when Xi Jinping 习近平 came to power. They’re worried that either their products are going to be blacklisted or they’re going to have to form joint ventures to continue operating in the country. This is total anathema to Cupertino — not what they want to happen.

This team of people that came into the company, either hired or appointed, called themselves “the gang of eight.” They’re the first people who are senior, living and working in the country. This is where you pass the baton from Foxconn to Apple. Apple realized: we can outsource our manufacturing to Foxconn, but we cannot outsource our political relationships.

They take ownership of this and do their own supply chain study effectively in tandem with the government affairs team. What they realized is that they’re investing $55 billion into Chinese factories — into factories operating in the country.

China has this concept called “registered capital.” Apple, in any given year, doesn’t make as much money as Walmart, and Walmart gets 60 to 70% of their products from China. But Walmart’s not making investments in the country. If you’re importing kitchenware and action toys from China, that doesn’t count as registered capital or investment — that’s just spend. You’re buying that stuff, putting it on a ship, and selling it in Arkansas.

If you were Volkswagen or GM, you’re training workers to run a production line. For a certain period — 18 to 24 months — it counts as an investment into the country. The training costs of setting up a production line count as fixed assets, the same way that putting machinery on the production line is counted.

Apple realizes their investment is more like Volkswagen or GM’s than Walmart’s. They’re not sitting in Hong Kong ordering parts from Shenzhen. They’re going into the factory and training them how to do all this stuff. Their training costs, which are effectively wages, should be counted as fixed assets, as registered capital.

But they go further. Unlike Volkswagen or GM, who set up a production line and have it running for seven years because the lifecycle of a car is about seven years, the iPhone never gets to that stage where it’s running. They’re always upending the design of every product in the portfolio. They’re able to count a whole lot of training costs and wages, in addition to billions in machinery that’s put on the production line, as registered capital.

Someone could debate whether all that spending should actually be counted as investment, but it’s not an argument with me — it’s an argument with Apple and with Beijing, which accepted the argument. That’s where the money comes from, why it’s counted as registered capital, and what distinguishes them from other companies that don’t do the training. Other companies are buying off-the-shelf parts, whereas Apple is working hand-in-glove with hundreds of factories.

That’s why they’re able to operate without a joint venture — because they can say, “You have no idea what influence we’re having on hundreds of factories across the country.”

Jordan Schneider: For Kyle and Patrick, the contract manufacturers that Apple shepherded — do you have a sense of how their relationship with Chinese handset manufacturers is similar or different?

Kyle Chan: There’s a lot of overlap. You can break it down — Lens Technology is a supplier not only to other Chinese smartphone manufacturers but to others as well. GoerTek and some of these suppliers supply across the board.

The order of operations depends, but some of them were brought up by Apple and then helped the rest of the Chinese supply chain. Some of them were already serving what were, for them, lower-tier customers like Oppo and Vivo before they got their Apple contract. It’s a mixed bag.

I also love the part where you mention Apple trying to navigate Chinese politics, because good luck to anyone trying to navigate that.

They tried to cozy up to Sun Zhengcai 孙政才, who was the party secretary of Chongqing. Apple tried to invest in data centers there, which at a local government politics level is great — you want to show that you’re bringing in investment, generating economic activity, especially when it comes to anything high-tech. That’s a bunch of gold stars.

That turned out to be a mistake because Sun didn’t end up becoming the successor to Xi — it turned out Xi was the successor to Xi. Sun didn’t even end up in the very top echelon of party leadership.

There are other ways where Apple has been incredibly adept, especially Tim Cook, at navigating Chinese politics and American politics at the same time. It’s not a job I would envy, although he’s probably pretty well compensated.

Mood Music:

This is a repost of an article by Jon Sine of the Cogitations substack.

And Suffering Will Be Your Teacher

Xi Zhongxun was born into a fallen world. That, at least, is something the father and the son, whose youth was forged in the Cultural Revolution, have in common. But if we speak literally rather than metaphorically, the world of the elder Xi was not fallen, but falling apart. He was born in 1913, in the desolate northwest of China, as a scourge of European guns, germs, and steel was unleashing forces that would sweep the world’s great agrarian empires—Ottoman Turkey, Romanov Russia, and of course Qing China—into the dustbin of history. These same forces would pull Zhongxun, like so many young radicals of his time, into the dark vortex from which modernity would crawl. The birth of modernity in China was a bloody and terrifying upheaval. Anything, Mao quipped, but a dinner party.

Modern minds may struggle to comprehend the youthful Zhongxun. One could begin, as The Party’s Interests Come First does, with the shocking story of how he tried to murder his teacher at the tender age of 14. One could describe how famine stalked Zhongxun’s family, distending his belly and those of his orphaned siblings, claiming several of them. One could note that Zhongxun’s first wife was only eligible because her first husband had his head severed from his body by one of the various warlords and militias—and it was she, an eighteen year old girl, who had to find and bury the carcass.

Images of death and suffering have burst on to our social media feeds in the 21^st century. For most of us they disappear as quickly as they arrive, a flash on the screen sent away by the next swipe of the finger. Inhabiting the mindscape of Xi Zhongxun is hard, perhaps impossible—it would require experiencing things we never have, lingering on them for more time than we are comfortable or perhaps capable. Terrible and turbulent days of unceasing insecurity, death, and suffering are mercifully, for most of us, foreign.

But Zhongxun and his revolutionary kin came to know suffering like a first language. It spawned in them a zeal for purpose and meaning; a drive to find something that could not only bring order to a chaotic present, but something that could redeem a fallen world, that could make sense of seemingly senseless suffering. Suffering shaped and inexorably drew people toward causes bigger than themselves. Toward things that, as Viktor Frankl would have understood, transformed the very meaning of suffering. For some, like Zhongxun, suffering became the crucible in which the meaning of their lives was forged.

There was nothing inevitable about Xi’s trajectory, however. Subjected to similar suffering, any number of men or women might have chosen a different path. For Xi, the road he walked was shaped as much by happenstance as by conviction—introduced to the basic concepts of communism, almost by accident, first by a teacher and then by an educated prison mate. As Torigian reveals in his prodigious excavation of Zhongxun’s life, Xi would later recall that he knew nothing of communism when he first joined the cause. It was not the Communist Manifesto, Lenin’s Imperialism, or any other Marxist-Leninist tract that first kindled his passion. It was his own suffering, reflected back to him in the pages of a novel: The Young Wanderer. “If other poets,” the author, Jiang Guangci, would write, “pride themselves on being artists ahead of their time—creators of beauty—then I pride myself on being a true son of the age, a singer of the storm.” Much as Stalin drew strength from Georgian heroic tales (from which he also took a nickname, Koba), Xi, too, found inspiration in literature. By chance and circumstance, his creed became the communist one, and his devotion was given to the Communist Party.

If Zhongxun’s conversion to the communist cause was shaped by chance and far from foreordained, it was also part of a larger pattern. Like the old Bolsheviks profiled in Slezkine’s House of Government, Xi was one of many youths who gave themselves wholly to a cause. And once he made that choice, he never wavered. The many bloody trials and tribulations he endured in service of the Party—the only force he genuinely came to believe that could save China—hardened a commitment that would prove unshakable. Part of this devotion was deeply personal. With both parents dead and little family to call his own, the Party became his surrogate kin. The “forging” he underwent in those early revolutionary years—from infiltrating the Nationalists in his teens and twenties, surviving multiple assassination attempts, to becoming one of the youngest pioneers of the northwest base where the Communists would eventually settle in Yan’an, a place where he would both purge and be purged—took on a fetishistic significance in later memory.

For many Chinese Communists, and revolutionaries the world over, suffering was their teacher. And it taught them that only total devotion to the cause, of which the Party was the sacred embodiment, could deliver salvation. A close reading of Torigian’s biography enables the modern mind to feel this.

Do Not Wait Until The Evening to See How Splendid the Day Has Been

As Torigian reminds us, Zhongxun is often remembered in popular conception as one of the most humane figures the Party ever produced—held up as a symbol of reform at its most principled, a legacy now sometimes invoked to cast Xi Jinping as an unfilial son who has betrayed his father’s path. But as Torigian painstakingly shows, Zhongxun defies easy categorization. The elder Xi’s life, Torigian writes, is “a powerful statement about the misleading nature of grand narratives” (page 535).

Xi the “reformer” initially opposed the household responsibility system. He despised materialism. He railed against the corrosive threat of individualism. As Torigian writes, quoting him intermittently:

“Individualism was a ‘germ.’ Even if small, this germ ‘does not fear the heavens, does not fear the earth, and it is extraordinarily daring.’ He continued, ‘Even if you weigh eighty kilograms, even if today you have only a tiny, tiny bit of individualism in your body, once it develops, it will devour you whole.’” (p. 194)

Zhongxun, for his part, routinely and enthusiastically affirmed his commitment to Deng’s Four Cardinal Principles. The tension between the “Three” and the “Four”—as Torigian frames it, the economically reformist spirit of the 1978 3rd Plenum versus the enduring imperative of the Four Cardinal Principles to uphold the Party’s authoritarian core—runs through the Party as well as Xi.

A Lighter Side to Xi Zhongxun: At Disney Land (1980)

Underlying it all is the deeper tension at the core of Zhongxun’s life—the conflict between humaneness (人性) and Partyness (党性)—between what one thinks is right and what the party demands or deems acceptable. Despite his difficult youth, Zhongxun somehow sustained a gentler, more conciliatory side. It was further honed through years working in the United Front, where he often favored dialogue and co-optation of local power brokers over coercion in dealings with ethnic minorities, religious groups, and other non-Party actors. One striking example comes from policy in Xinjiang, where, as head of the Northwest Bureau, Xi intervened all the way to the top to overturn a hardline approach pushed by Deng Liqun and Wang Zhen, figures typically cast as staunch conservatives. In doing so, he was willing to bear the enmity of Deng and Wang (which would surface in due time) so as to forestall a more repressive campaign against religious believers and nomads there, favoring instead a more peaceful strategy of co-opting and courting local power brokers.

And yet, as the aptly named book reflects, when the Party’s interests were on the line, they always came first. If the Party needed someone eliminated, Zhongxun would—and did—comply. As he did in Xi’an in the early 1950s, fulfilling Mao’s mandated execution quotas, and earlier still during the Shaanxi base area purges of the 1940s, when Xi was county secretary of Suide. Zhongxun also remained conspicuously silent during the Tiananmen crisis, despite holding the prominent, and at the time very relevant, post of NPC Vice Chair—perhaps shrewdly foreseeing Deng’s violent verdict and not wanting, once more, to end up on the wrong side of Party history.

A Tougher Side of Xi Zhongxun: His Portrayal In Hagiographic New 39 Episode Miniseries (2024)

The prodigious depth of Torigian’s research invites comparison with Robert Caro. In Working, Caro recounts a formative piece of advice from his early days: “Turn every page.” Across roughly 2,000 endnotes and an extraordinary range of Chinese-language sources, Torigian shows the same tireless commitment. His book is not only a biography, but a repository of primary sources and a work of translation, with many materials rendered into English in full or in part for the first time.

Amid the mosaic of sources, three stood out to me. First is Xi’s official Chinese biographer, Jia Juchuan, whose three-volume life of Zhongxun is thoroughly mined. Torigian describes Jia as “more likely to omit than mislead,” and his account is triangulated with other key works like The Chronicle of Xi Zhongxun (习仲勋年谱) and Biography of Xi Zhongxun (习仲勋传). Second is Li Rui—former Mao secretary and high-level Party insider—whose private diaries, now archived at Stanford, offer invaluable glimpses into elite politics. Torigian draws on them carefully to add texture and insight. Third is Warren Sun, a towering figure in Australian China studies. Alongside Fred Teiwes, Sun has long set the standard for rigorous political analysis, and has done more than most to recalibrate Deng’s legacy—as both architect of reform and calculating autocrat, including his role in sidelining the more consensus-minded Hua Guofeng.

But if Torigian’s research ethic recalls Caro, his narrative style diverges sharply. There is no luminous epigraph urging readers to “wait until the evening to see how splendid the day has been,” as in The Power Broker. Instead, contradictions are foregrounded, ambiguities embraced, and tidy storylines deliberately refused. This is a book that resists resolution. Considering how often—and for how long—both internal participants and outside observers have misunderstood Chinese politics, such caution is not only reasonable but necessary. So many statements must remain provisional, with multiple streams of evidence pointing in suggestive directions but rarely converging into certainty. Such are the realities of writing about a man operating at the center of a system that is built around secrecy.

One of the central lessons of The Party’s Interests Come First is that the familiar labels used in China-watching—reformer versus conservative, or more morally charged binaries like good versus bad—often collapse under scrutiny. Torigian dismantles these categories, showing how supposed heroes were less heroic than assumed, and villains less villainous. At the top of the regime, everyone was, at one time or another, sometimes simultaneously, a victim and a perpetrator. That is the nature of the Party system. No one’s hands are clean, though some are more stained than others.

Inside the Leninist Machine

A focus on the characteristics and pathologies of Leninist systems, central to Torigian’s excellent first book on succession in China and the USSR, remains at the forefront of The Party’s Interests Come First. This new work, however, offers a more intimate portrait of how easily one can misplay their hand in the murky world of Leninist power politics—a setting in which prestige, manipulation, and coercion prevail. The intentionally hierarchical design of these regimes makes them especially leader-friendly: the top leader reliably stands above the rules and norms, able to reshape them at will. As a result, institutionalization at the apex is exceedingly difficult, if not impossible. Even among the highest ranks, elites operate in an environment of profound opacity, often unsure of the core leader’s true intentions, the current alignment of political forces, or what must be said or done to maintain their position.

Twice in his career, Xi Zhongxun served as a chief implementer to the regime’s chief implementers: first under Zhou Enlai in the State Council of the 1950s, and later under Hu Yaobang in the Secretariat of the 1980s. In both roles, Xi witnessed firsthand how precarious elite politics could be. In early 1958, Mao turned sharply against Zhou for trying to moderate the Great Leap Forward (“Oppose Rash Advance”), stripping the State Council of its economic authority, creating five new small groups to oversee government work, and handing control over the economy to Deng Xiaoping and the Secretariat (who then presided over the most unrestrained phase of the disastrous campaign). Decades later, as a member of the Secretariat, Xi again observed how the Secretariat and the State Council, now under Zhao Ziyang, vied for influence, and how Hu Yaobang—often described as the conscience of the Party—was ultimately purged by Deng. Ironically, as Torigian determines, Deng repeated a pattern he had twice suffered himself under Mao: purging a deputy not for disloyalty or policy differences, but simply because his confidence in him had mercurially wavered (pp. 472–3).

The Old Revolutionaries, Before Their Scattering, On Their Way to Physical Labor

The much-touted “institutionalization” of Chinese politics under Deng Xiaoping is revealed as largely illusory. Torigian’s detailed reconstruction of events surrounding Deng’s autocratic and often arbitrary purges—of Hua Guofeng, Hu Yaobang, Zhao Ziyang (and, as an aside, nearly Jiang Zemin)—paints a far less orderly picture. Deng emerges, in the words of Li Rui, as “half a Mao”: a leader who deliberately preserved a two-line system that concentrated immense discretionary power in his own hands while leaving others to operate in a state of calculated uncertainty. When he did intervene, as in the decision to use force at Tiananmen, it was often abrupt, unconsultative, and final. In this light, elite politics under Xi Jinping appears less an aberration than return to form.

Throughout the book, Torigian revisits key inflection points in Party history: the fall of Gao Gang, the Great Leap Forward, the purge of Peng Dehuai, the emergence of the Special Economic Zones (SEZs), and the ouster of Hu Yaobang, among others. In each case, he draws on an extensive array of sources, often supplemented by anonymous interviews with insiders, to show just how easily both outsiders and insiders have misunderstood events. One striking example is the birth of the SEZs, where Hua Guofeng—alongside Xi Zhongxun in Guangdong—emerges as a central architect. Deng Xiaoping, contrary to the prevailing view, did not formally endorse the zones until relatively late in 1984, after their early success had become evident. “By then,” Torigian notes, “Deng was taking credit for everything. “I proposed setting up special economic zones,” he said. “It appears that the correct path was taken” (page 285). Western textbooks have since gone along with Deng and Party historiography in doing Hua a lasting disservice, reducing him to a caricature as the Maoist ideologue clinging to the “Two Whatevers,” remembered if at all through a dismissive throwaway line.

Factions play a surprisingly minor role in this account, in contrast to works like Victor Shih’s Coalitions of the Weak or Cheng Li’s earlier scholarship, which emphasize the role of patronage networks and elite affiliations. Patronage networks, often assumed to be relatively coherent, emerge here as more fluid and contingent. While factional analysis remains appealing—particularly because ties can be quantified through shared hometowns, schools, or bureaucratic overlap—Torigian shows that many presumed alignments, such as Xi Zhongxun’s with Gao Gang or Peng Dehuai, were overstated, misunderstood, even among Party insiders. Old networks did matter, but they were also often unreliable. Just as typical, power and alignment shifted through opportunism or convenience, as in the case of Ye Jianying’s unexpected role in restoring Xi Zhongxun and landing him in charge of Guangdong, and in the case of Wang Zhen who, despite the seeming historical enmity regarding Xinjiang policy, became the first person to speak up for Zhongxun’s rehabilitation in 1977 (page 251).

Ideology, too, comes in for scrutiny. The arbitrariness of ideological labels often obscures more than it reveals. Party leaders, like the institution itself, frequently held contradictory positions simultaneously. As today, this does not tend to produce dialectical synthesis so much as oscillation and confusion on the part of internal participants and external observers. Torigian is especially pointed in one aside: “Guessing about whether Jinping cares more about ‘ideology/security’ or ‘development’ is a distraction from the basic point that the Party has always cared about both, even though the pursuit of two such goals simultaneously inevitably creates tensions” (p. 543).

What It Is, And What It Isn’t

This is not a book about Xi Jinping—though, if judging by much of the popular media coverage thus far, one could be forgiven for thinking otherwise. There are, to be sure, sections that probe deeply into the younger Xi, most notably Chapter 21, Princeling Politics, which unearths fascinating details about his early rise through the Party in the 1980s. Torigian draws on sources affiliated with the Organization Department and the Young Cadre Bureau (a bureau responsible for identifying and fast tracking promising cadres) to reconstruct how Xi was perceived by contemporaries at the time, including revealing diary entries from Li Rui, who was then working in the OrgDep.

There are also moments of striking personal detail I had never encountered before—such as a scene near the end of the Cultural Revolution, when Xi Jinping visits his father in exile. In a sweltering apartment in Luoyang, where the elder Xi had been sent to labor in a tractor factory, father and son sit smoking cigarettes in their underwear as the younger Xi recites Mao’s speeches from memory, with Qi Xin watching.

Subscribe now

The book does not examine Xi Jinping’s policies today, but it offers a window into how he has internalized his commitment to the Party. Jinping once spoke of confronting profound doubt during his years as a sent-down youth in Liangjiahe. But as with his father, it ultimately appears to not have shaken but deepened commitment to the party. Suffering, for him, became meaningful when understood as a sacrifice. His father never abandoned his loyalty to the Party, despite being purged over a novel, spending sixteen years (1962-1978) in the political wilderness, and suffering greatly during the Cultural Revolution. Torigian counsels: “While some may wonder why Jinping would remain so devoted to an organization that severely persecuted his own father, perhaps the better question is, How could Jinping betray the Party for which his father sacrificed so much?” (p. 539).

Xi Zhongxun and Xi Jinping in Luoyang (1975)

Finally, beyond Xi Jinping, the book is not even best understood as a biography of Xi Zhongxun. Its deeper purpose is in using Zhongxun as a lens through which to examine the history and internal contradictions of the Chinese Communist Party itself. What the book delivers is not just a portrait of the man who fathered China’s current leader, but a window into the moral and political structure of the Party that shaped them both.

Forging Red Genes

If there is a single throughline in Xi Zhongxun’s life, it is this: devotion to the Party above all else. The forging of loyalty through suffering did not remain his legacy alone. It reverberates in Xi Jinping’s own mythology as well. For both father and son, “struggle” and transformation into a loyal cadre take on a near-fetishistic significance, each held up as a model of what a true revolutionary should endure, and ultimately become.

Yet, as Torigian warns, there is no determinism in this process of forging. There is no guarantee that those who “suffer for a cause” emerge more committed. Just as easily, they may come out disillusioned, embittered, or broken. Zhongxun’s own children illustrate this variance: one committed suicide, one became a rule-of-law advocate, several chased money and pleasure, and one pursued power. Xi Jinping may have internalized his father’s legacy—may even have “inherited the red genes”—but the real cliffhanger is whether the next generations can or will.

This remains a central quandary for the Communist Party today: how to cultivate loyalty through struggle in an era defined not by war or revolution, but by peace and development, and how to do so without alienating the very people it seeks to inspire. In China today one frequently encounters the slogan 永远跟党走—“Forever walk with the Party.” But one finds it hard to imagine that China’s increasingly urbane, educated, and independently-minded elite dream of goose-stepping into eternity.

One wonders if Xi Jinping sees the germ of individualism spreading, as his father once did, threatening to devour the party whole. In 2017, a new slogan was popularized: 幸福都是奋斗出来的—“All happiness comes from struggle.” But what kind of meaningful struggle do today’s rising cadres face? What crucible of hardship might shape them into passionately devoted party members, as it once did their leader? And to what lengths might Xi be willing to go to find out?

Coda

Be warned: Torigian’s is not a beginner-friendly book. Readers unfamiliar with the terrain of modern Chinese history or the historiography laid down by earlier giants like Roderick MacFarquhar, Ezra Vogel, and Richard Baum may find themselves overwhelmed by the detail, unsure of the stakes, and unaware of previous interpretations in the field. Likewise, media headline writers and soundbite chasers will find little to grasp onto—and what they do seize upon, as early evidence indicates, may not well reflect the book’s actual content. Torigian refuses sweeping statements, brash generalizations, or seductive narrative arcs. “In many cases,” Torigian writes, “the vagaries of evidence and intention simply cannot bear the weight of the big questions we would most like to ask” (page 536). But those willing to forge insight and meaning out of complexity, contradiction, and contingency will likely recognize The Party’s Interests Come First as a major achievement—one of the finest works on China in the past decade, written by one of the greatest American China scholars of his generation.

Buy Joseph’s book!

And Subscribe to Jon’s substack

Cogitations

Writing that excavates the social scaffolding.

By Jonathon P Sine

RAND’s Compute Cluster is Hiring

We’re particularly excited about applications from ML engineers and semiconductor experts eager to shape AI policy as well as seeking excellent generalists excited to join our fast-paced, impact-oriented team. Find an overview here. More details below.

The team's work focuses on using compute as a governance tool, but extends to technical AI governance more broadly, including: technical mechanisms for AI governance (e.g., verifying AI agreements, hardware-enabled mechanisms), AI infrastructure (policies, trends, and forecasts), and export controls (designing effective restrictions, assessing their impact, and fixing them). We've achieved strong "product-market fit" for our work—our technical analyses inform major policy decisions and leading AI companies and governments regularly seek our input.

The roles are: Technical AI Policy Associate & Technical AI Policy Research Scientist (senior). Applications are accepted on a rolling basis.

Why China Wants to Steal the Secrets to a Chunky Soviet Hovercraft

Lily Ottinger reports:

Earlier this month, the New York Times obtained an internal document from the Russian Federal Security Service detailing the threat of Chinese espionage. The report specifically outlines a Chinese campaign to snag Soviet aerospace engineers:

China has long lagged behind Russia in its aviation expertise, and the document says that Beijing has made that a priority target. China is targeting military pilots and researchers in aerohydrodynamics, control systems and aeroelasticity. Also being sought out, according to the document, are Russian specialists who worked on the discontinued ekranoplan, a hovercraft-type warship first deployed by the Soviet Union.

“Priority recruitment is given to former employees of aircraft factories and research institutes, as well as current employees who are dissatisfied with the closure of the ekranoplan development program by the Russian Ministry of Defense or who are experiencing financial difficulties,” the report says.

An ekranoplan (literally “screenglider”) is an airplane-esque vehicle designed to fly closely above a body of water, utilizing the ground effect to reduce drag and achieve greater fuel efficiency. From 1966 to 1988, the world’s largest and heaviest aircraft was a classified Soviet ekranoplan dubbed “The Caspian Sea Monster,” which had a maximum takeoff weight of 544,000 kg and a wingspan of 37.6 meters. Since they fly just a few meters above the water, Ekranoplans operate outside the range of detection for many radar systems.

But why is China so interested in acquiring this technology?

Ekranoplans could possibly be used to ferry troops across the Taiwan Strait (the US Navy estimated that some Soviet ekranoplans could carry up to 850 troops or two tanks), although flying over the open ocean can be challenging for ground effect vehicles. The water below the craft must be calm, with waves below 1.25 meters in height; otherwise, the air cushion becomes unstable.1

Regardless, it would technically be possible for Ekranoplan-style warships to fly over the Taiwan Strait on calmer days, and the PLA isn’t considering launching an invasion in the middle of typhoon season anyway.2

With influence from Soviet designs, China has built smaller Ekranoplans like the DXF-100, the Albatross-5 (信天翁5), and the Neptune-1 (海王一号), which can hold 15 to 20 passengers. But it seems that China is more inclined to apply techniques of Ekranoplan design to other technologies. Ekranoplan engineers are intimately familiar with both hydrodynamics and aerodynamics, so perhaps China’s simply believes that these are the most cost-effective engineers to target. But apart from generic overlap with shipbuilding and aircraft design, there is a direct technological crossover between ekranoplans and wing in ground effect drones (WIG UAVs), which are basically like tiny unmanned ekranoplans. The first reports of Chinese military WIG drones surfaced in 2017, and were quickly recirculated by state media rather than being censored. WIG drones are also being developed by Gdańsk University of Technology in Poland and a Danish startup. These drones could be used for naval reconnaissance, transporting goods, or delivering payloads, all while flying outside the range of aircraft detection radar.

While the extent of China’s espionage activities in Russia doesn’t bode well for their partnership (I highly recommend you read the whole NYT article), China appears to have extracted plenty of value from Russian scientists already. Hopefully, Taiwan has a plan to deal with low-altitude amphibious drone swarms. Who knows? Maybe Taiwan has its own team of disillusioned Soviet scientists waiting in the wings.

How China’s Gen Z Is Exporting Chinese Soft Power to the World

Selina Xu is a writer and researcher on technology. She was a former China reporter at Bloomberg News.

Helen Zhang is the co-founder of Intrigue Media and a non-resident fellow in the United States Studies Centre's Emerging Technology Program. She was previously an Australian diplomat.

Since America’s “Liberation Day” tariff blitz, a lot has been said about China’s economic and technological self-reliance, which has given it more leverage in this trade standoff. Under Xi Jinping, China has steadily focused on reducing dependence on Western supply chains and the U.S. dollar, while swamping the world with goods.

Much less has been said about China’s growing cultural self-sufficiency and ability to export soft power. Just a decade ago, Marvel movies topped the Chinese box office while Japanese video games and Taiwanese soap operas occupied the pastimes of youths. In 2025, the highest-grossing movie in China (and in the world) is Ne Zha 2, an animated retelling of a traditional Chinese myth. On Youtube and other streaming platforms, historical costume dramas — often featuring palace intrigue or celestial romance — have gained traction with overseas audiences, especially in Southeast Asia. On phones and PCs, Chinese games like “Genshin Impact” are winning hundreds of millions of players at home and abroad.

The ascendance of domestic content is in part a result of the Chinese government’s push for national rejuvenation through restriction of foreign content—for example, a nearly decade-long unofficial ban on Korean entertainment, including K-pop, when South Korea angered China by agreeing to allow a U.S. missile-defense system on its soil. On April 10, China said it would cut back Hollywood films in retaliation for U.S. tariffs.

But part of this phenomenon is driven by China’s Generation Z, a 270-million-strong cohort born since the mid-1990s, who are more culturally confident and cosmopolitan in their tastes — and willing to pay for good content. Already, Gen Z accounts for 40% of consumption in China, and their influence will only grow, with spending set to surge fourfold to 16 trillion yuan ($2.2 trillion) by 2035.

In recent years, blockbusters like Chang An and Jiang Ziya and popular TV series like Empresses in the Palace and Nirvana in Fire underscore how the youth are gravitating towards traditional Chinese heritage. Some of these draw from epics like Journey to the West, which is one of China’s most-read literary masterpieces, and The Investiture of the Gods, a 16th-century fantasy novel about gods and demons. Others are set in various historical dynasties, but one that comes to the fore is the Tang Dynasty (A.D. 618-907) — dubbed China’s golden age — when its empire was at its most powerful, and when the ancient Silk Road was at its peak. Some have attributed this solely to nationalism, but Gen Z’s love of history is authentic, imbued partly by an education system emphasizing “five thousand years of Chinese civilization.” In 2024, over 62 percent of visitors to China’s national museum were under the age of 35. Wearing hanfu — a style of clothing with flowing robes that dates back more than two millennia — has also ballooned from a niche hobby to a billion-dollar market in China and a global movement on TikTok. When one of us visited Xi’an last year, the ancient capital’s city walls were overflowing with young people dressed in hanfu.

China isn’t alone in indulging in nostalgic, domestic revival — in some ways, this isn’t too different from Donald Trump’s “Make America Great Again,” Britain’s Brexit reminiscence of its pastoral and colonial past, or Putin’s harking back to imperial glory. But Xi Jinping’s ambitions go beyond a mere exercise in nationalism, and should not be construed as isolationism. In his April 16 essay in Qiushi Journal, the flagship magazine of the CCP, Xi wrote, “To build a culturally strong nation, we must more proactively present China’s perspectives, spread Chinese culture, and showcase China’s image — ensuring our soft power matches our hard power, converting our developmental advantages into discourse power.” Beijing is keen on deepening ties with regional neighbors, especially Southeast Asia, which has also been hard-hit by American tariffs but already has a glut of Chinese goods. To bind South-East Asia’s economy more tightly to China’s, soft power can grease the wheels.

One area where Chinese soft power is growing is the gaming industry. By some counts, over 500 million people in China are consumers of anime, comics, and gaming, most of whom grew up watching Japanese anime but are increasingly embracing local content. The rise of the genre can be seen in the trajectory of Bilibili Inc, a Chinese streaming platform that started off as a niche site for anime and gaming fans but has now become a $9 billion public company that dictates mainstream trends — the platform has about 30 million paid subscribers (that’s more than ESPN), who are mostly Gen Z. Young Chinese men are playing “Genshin Impact” and “Honor of Kings” — two of the world’s most lucrative mobile titles and both Chinese-made — while women are playing Chinese otome games that have interactive romance storylines. As local game studios beef up to cater to increasing interest at home, many of these games are also making waves abroad. For instance, the wildly popular “Love and Deepspace” became the most-downloaded and top-grossing interactive story mobile game in Japan last year. Four of the ten top-grossing game publishers in the world last year were from China, according to analytics company AppMagic.

After decades of importing content from abroad, China is now exporting culture to the rest of the world.

China Gen Z’s tastes in apps and brands are also making inroads overseas. Rednote, which has billed itself as a “lifestyle bible” and is especially popular among young women, is China’s fastest-growing social media platform. The company, a surprise winner of America’s early-2025 TikTok ban, has seen global daily active users up 28% in March from last December. On the app, users share lifestyle content featuring a dizzying array of Chinese brands — many of which are now coming to the West. One example is Pop Mart. The maker of Labubu dolls saw its non-mainland revenue grow by 375% in 2024, accounting for about 40% of its total revenue. Another example is Chinese bubble tea — including brands like Molly Tea and HEYTEA — which have popped up on the streets of New York and California, with distinctive aesthetics, lounge-like ambience, and some selling branded tote bags and cups à la Starbucks. In a sign of their growth, at least four Chinese bubble tea brands are preparing to go public in Hong Kong.

To be sure, the government has been a visible hand guiding tastes, though not often successfully. In recent years, alongside a tech clampdown, the Chinese government has tightened its grip on cultural industries, banning “effeminate” men and hip-hop culture on TV, cracking down on idol fangroups, and championing programs that “vigorously promote excellent Chinese traditional culture, revolutionary culture and advanced socialist culture.” More nationalist epics featuring Chinese resistance efforts during the Sino-Japanese War have dominated the silver screens, alongside anti-corruption TV series like In the Name of the People and The Knockout.

As the U.S. turns more isolationist, slashing foreign aid and imposing tariffs on developing countries that depend on export-driven growth, China now has an unprecedented soft-power opportunity to fill the void. While China’s ability to step in could be constrained by the economic challenges it faces at home (the country has scaled back on big infrastructural loans), cultural and technological exports — from games and movies to TikTok and RedNote — will be one way for China to draw closer to the Global South. As Beijing looks to find other outlets for trade, we expect it to wield more soft power, turbocharged by Gen Z consumption, especially in fast-growing markets like Southeast Asia, Latin America, and Africa. After decades of Hollywood and Silicon Valley’s dominance, the world is now standing on the cusp of China Inc.

Subscribe now

Dissecting Taiwan’s Chip Industry

Aqib is a graduate of Harvard University from the Regional Studies—East Asia program. Today, he presents his research on Taiwan’s semiconductor industry.

With waves of export controls from the United States and economic pressure from China, Taiwan’s semiconductor industry — its crown jewel — has been facing the heat from both sides. But headlines often miss the mark by addressing the industry as a monolith. With lengthy supply chains including electronic design automation (EDA) software, equipment, design, manufacturing, and packing, the different sub-industries all have their own economics, and their experiences all vary.

So let’s look at a few examples from the most-discussed sub-industries: advanced manufacturing, mature manufacturing, advanced design, and mature design. Their realities have all been warped in different ways by the recent geopolitical landscape; some have been cruising while others have been collapsing, yet two factors commonly mark their experiences: the tenacity of Chinese companies to get the chips they want, and the AI boom creating profits for anyone who can latch onto it.

Advanced Manufacturing: TSMC

TSMC, has fared remarkably well since BIS export controls. Recent news of the U.S. banning TSMC from all AI chip exports to China in 2024 and a potential billion-dollar fine may seem frightening, but these incidents are overshadowed by TSMC’s continual growth in the Chinese and global market from AI demand. The reason for TSMC’s staying power in the Chinese market (as shown below) is that there is simply no other alternative. If you want to make an advanced chip for AI or other high-performance computing (HPC) applications, TSMC is the only company that can do so. Despite export controls, TSMC’s revenue from China has only increased, and the share of revenue from China has remained relatively steady.

In the face of export controls, how do Chinese designers place orders with TSMC? The answer lies in downgrading and going for so-called efficiency rather than raw power. Chinese AI companies like MetaX (沐曦) and Enflame (燧原科技) reportedly downgrade their chip designs to be just within performance restrictions enforced by BIS export controls. Besides simple downgrading, Chinese companies have begun to focus more on ASICs and FPGAs, less versatile yet still strong chips that can be programmed for specific applications. BITMAIN (比特大陸), which was the cause of TSMC’s recent explosion of sales to China, has been able to buy up leading-edge 3nm chips from TSMC by designing ASICs for Bitcoin mining and AI applications.

How effective these downgraded chips and ASICs are is still an open question. Of course, Chinese companies will say their chips are comparable to GPUs from NVIDIA, and, in theory, Chinese firms can go far with such chips. Basically, instead of asking for a juiced-up GPU that can do everything, they are designing an ASIC that can do a limited set of tasks just as well but flounder at everything else. This strategy could allow China’s AI push to persist despite export controls, especially if they can make up for the weaker semiconductors with better code.

But regardless of whether these chips accomplish their goals, Chinese firms continue to buy them, and thus, TSMC continues to prosper.

Mature Manufacturing: Powerchip

However, the same cannot be said for Taiwan’s mature manufacturing foundries. A perfect storm of COVID-19 and increased Chinese competition has plunged companies like Powerchip into darkness, as the graph below shows.

Mature node manufacturers like SMIC and Hua Hong have been running Taiwanese firms out of business. With subsidies enabling Chinese fabs to cut costs and pressure for Mainland companies to “buy Chinese,” Powerchip is losing the battle for the Chinese market, and Taiwan’s mature chip industry needs to find business elsewhere.

Despite the downturn, Powerchip has found a few growth strategies that serve as a model for Taiwan’s other mature foundries, like UMC. One of these ideas is the Fab IP model. As governments increasingly treat semiconductors as a national security product, Powerchip is attempting to monetize their experience in making and running fabs.

In the Fab IP model, Powerchip signs agreements with other countries and assists in fab planning and operations, while ideally avoiding the construction and operating costs. Powerchip signed such an agreement with India’s Tata Group, which agrees to pay Powerchip royalties for technology transfer while raising the funds for the fab itself. The Indian fab won’t be operational until 2026, but the Fab IP model opens doors for Powerchip’s business. They can no longer compete with Chinese firms on price, but maybe they can compete vicariously through Indian or other foreign fabs. Powerchip is reportedly in talks with Thailand, Saudi Arabia, Israel, and Poland over similar agreements.

Powerchip’s other method for survival is latching onto the AI boom. For TSMC to make AI chips, they require CoWoS packaging technology, which relies on relatively unsophisticated silicon interposers. With CoWoS demand greatly surpassing supply, Powerchip has been able to insert itself into the AI supply chain. In 2024, PSMC opened a new fab in Tongluo, Taiwan, dedicated partially to manufacturing the silicon interposers required for CoWoS for TSMC. Thus, PSMC can still ride the advanced manufacturing AI wave. Advanced packaging demand has shown no signs of slowing down. With TSMC intending to ramp up CoWoS nearly threefold by 2026, Powerchip’s Tongluo fab will certainly be needed.

Advanced Design: Alchip

Taiwan’s advanced design industry has been able to thrive despite export controls. Let’s take Alchip, Taiwan’s #1 AI company, as an example. Their revenue has skyrocketed in recent years, and they have been able to pivot from the Chinese market to the American one as an engine for its growth. (This pivot accelerated when the U.S. placed their biggest customer, Pythium, on the Entity List in 2021.)

Alchip’s success is partially based on its unique position as an ASIC designer during the AI boom. Chinese customers, particularly automakers, still like to use Alchip, since their products are usually not restricted by export controls. However, although the cost of designing a leading-edge chip can run hundreds of millions of dollars, other Mainland designers exist at the leading edge.

Besides its specialty in ASIC design, Alchip has found success through a partnership with TSMC. As a “pure-play design company,” Alchip maintains a close partnership with TSMC, a pure-play foundry, and the design company has a knack for reserving limited fab capacity at TSMC. In particular, Alchip has often been able to gain “capacity support” for the critical CoWoS packaging mentioned earlier.

As American companies are chomping at the bit for AI chips, Taiwanese design benefit from being right next door to TSMC. Alchip has assisted Amazon and Intel in designing their own AI chips to compete with NVIDIA’s GPUs. For the next two years, Alchip orders are skyrocketing with chips for just these two companies, and these orders enable Alchip to keep pushing to the next node.

Taiwan’s advanced design companies have lost out on Chinese business either from customers getting Entity Listed or from Mainland competitors, but these losses have coincided with explosive growth from AI demand. The growth has greatly outweighed the losses, and advanced designers do not seem to be under fire.

Mature Design: Weltrend

The tragic character in Taiwan’s semiconductor soap opera is the island’s mature design sector. This sub-industry has historically been the most reliant on the Chinese market, and these firms are the ones facing the toughest fallout from Chinese competition.

These companies have limited options for survival. Some are attempting to switch to using Mainland fabs to manufacture their chips, risking unintentional technology transfer for only marginal benefits in cost. Mainland competitors can sell chips at a price that would only cover the production costs of Taiwanese mature firms, thanks to subsidies and government help. Taiwan’s mature companies are not as easily able to pivot to the world market either — in the mature chip market, cost is everything. No one wants their TV or microwaves or other analog products to be more expensive.

So how can these companies survive? Taiwan’s Weltrend puts forth one route for relief: taking advantage of the AI boom. Unable to design leading-edge GPUs or ASICs, though, Weltrend is attempting to cement its niche in server cooling fans.

By offering the best chips for server cooling by combining their design skills with developed algorithms, Weltrend hopes it can raise sales for its products as the rest of AI sales go up. These server cooling chips are needed in every data center. This kind of niche in the AI periphery is what some mature firms call the “garnishes for the steak.” They cannot compete with advanced nodes to be the main show, and they cannot compete with China on price. But by finding an irreplaceable position in the AI ecosystem, perhaps mature companies can survive.

Is This a Problem?

Is the collapse of Taiwan’s mature design companies a crisis that must be averted? Depends on who you ask. When speaking to some representatives of advanced design companies, I’ve heard people say that they “want mature designers to die” so that profitable companies can soak up their valuable talent. If this is the case, then perhaps it’s okay for mature designers to dwindle. Maybe Taiwan is simply moving up the supply chain to advanced nodes and leaving the cheaper mature nodes to China.

Of course, mature design companies don’t see it that way. Many are convinced that mature chips must be afforded the same protections as advanced chips. Perhaps it is a national security risk if all our server cooling chips can only be made in China.

If mature chips are just the garnish, then maybe it’s okay for them to fall, as long as we have the steak. But mature chip companies also tend to liken the industry to cars. The car needs its hood and headlights too, not just the flashy engine.

This plight opens new questions for policymakers in the U.S. and Taiwan. Should we be protecting mature chips? If so, how can we protect mature chips? It’s hard to ban based on performance without banning everything under the sun, so policymakers will need to find creative ways to protect the industry.

Tariff-proof Tea

Bryan Cheong is from Singapore, and lives and works on software in San Francisco. You can follow him on X here.

Most of the overseas tea merchants with sources or warehouses based in China that I love have paused their shipments to the United States after the latest tariffs were imposed and the de minimis exemption was struck. But the US has no shortage of tea collectors who have amassed vast stores of aged tea, who found customers even from across the Pacific. Two such collectors are based in the California Bay Area. Probably the second-largest of these belongs to Roy Fong, the founder of the Imperial Tea Court in San Francisco’s Ferry Building, who counts among his stores a 1980s puerh collectionfrom the Menghai Tea Factory, which crossed the border more than 40 years ago and is safe from additional customs and duties. This does not mean that the tea is cheap — aged puerh tea has been prized particularly by the Cantonese in Hong Kong and Southeast Asia. Roy Fong had also previously invested heavily in trying to grow tea cultivars in California, but the arid conditions of the state proved a difficult environment for the plants to flourish in. Nevertheless, tea is an adaptable plant, and forcing it to try to grow under new conditions is how we get new cultivars like Taiwan’s high mountain oolong bushes. Fong might have succeeded eventually, but alas, a fire in 2017 destroyed most of his tea plants and a portion of his puerh collection, so California’s aspirations for domestically producing tea will have to wait for another pioneer. Notable among the Imperial Tea Court’s offerings are the Special Reserve Ripe Puerh, available in cake form and loose, which have been collected and stored in California for the last 40 years. Puerh tea is made from the large-leaf variety of the tea plant, and is grown in Yunnan province in China. Yunnan is the ancestral heartland of the wild tea tree, and ripe puerh is an artificially fermented tea that mellows the large and astringent leaves into an earthy plum-coloured brew. The Special Reserve Puerh can be ordered online, but if you are in San Francisco, I invite you to try it in person at the Hong Kong-style teahouse at the Imperial Tea Court. On the nose, it is like dry leaf litter mingled with moss, sprinkled with a dusting of Ceylon cinnamon. On the tongue, it is sweet and clean, its age has mellowed any muddiness or bitterness and turned the tea rich and smooth with mineral undertones, and it does not turn bitter no matter how long you steep it. In the stomach, it is comforting and warming. The leaves will survive many, many steepings.

I spent two weeks in May visiting family in Tel Aviv. What follows are some scattered impressions.

A week before my flight the Houthis hit Ben Gurion airport so all flights were cancelled except El Al’s. Half the gates were unused at the airport and there were no non-Jewish tourists.

Houthi missiles were better from a lifestyle perspective than Hezbollah or Hamas ones as you get five minutes’ warning instead of just 60 seconds. It surprised me how diligent most people still are in going into shelters. Waking up at 3am is no fun but at least the baby’s crib was in the shelter already.

I met up with an Israeli-American reporter who covers politics for western media. “When my mom in Maryland calls me up to complain about what she saw Trump doing on MSNBC, I pat her on the head and let her know how much worse it can get.”

The hostages occupy enormous mental headspace in every Israeli. Murals and bumper stickers drape the physical space of the city, parents and relatives are constantly on tv. The hostages and their stories seem like a shared language (like how this book review characterized the Marvel Universe today) where first names hold totemic value.

At a playground one day a helicopter flew over and everyone got concerned as they knew it was a military medical transport flying back from Gaza.

Vibes

Is there a city that does good weather better than TLV on a Thursday night? Restaurants and bars pour out into street after street, you see groups of 20+, often intergenerational, out together. Fridays are much quieter as many go home to see their families. Does Israel have the answer for the loneliness epidemic?

No international djs come anymore. Boiler Room memory holed the shows it put on in Israel, but they’re still getting protested because they got bought out by KKR who is apparently too pro-Israel for their audience.

Tel Aviv needs to get abundance-pilled. Rent and housing costs are both maybe 85% of new york city.

Cigarettes are everywhere. Ten percent of Israeli teens smoke, not France’s 15% but far higher than America’s 1.5%. Fancy restaurants have outdoor smoking and non-smoking sections.

Folks called up from reserve duty go out with their gun in flip flops because they have to report the next day and it’s more convenient to carry around.

Taxis charge a 25% surcharge on shabbat.

I ran into one Chinese guy, maybe 30, who’s in town to sell lightbulbs (“we’re midmarket, can’t compete with Philips in America but too fancy for Africa and LatAm!”). He had no idea about the missile warnings.

Food

I do not like salads in general but immensely enjoyed every one I had in TLV. Dishes were always just on the edge of being over-salted. The country has share plate-maxxed.

The burger culture is extremely strong. The most outlandish attention-getting twist on the formula I saw was one shop sprinkling sugar on the bun.

Many restaurants were short-staffed due to reservist call-ups.

Babies

Secular Israelis have more babies than any other comparably rich country. You feel it in the street with lots of families and young children everywhere with even upscale restaurants all welcome to kids. I wonder just how having such ultra-fertile religious sects pushes these numbers up.

People are far friendlier to my 9-month-old than in new york and they’re also more than happy to tell you how to parent. I got told off three times for having a Doona stroller as apparently it’s bad for babies’ backs?

Seeing my child play with other kids too young to know war when Gaza is a 90 minute drive away redefined cognitive dissonance for me. While I was there, former Prime Minister Ehud Olmert published an oped saying that “What we are doing in Gaza now is a war of devastation: indiscriminate, limitless, cruel and criminal killing of civilians…. Yes, Israel is committing war crimes.”

And speaking of war crimes…

Books I Read

Arendt—Eichmann in Jerusalem

I would have read this years ago if anyone told me how funny it was! As she wrote to a friend, “You are the only reader to understand what otherwise I have never admitted—namely that I wrote Eichimann in Jerusalem in a curious state of euphoria.” It shows and is so much better for it.

World-historically funny.

And now a little darker…

This is in dialogue with what Olmert said on Ezra recently.

You have to change the nature of the dialogue and the appeal to the Israeli people and start to talk in a different way. Instead of warning us all the time that we are on the verge of destruction, which is what this government is doing now for 15 years, not just in the last couple of years.

I remember the days when I was fighting Hezbollah. After my war against them, everyone said how we failed and so on and so forth. But a few years later, I started to hear that they are so powerful, that there is a danger to the very existence of Israel if Hezbollah attacks Israel. And we keep hearing all the time that Iran is threatening the very existence of the state of Israel. And we hear also about Hamas today — these days, when Netanyahu talks, why does he need to explain the war? Because Hamas can become a danger to the very existence of the state of Israel.

This has to change. You have to open a dialogue with Israeli society on a different basis, on the basis of hope — something which will change the lifestyle and the hopes of the younger generations. Then we will not have to fight all the time.

“Look, the guy keeps winning, he must be doing something right!”

On the moral hollowness of the generals who waited until 1944 to turn on Hitler.

Career myopia, Nazi-style

Can’t stop pulling quotes from this book it is just too perfect. Of course, literally no-one who made it to Nuremberg has enough guts to defend Nazi ideology—otherwise they would have stuck it out in the bunker.

She closes chapters better than anyone I’ve come across. “Totalitarian domination tried to establish holes of oblivion into which all deeds, good and evil would disappear, but holes of oblivion do not exist. One man will always be left to tell the story.”

TLV Restaurant Recommendations and some more clown excerpts from Eichmann in Jerusalem

Read more

Eva Dou is the author of The House of Huawei, an excellent book covering the personal, economic, and geopolitical arc of Huawei, China’s most important company.

We discuss…

• The life of Huawei’s founder, Ren Zhengfei, who rose from Cultural Revolution disgrace to become one of China’s richest businessmen,

• How Ren built Huawei, and what makes their corporate culture unique,

• Huawei’s strategic entry into developing and high-risk markets like Libya, Iraq, and Iran, and whether the controversial deal with the UK is a threat to national security,

• How Huawei outcompeted Chinese state-owned telecom companies and eventually achieved national champion status,

• How Ren’s personal interest in foreign art, music, and architecture advances Huawei’s market share.

Co-hosting today is Kyle Chan, a postdoc at Princeton and author of the High Capacity Substack.

Have a listen on Apple Podcasts, Spotify, or your favorite podcast app.

Forged in Fire 浴火重生

Jordan Schneider: Eva, let’s start with the Cultural Revolution. How did it impact the future founder of Huawei, Ren Zhengfei, in his early years?

Eva Dou: Huawei’s founder, Ren Zhengfei, comes from quite a humble background. He grew up in rural Guizhou Province, where both his father and mother worked as schoolteachers. His father eventually became a principal, and they instilled in him the importance of learning and a love of reading that continued throughout his life.

At the same time, that academic background made them a target when the Cultural Revolution came around during Ren Zhengfei’s teenage years. His father had worked briefly for the Nationalists earlier in his career, making him one of the many people targeted during the Cultural Revolution. He was struggled against at his school, criticized by students, and eventually put in a labor camp.

This affected his entire family — Ren Zhengfei and all his siblings. It impacted their work prospects throughout his early career. He would later complain that he was unable to get the promotions that his peers received, no matter how hard he worked. His company Huawei is now considered case number one of a national champion in China, but he really started out on the outside of the system, looking in. That’s part of what makes his story arc so dramatic — how far he came.

Jordan Schneider: There are a number of passages that were really evocative for me. There’s this theme of suicide that runs through the book. Ren Zhengfei is an incredible workaholic with multiple divorces who at one point said he was excited to use Chinese national holidays to travel abroad and get more work done.

This connects back to his father’s experience during the Cultural Revolution. You have this line that said Ren Moxun 任摩逊, his dad, also considered ending his life, but he didn’t want to die before his name was cleared, as it would leave his wife and children with a cloud over them. Ren Zhengfei explained that if his father died, his children would have to carry this political burden. “He endured 100 tortures, but would not kill himself.”

Later, Ren Zhengfei said, “The Cultural Revolution was a disaster for the nation, but for us, it was a baptism. It made me politically mature so that I wasn’t a simple bookworm”

When you look at Huawei from the outside and see the handsets, AI accelerators, and base stations, at its core is this stupendously driven human being who was shattered to his core. There are many business stories in China where people come out of the Cultural Revolution — Xi Jinping as well — forged into these super-driven individuals. But obviously tens of millions of people didn’t turn out that way. Even the ones who did end up succeeding beyond their wildest teenage imaginations carry very deep scars that play out over the subsequent decades of their lives.

Eva Dou: Many people have pointed out that Ren Zhengfei, as an entrepreneur and business person, carries this pessimism with him that some attribute as one reason for Huawei’s success. He’s always looking at the worst-case scenario that could happen to his company. Even in the years when business was flush, he would warn his employees that something bad could always happen — they could go bankrupt at any moment and had to be prepared for anything.

These early experiences of facing life-and-death stakes and seeing how fickle the political environment and business can be really did inform him. In many ways, it helped make Huawei a company that could survive unexpected things, which we’ve seen through two Trump administrations and many unexpected developments.

Jordan Schneider: There’s a bit you have later where you talk about how he painted black swans everywhere once he built this giant campus. The Cultural Revolution was the ultimate black swan. Once you live through that, what’s a couple of sanctions violations?

Another illustration of the psychological jumps you have to make to process and move forward from living through that time comes from this incredible quote from his dad when reflecting about his experience being struggled against in middle school. You write: “One day, one of his old students visited him and recalled the struggle session where a classmate had beaten him with a wooden stick until it broke. Ren’s father smiled grimly. ‘I have to thank that piece of wood. Had it been sturdier, I surely would have been beaten until something was wrong with me.’”

It’s a lot. Kyle, are there other business stories that come to mind — aside from Xi Jinping and Ren Zhengfei, which we’ll be exploring in an upcoming episode with Joseph Torigian — about this kind of dramatic arc for Chinese business people?

Kyle Chan: What’s really interesting is that there’s a strong parallel between Ren Zhengfei’s worldview and the view of many Chinese entrepreneurs during these tougher years, as well as Chinese political leaders in their attitudes towards China’s development. This idea especially that we are operating in a very tough world, that we have to rely on ourselves, that there’s danger at every corner and risk, and what to do about that preemptively.

Jordan Schneider: On the arc towards being a national champion, Ren works for the PLA as an engineer for a few years, finally makes it into the party — which was a dramatic arc in and of itself — and then gets invited to this famous Deng National Science Conference in 1978. There, Deng Xiaoping essentially blessed everyone, saying it’s okay to start businesses and do science — you don’t necessarily have to live inside the system to serve the country. That was a big psychological unlock, almost in the way you describe how he reflects on that. Can you talk about these years? Let’s bring the timeline up through 1978.

Eva Dou: The first part of Ren Zhengfei’s career was as an engineer in the engineering corps for China’s military. During this period, he was largely low-key. Occasionally he would gain kudos from superiors for his work, but he wasn’t anyone famous by any means. He probably thought that’s where he was going to stay for his entire career, his entire life.

This abruptly changed because Deng Xiaoping decided to make this switch to begin the capitalist experiment in China. This was coupled with a dramatic downsizing of the military, and he was laid off from the military and sent to Shenzhen to work in this burgeoning private sector.

He’s talked about feeling very disoriented at the time. This was similar to the experience of many people in China who had been in a system that was only a planned economy and suddenly it was all different. They had new ideas, new fashion styles, and new trends coming in from Hong Kong over the border to Shenzhen. He had this sense that he was really far behind — he was already almost 40 at that time — in trying to adapt to this totally new world.

Jordan Schneider: You have this quote where Ren would later tell his colleagues that he’d wasted time during those years when he was outside the establishment, not as a party member. You can read this quote two ways: “I was a soldier for all those years I didn’t join the party. My life was full of adversity. When I think of all that wasted time, I wonder how could I have been so naive and ridiculous that I didn’t understand it all being about compromise and shades of gray."

On one hand, I see that as maybe this guy’s still kind of pissed that the party ruined his parents’ life. But also, there was no way he was going to be accepted as a party member, even though he was an engineer for the PLA, because of his class background — his nationalist background. What’s the right way to understand that quote, Eva?

Eva Dou: I agree, it’s a bit ambiguous. He does give this kind of counsel to his staff repeatedly through the years, advising them basically that politics is not fair. He’s told them straight out that life isn’t fair and you have to keep in line with the political system that we live in. He said Huawei is a Chinese company and it’s a requirement to be patriotic to be part of the company.

That’s informed partly by his earlier experience with the Cultural Revolution. But there’s another part that isn’t talked about much, which is the Tiananmen democracy protests in the late 1980s. It’s remembered largely as an event in Beijing at Tiananmen Square, but these protests were going on across the country. Some of the largest ones were in Shenzhen, where the young people there were very progressive, international, liberal, and bold. They held some of the largest protests.

That was at the time that Ren’s daughter, Meng Wanzhou, was in high school and about to go to college in Shenzhen. Those protests were quashed in Shenzhen as well as Beijing. That also informs the approach and worldview of Ren Zhengfei and other business people of his generation. They remember these things.

Jordan Schneider: It comes back to the thing that Kyle brought up earlier: to what extent does he really believe the nationalist stuff versus just being a business person who wants to do cool science and compete with Ericsson because that’s cool and will make you money and make you remembered?

You have all these quotes where they’re laying it on too thick at times with how much they’re using the nationalist angle and the national strengthening angle to motivate employees. But for anyone in that generation, there has to be some kind of internal narrative that not all is right in the state of Denmark. I don’t really know where to go with that.

Eva Dou: There is one comparison that I look at in the book. Before Huawei in the 1980s, China’s most promising tech company was this company called Stone Group. The founder of that company was bold and radical in his political beliefs. He actually supported the pro-democracy protests at Tiananmen, and it basically ended up with his company being wiped off the map in China.

He ended up having to flee the country shortly after the Tiananmen crackdown. He was never able to return to China again and has lived as an exile ever since. Ren Zhengfei’s generation of entrepreneurs grew up revering this company and these entrepreneurs, and they saw what happened there. Since then, it’s been taken for granted that you have to bend the knee to the political line to build a successful business in China.

Jordan Schneider: One of the interesting leitmotifs of this book is Huawei’s relationship to the state. The story that you talk about, and that Doug Fuller goes into more deeply in his book, is the fact that Ren actually kept the state at much more of an arm’s length than many of his competitors in the 90s and 2000s. He focused on export discipline and investing in R&D as opposed to the cheap, easy money that you can get from government contracts. That was the way he kept Huawei advancing at the technological frontier.

Let’s give a little sense of what the industrial upgrading arc that Huawei went on in the 80s and 90s looked like, and what decisions they made different from competitors who ultimately weren’t able to achieve the levels of greatness that Huawei did.

Kyle Chan: Just to jump in with some broader context here: with China’s effort to develop its own telecom equipment industry and its broader tech industry, there’s this really interesting connection between the role of the state and the role of these different businesses. At various times you have other companies, other state-owned enterprises like Great Dragon that were the real national champions originally. There’s this idea that some companies were seeking to form connections with the state, while others were actively presented as the spearhead for China’s technological and economic development versus Huawei at this time.

Now we know Huawei as playing that role much more prominently. But back in those early days, Huawei was one of many competitors and was actually the underdog. To build on what Jordan was saying, Eva, you had a great passage about how Ren Zhengfei was even worried about taking loans from state banks, wary about becoming too entwined with state interests and all the strings that are attached. Could you elaborate on some of those early years and eventually reaching that status that we now associate with Huawei of dominating the industry? Really, in those early days, that was a story that was still unfolding.

Eva Dou: Especially in those early days, China’s state sector — these state-owned enterprises — had a reputation of being very sluggish and uncompetitive, basically a place where people knew they would stay employed whether they worked or not. This was something that Ren Zhengfei knew meant the death of his company if they became like that. He was quite wary of being too intertwined with the state, especially in Shenzhen, which is where this capitalist experiment was beginning in China.

They saw themselves as different, as forging a new kind of business system than the rest of China. They would call the rest of China “the inland.” Huawei was very aggressive, very vicious in how they went about doing business and also in their expectations for their own employees. They were notorious for firing people summarily if they didn’t meet expectations.

For a while they were doing joint ventures with state-owned companies across provinces in China to try to expand their footprint. That was a big culture shock for the state-owned employees. You have these stories that they were used to taking their daily naps during the day when they got sleepy. Then the Huawei management would come in and clean house, and things would be totally different.

Jordan Schneider: I love the part where, when they start winning, the state-owned company is like, “Hey, come on, this is a private company. Why are we letting them win these contracts?” Can you tell that story? I really liked it.

Eva Dou: It was one of the local officials in Shenzhen who had written a memoir after he retired with some of his recollections of dealing with Huawei and these other companies. He talked about how the state-owned companies after a while started coming to him and complaining that he couldn’t just be supporting Huawei, that he should be supporting them too. He put them in their place and said, “Well, you guys have been getting these subsidies and all these benefits for so long. If you can’t compete with Huawei, you have to look at yourself and see why that is."

Jordan Schneider: They sell cheaply to get market share. Someone at Great Dragon complained to Shenzhen’s Science and Technology Bureau that the local government shouldn’t just support privately owned enterprises. I mean, sorry man, sucks to suck. You could have been Huawei in the 80s, but instead you let Ren do the thing.

We’re in the 90s now. Huawei is a going concern, on the up and up. There is this fascinating blend of Western culture and PLA culture that Ren tries to inject into the lifeblood of how Huawei operates. Can we talk about the different influences that he tries to incorporate and put into his company and how it does business?

Eva Dou: He and his deputy team really felt that they were forging a new kind of business model in China — one that was not the state-owned enterprise and also not purely the Western style of private enterprise. He was seeking wisdom from all different places in the mid-90s and late 90s when he was working on this.

Famously, IBM was the Western company that they hired as a consultant to teach them how to transform from a startup into a multinational company. There were just many things that Ren Zhengfei knew would take them too long to figure out on their own: How do you meet production schedules year after year? How do you plan product launches to remain on schedule? How do you manage your supply chains? They went to IBM and some other Western companies to try to learn this as quickly as possible. At the same time, he was also meeting with leading Chinese entrepreneurs to try to understand what the relationship to the state should be at that time.

Subscribe now

Jordan Schneider: In 2017, I was at PKU in grad school and they have all these recruiting events. Of course I’m going to go to the Huawei one, and they hand out this book that they put out, “The Huawei Way,” where it’s these chapter-long essays of employees reflecting about their experience in the culture. I’m sure you’ve read it.

It’s funny, Eva, because usually when you see business people and they’re defining a new corporate culture, it’s bullshit and they just want to look back on themselves as intellectuals and business innovators. But in all of my reading about Huawei and diving into this book, I mean it is a weird artifact, right? This company and the way it goes about things.

His openness to Western management practices, the eagerness to go overseas and localize while at the same time the wolf warrior, hardcore zaibatsu energy but also up-and-out and “we’re going to fire you if you don’t deliver" — all of that is not something that you see anywhere else and only comes from Ren personally pushing to synthesize his personal experiences and his drive to synthesize a new way of being from a corporate perspective. That’s one of the most fascinating parts of his story and your book in particular.

Eva Dou: In most countries, there are older businesses, businesses that have been around for many years to look back to. In China, just because of the Cultural Revolution, that was all wiped out and there was a sense they’re starting again and have both the privilege and the responsibility to build a new model. Ren has said many times that he wanted to build a company that could last for 100 years. It’s been very much an open question how you do that. It’s only been a few decades since the end of the Cultural Revolution really, so it’s still an ongoing experiment of building a private enterprise in China that can last for a century.

Jordan Schneider: Kyle, do you want to do the East Asia comparison? I think Samsung is kind of the closest analog from a crazy founder, hardcore founder perspective. What other parallels do you see?

Kyle Chan: There are other ones too, like the founder of Honda and also the founder of Toyota, breaking the rules in the early days, not being allowed actually to have a license to produce cars but defying that and then eventually becoming the star player in the industry. You see this again and again, and actually another strong parallel is just the expectations — these sky-high expectations that you get now in Silicon Valley.

I jokingly have talked to people about this book and said, “Wow, Elon Musk would love this approach to just hard charging, ‘Go to Mars’ mentality that Huawei seems to instill in its work culture.” Some of this is nuts and bolts of how you run an organization, but some of this really seems to be, “We can build, we can defy the EUV lithography export controls” — which we’ll get into later. But all those things are just mere speed bumps along the way to these greater global ambitions.

Jordan Schneider: When you read the early Apple stories of just the intensity and the near-death experiences, you get a bit from Steve Jobs at his most unleashed peak. But I really think Samsung is the closest comparison. They have a similar story in their arc to the one that you wrote about, Eva, where everyone signs their resignations and says, “I will sacrifice.” The importance to the country arc that ends up developing is not quite as present in a Tesla or a Honda as it is in a Samsung, which modernized an entire country for better or for worse.

Eva Dou: That’s interesting, the comparison to Samsung. Huawei’s mission has shifted over the years in that in its early days it was not trying to be a company like Samsung. In fact, Ren was very adamant that they were going to be very narrow in what they did — that they were just going to do telecommunications equipment. They didn’t want to be this sprawling multi-armed conglomerate because they felt they would end up not being good at anything.

It’s been a process where they’ve gradually expanded. During the 2000s they began making mobile phones and then smartphones. That was something that generated a lot of debate inside the company. Ren at the beginning was against it. He thought they were diversifying too much and was eventually won over by some of his deputies that this was the way to go. More recently, because of sanctions, they’ve been forced to make more of the nuts and bolts themselves, especially chips.

By how this worked out, they’re now a much more direct comparison to a company like Samsung — they have so many different lines of business. What’s interesting is they didn’t start out that way, and for many years that’s not what they wanted to become.

Jordan Schneider: It’s interesting also because this happens over the course of Ren getting older and older. He has this great line where you said he suspected that smartphones were overhyped. “We believe that the Internet has not changed the essence of things,” he told staff. “A car must be a car and tofu must be tofu."

But if he succeeded, which he did, in hiring all of these hard chargers who are going to want to spin up new industry verticals or whatever — just having them all fight for the same three slots running different verticals in telecom — is actually not sustainable. One of the remarkable things is that he was not stubborn enough to shut down these new business lines. His commitment to investing — what’s the number? Like 30% of annual revenue into R&D — some crazy stat which basically no other Chinese company did in the 90s and 2000s and even into the 2010s — is something where if you’re going to end up doing all that, then yeah, you’re going to end up wanting to branch out into new businesses and grow in different directions.

Kyle Chan: Speaking of branching out to new businesses, I was wondering if you could say more about this moment in Huawei’s rise where they’re getting into a whole bunch of different areas. While they began with switches and telecom equipment, they really started to expand into everything from undersea cables, and then eventually EVs, AI, and semiconductors. Actually, your book points out that Huawei was working on semiconductors from a very early stage and was quite important in China’s broader push into semiconductors. Could you say more about that period and what drove some of that expansion? What was that like for a company that began with such humble beginnings? This is part of the arc leading up to a $100 billion per year revenue business.

Eva Dou: One way to look at it is that their customers — their international customers — are often buying all these different things. From that sense, it makes a certain logic. Smartphones and telecom gear seem so different — one’s very specialized, one is for the consumer — but it actually ended up being a savvy way for them to get their foot in the door into markets where it was quite difficult to sell their telecom gear, such as Europe, because smartphones were seen as much less sensitive technology.

In countries where they weren’t able to sell telecom gear directly at first, they were able to sell small, cheap smartphones or mobile phones. Actually, the customer is the same — it’s these telecom operators who run these mobile shops, where, as a consumer, you would go in and buy a phone. That allowed Huawei to develop these relationships with these companies. In some cases, once they were selling phones, they were eventually able to sell the telecom gear.

Chips are a little different. It started out partly for cost reasons. For some cheap chips, they found they were able to save costs significantly if they were able to produce them themselves instead of relying on foreign suppliers. There’s probably a security component to that too. He said from an early time he’s pointed out the national security implications of telecom gear, and it probably helped assure Chinese officials if they were making some of their own chips. In fact, their early chips were often used for surveillance cameras. That was one of the early lines of business for their chips.

They’ve gradually found that one line of business can help the other line of business. They’ve expanded probably beyond what Ren would have been comfortable with back in the 1990s.

Spare Tires and War Zones

Jordan Schneider: I love this line from the head of HiSilicon in 2019, which is Huawei’s chip development arm, who said, “For years, the unit had worked on spare tires 备胎. But now, today is the decision of history. After Trump’s export controls on the company overnight, all the spare tires we built have become the main tires. Our years of blood, sweat, and tears have been cashed in overnight to help the company fulfill its commitment to keep serving customers."

It’s very emotional working at this place. On the one hand, you have these stories of people — one of the executives talks about how his annoying family will never forgive him for missing the birth of his two children. But on the other hand, it’s clear that for these employees this is not just a paycheck. There is a lot of life fulfillment and meaning that these folks derive from this work. Maybe I was a little too dismissive earlier about them laying it on too thick because, look, we’re in a country where religion isn’t really a thing and you’ve got to find it from somewhere to get you motivated enough to work 80, 100-hour weeks past the point where you’ve already met your creature comforts.

Eva, what are your thoughts on Huawei as a spiritual totem for its founder and employees?

Eva Dou: Definitely, part of Huawei’s success has been very much its founder’s charisma and his ability to motivate his employees to work incredibly hard and sometimes take enormous personal risks in working for this company. There is a spiritual component to it where he is telling them they can find meaning in helping their country become technologically sufficient in key technologies. The United States has been this bugaboo throughout its history that’s held up as a country that is hostile to China’s development, that wants to hold the country back and that they at Huawei would have to surmount.

Through the years some of that has sounded melodramatic and overblown. When the sanctions came down during the first Trump presidency, that made people believe that Ren wasn’t just overly scared, that he had indeed been preparing for what was going to happen.

Jordan Schneider: Another Elon parallel, right. The American company that gives the most spiritual energy to its employees to work really, really hard is SpaceX, with this incredible vision of making us an interplanetary species.

My favorite illustration of the truly devoted service to customers — not just the Chinese government — is during the Libya civil war. Literally everyone and their mother was pulling out of the country. The Chinese diplomats pulled out. Same with Bahrain when they were going through their civil disturbances. But Ren Zhengfei and Huawei were like, “No, we’re going to keep our engineers here. We signed some contracts. We’ll just have two businesses, one for the government, one for the rebels."

You have this great interview with the wife of one of these engineers saying, “Yeah, every time we call him, we hear gunshots in the background.” Libya is not this huge growth market, but it is a fascinating illustration of just how committed this company is to expanding and succeeding.

Kyle Chan: This is a really interesting segue because so much of this book is looking at Huawei’s trajectory over time in parallel with China’s. Some of these twists and turns and even geopolitical strategy at the company level mirrors that of China’s approach. Huawei, especially in the early days when it was really trying to establish itself internationally, was entering these markets that were a bit too risky or too volatile perhaps for some of the more established players. It gave Huawei an opportunity to start working outside China.

It parallels China’s efforts to develop relations with different parts of the Global South — Southeast Asia, Africa, Latin America, Middle East — while gradually seeing ties with the West as being a longer-term project, but one that would be much harder to tap into early on.

Could you talk more about this opportunistic phase in Huawei’s trajectory in terms of its global expansion? Later on, when you see one of these maps of where Huawei ended up, it’s all red Huawei or ZTE 5G equipment all over the world. That’s the later endpoint, but along the way, what was Ren Zhengfei trying to do and how were they trying to make inroads piece by piece into the international market, culminating with the British Telecom contract, which gave them that kind of legitimacy that parallels China’s quest for legitimacy on the international stage?

Eva Dou: Certainly. Huawei’s international expansion can very much be seen as a proxy of China’s broader international relations effort over the decades. The company was founded in 1987, and at the beginning they only focused on the domestic market. Then in the mid-1990s, they started looking overseas.

At the very start, Russia was one of the first countries they looked at. That was partly through diplomatic help — the Chinese Foreign Ministry helped them go over for a trade show and helped smooth the way to their first very small contract there. At the same time, they were looking to countries in the Middle East and Africa in the early days.

There is some history of their eventual run-in with sanctions issues with the United States, which is that some of their earliest customers were countries where the major Western companies were a little more reluctant to go for whatever reason. That included Iraq, Iran, Afghanistan, North Korea.

Then you get to the point where China is entering the World Trade Organization in the early 2000s. Huawei jumped on this trend of internationalizing. That was when you saw them start to push into developed markets, into Western markets more aggressively. Of course, it took many years for them to be established in Europe. But if you look today, there’s hardly any country in the world that doesn’t use Huawei telecommunications equipment to a greater or lesser extent.

Kyle Chan: In parallel with this rise, what’s really interesting is that as Huawei is expanding globally and as Huawei’s star is rising in China as a symbol, you have growing suspicions from the West, especially from the US, about what Huawei is doing and what its real relationship with the Chinese government is. There are concerns about surveillance technology. You have the rollout of smart cities and safe cities programs from Huawei. This ties into the very markets that Huawei is deciding to enter.

In the early years, as you mentioned, Iran and Iraq were some of the earliest cases of Huawei coming on the map in terms of US concerns about security. How did you see this reaction to Huawei’s expansion, especially from the US? You have congressional hearings later on, which is a high point of the book, a point of high drama, where Huawei and ZTE executives are brought in front of Congress to testify and answer basically a pretty tough interrogation session. Something similar happens with British Parliament to a certain extent.

There are growing suspicions rising almost right alongside Huawei’s growing prominence on the world stage. Could you describe some of that and where it’s coming from? Obviously Huawei’s own activities within China contribute to this. It’s not just what Huawei’s doing abroad, but later on, you were one of the earliest reporters on use of Huawei surveillance equipment in Xinjiang, for example. Could you tie all those different pieces together?

Eva Dou: The first incident where Huawei really got on the map for the US government was in Iraq in the early 2000s. Huawei was helping the local government and military build a fiber optic telecommunications network. If you think about it, these older systems where they’re using radio communications — that’s something that is more easily surveyed or hacked. That’s what the US government had been doing. They’ve been using it to keep track of what Iraq’s military was doing. Suddenly those conversations were going underground. Huawei was helping them build these fiber optic lines that ran underground and that were much more difficult to tap.

That was when George W. Bush gave the order to bomb those installations, which would force Iraq to use radio again. This was later explained by some US officials. That was something Huawei was doing for governments around the world, helping them build these more secure communications networks that were harder to survey for Western governments. That was how they came into conflict with the US at the start as a national security concern for the United States.

Later on this became other things. You mentioned Xinjiang and surveillance. Their surveillance systems became a significant line of business for them, both domestically and in countries around the world. These are the modern surveillance systems for an entire city. There are video cameras, but more importantly, there’s the software on the back — there’s facial recognition, there are AI algorithms to help track trends.

That is a more recent iteration, and what sparked the sanctions was the 5G generation of networks being laid out around the world and Huawei seeming poised to win an enormous chunk of those orders. That set off alarm bells in Washington at the time.

Kyle Chan: This gets to a whole question about the role of technology today and to what extent can we separate some of these core functions of what telecom equipment is supposed to do versus issues of data privacy and national security. I wanted to read a quote from the head of global cybersecurity for Huawei at a UK parliamentary hearing. He’s asked about whether Huawei’s telecom equipment could be used for surveillance and data collection and sending that back to China. His answer is: “It therefore does know where you are because it knows where the information is coming from. In that context, telecommunications networks from all vendors know where you are so as to connect you to those networks. Huawei’s equipment is no different from anyone else’s equipment."

This was just a question about whether Huawei’s telecom equipment can track you. His answer seemed to be, “Well, in order for a cell tower to work, that’s really what it does. It has to triangulate your position and make sure you receive and get the signal."

How much of this seems to become only a bigger issue over time? Now we have a whole bunch of questions about not just telecom equipment, but this has entered into many different domains. Everything from TikTok to connected vehicles, DJI drones. The latest one was TP-Link routers. It’s this growing concern from countries like the US about what will happen with data that’s passed through or collected through these systems.

Is there anything that Huawei could do to reassure the outside world, especially in the West? There was a really great example where Ren Zhengfei or a Huawei executive offered to share source code or do a tech licensing deal to just put it out there and say you can look through our source code and see what’s going on, if that’s any kind of reassurance. There are different ways of trying to deal with this issue. But to what extent is it just something like a Gordian knot that can’t be untied?

Eva Dou: The comparison to how US and UK policymakers have looked at this issue is pretty interesting, because until recently, they’ve taken very different approaches. The UK has taken the approach that these security risks can be mitigated, and we are comfortable with it. The US has taken the approach that this is an intractable problem — these security issues cannot be mitigated.

Cybersecurity experts in both countries receive very similar training. Ultimately, it is kind of a political question and a political answer of what kind of relationship you’re willing to have with China, what level of “risk” you’re willing to take on, and to what degree do you see China’s government as an enemy that needs to be blocked by all means necessary?

Until recently, the UK approach had been that they have this center where Huawei has its source code for its products and where UK officials, including UK intel officers, can go take a look at the back end and comb through it and resolve any cybersecurity concerns that they feel they have with it, and they were okay with it. Actually, today, a number of other countries do use this model, and they feel that mitigates whatever hacking risks that they feel there are.

Subscribe now

The US has been on the extreme end of the spectrum compared to other countries in feeling that these are unmitigatable risks that they cannot — that we just can’t use this equipment. That speaks partly to just how the US and Chinese governments see each other, that they see each other as direct rivals, whereas for other countries, it’s not the case.

Jordan Schneider: It’s funny, because there is kind of this spectrum, right, where you have these African countries and Malaysia being like, “Whatever, spy on me. What do I care? It’s not like I could stop them anyway. Might as well save the money with the Huawei kit.” Then we have the UK kind of in between. You’ve got this great back and forth between John Bolton and a senior UK security official saying, “What I don’t understand is why out of all the things — chips, AI, rare earth minerals, whatever it is — your administration has decided that a modest amount of base stations on hilltops in England is the epicenter of your new declared war on China. Why?"

"You got to pick something,” John Bolton replied, but that feels like a facetious response. The deeper one, which is the one you alluded to earlier, Eva, is that losing this entire industry of telecom broadly and base stations in particular to a Chinese player is something that the US couldn’t really countenance as the country that feels like it has more to lose from being spied on by China than even folks who maybe aren’t all the way at the side of Malaysia, but the UK that feels like they can kind of deal with it. Even if they can’t, how bad is it going to be? It’s not like they’re the ones upholding the industrial base of the liberal democratic order or what have you. Kyle?

Kyle Chan: It does tie into this question from the US side too of what are the motivations for each of these things? I see a lot of parallels for the EV industry. There’s a lot of debate about to what extent were the Biden era tariffs on Chinese EVs and then later on the connected vehicles ban motivated by security issues and to what extent was it motivated by economic competition issues about worrying about what will happen to American automakers and US auto jobs along the way.

That just puts the US in a different position than other countries when it comes to its relationship with China and Huawei and a lot of these other rising Chinese tech firms that seem to be now entering into spaces that US companies were comfortable being dominant in for a long time. There’s also that parallel with Europe too, taking a different approach, again with EVs. They have similar concerns, but there seems to be at least a greater interest in investment and production of Chinese EVs within Europe. All of that is just to say it’s a lot messier than merely “let’s block everything Huawei” or whatever, “anything goes.” There’s a full range of different approaches out there across the world.

Jordan Schneider: I want to talk a little bit about the organizational, senior management structure and how it pattern matches to the way the Communist Party structures itself. Can you talk a little about that, Eva?

Eva Dou: Huawei has for years adopted this kind of collective leadership model, which in many ways is similar to how China’s government is run. There’s a senior group of officials, both explicit and also unofficial, semi-retired officials who all have a role in deciding the direction of this company. On purpose, it is a little vague exactly where those lines are drawn of who’s in charge of what.

For now, Ren Zhengfei is still the top guy. He’s talked for years that he’s going to be retiring one day and now he’s in his 80s, and it’s still unclear when he’s going to be retiring. That’s going to be the big test for Huawei, of course, for any company — that initial, the first handover from the founder to the next generation of leadership and if they can keep the company running to the same degree of success.

Jordan Schneider: I want to close on the human arc a little bit. A lot of Western media makes fun of the Huawei campus. The fact that it’s kind of this Disneyland — it has a Versailles, it has a Kremlin, it has all these international styles. This moment that you write about, about him moving villages in middle school, you write that he was astonished when they moved to the county seat and saw a department store. It was the first time he saw a two-story building.

Then, a few decades later, he goes to the US for the first time, sees Las Vegas and thinks it’s the most beautiful city he’s ever seen. These giant buildings and the pyramids and whatnot. Then he builds this incredible company and decides that he wants to replicate all of these architectural wonders from the West. Not really from China, but from the West.

Just that arc of this boy whose dad was principal and was beaten to within an inch of committing suicide during the Cultural Revolution to go from having this really complicated, admiring, but also rivalrous relationship with the rest of the world and looking up to so much that the West has brought, but also really wanting to be able to compete and strive to stand at the same level, if not necessarily playing by all the rules that the Western countries and companies would want them to — that’s just a remarkable arc. That is probably the thing that’s going to stick with me most from this book. When Ren, who’s pretty old now, does end up passing from the scene.

Kyle Chan: This image of him from his early childhood days, really the rural boy from out in the provinces to then later on being able to be this savvy political operator at the national and international stage and basically getting on side with multiple Chinese leaders over time. You write about how he is going out doing deals with, from Jiang Zemin, Hu Jintao, and now Xi Jinping.

We recently had that big meeting with the tech executives, Chinese tech executives meeting with Xi Jinping. Who was there, sitting in the very front and center, was Ren Zhengfei himself. Who could have imagined, probably not him from his early days, that he would be at the center of so much of the Chinese leadership’s ambitions, US-China competition tensions, and part of this story of China’s rise over time.

Eva Dou: It is kind of staggering, the distance he’s traveled since his childhood to where he is today, both personally, his company, and also China, that China is now a true technological competitor on the world stage. That’s part of what interested me in this project — just understanding how much has changed and how it changed.

As far as the palaces that you alluded to that Huawei has built on campus, Ren has said — it’s not only the Western media that critiques it, sometimes his own employees do. He’s talked about this and said, “You guys can gripe about these palaces, but we build them for the customers."

It’s not just him himself. This has been China’s experience, broadly, this transformation. When they bring government officials and telecom officials from across the nation to the campus, this is the same transformation that they are experiencing themselves. It is appealing to a certain part of his customer base, all the glitz and glamour. Most of his international base is the developing world. Huawei is famous for bringing foreign officials on these junkets and wowing them with both the campus and the food and the hospitality.

Jordan Schneider: Eva, are there any stories you want to share from the reporting of this book? In particular, I’d be curious about, in reading a lot of these memoirs and Chinese coverage of the company, what are the narratives that get a lot of play in mainland China that you don’t end up seeing written up much in the Washington Post?

Eva Dou: The most fun part of this was doing archival research for me and trying to find sources that even in China are a bit obscure. You do have the easy narratives, both the international ones about Huawei being a national security threat and within China, Huawei being this great company that’s getting better and better. Part of what I try to bring to this is the smaller voices.

I really enjoyed reading some local government officials’ memoirs in China. As Huawei has interacted with so many officials both within China and around the world over the years, I wanted to bring some of their recollections of the experience of meeting with Huawei, both good and bad. In China, as you know, there’s pretty strict censorship. In memoirs, when people are late in their life and no longer worried about keeping their job, they are a little franker sometimes in their recollections. That’s been an enjoyable part of this process.

Kyle Chan: I was astonished looking through the sources that you cite and the amount of material that you had to go through. I’m sure the sources that you cited are just the tip of the iceberg for everything that you were wading through and trying to sift through. You even chased down a Harvard Business School classmate of a Huawei executive and interviewed them. I was astonished by the lengths you went to really get to all the details of this story, including a ton that just has not been talked about before.

Subscribe now

Eva Dou: Thanks so much. The process of digitization of archival material around the world was really helpful in this project. So many times during the process, I thought about how if nothing was digitized and I had to go in person and just read each page one by one, how long would this project take? It would have taken years and years more than it did. It really gave me an appreciation of the earlier generation of researchers who were doing everything in analog, in person, archival research.

Jordan Schneider: But they didn’t have Twitter to distract them. Eva, how long do you think you’d last at Huawei?

Eva Dou: Oh my goodness.

Jordan Schneider: Say your 24-year-old self. We’ll give you a little boost of energy.

Eva Dou: Well then good. I probably could hack it a couple years, but they are ferocious in weeding people out.

Jordan Schneider: They’d catch on to you eventually.

Eva Dou: Yeah, probably.

Jordan Schneider: Is there a moment or a meeting that you wish you could have been a fly on the wall for?

Eva Dou: There are a few points in their early history where they changed the company’s structure or ownership and I’ve always been pretty curious about what exactly those conversations were. They started out as a private startup — it was a pilot program for private startups. No one knew how long that would last. At some point they shifted to where they were almost more state-owned for a few years and then they went back to being a private corporation.

Kyle Chan: One funny thing that stuck out to me is that Eva, you have pointed out the pronunciation of Huawei among Americans. So many people say “Wawei” instead of “Huawei.” Once you pointed that out, I can’t stop hearing that. People do it all the time and I don’t know where it came from.

Eva Dou: Part of it is the company itself. At some point their PR team was going around teaching people to say the company name this way.

Kyle Chan: That’s so interesting.

Eva Dou: Which is a curious thing.

Kyle Chan: Wow. It reminds me of Hyundai trying to be like, “We’re not Honda, we’re not ’Hi-yun-dai,’ it rhymes with Sunday.” It’s tough.

Eva Dou: They thought about changing the name in the early days when they realized it was so hard to pronounce for international audiences and they batted a few different things around. But in the end it seemed too difficult to change, and we’re stuck with Huawei.

Jordan Schneider: One of the daughters is a singer, right? We have to do that as our outro music. But are there — did Ren ever talk about liking music? Did he have any favorite songs or genres or anything?

Eva Dou: Interesting. I don’t know off the top of my head about music, but he’s always been an admirer of the arts and that comes from his parents and this desire to be a cultured renaissance man. That’s something he’s really encouraged with his employees — for them to spend their spare time cultivating themselves, listening to different types of music, looking at fine art, things like that.

Jordan Schneider: Cool. Aside from the architecture thing, which people are clear on, are there painters or writers? Give me a sense of the Ren Zhengfei cultural constellation.

Eva Dou: Well, classical style art is something he quite appreciates. There are stories of him collecting paintings from around the world and having his employees help him bring oil paintings over from abroad. He’s talked about his family quite liking Europe, liking to go to Europe on vacations. His youngest daughter grew up largely between the UK and China. He often talks about these cultural things to make a point, as part of his outreach to different markets. He talks about international culture that he likes probably more often than domestic.

Renaissance Philanthropy — in my opinion, the most exciting philanthropic venture in the US — is getting a one-year check-in. Kumar Garg first appeared on the show right before I went on paternity leave, and now we’re back for round two. Before founding Renaissance Philanthropy, Kumar worked in the Obama Office of Science and Technology Policy and spent time at Schmidt Futures.

We discuss…

• How Renaissance catalyzed over $200 million in philanthropic funding in its first year,

• The goals of the organization and how it has responded to Trump’s S&T funding cuts,

• What sets Renaissance apart from traditional philanthropic organizations, and lessons for China-focused research foundations,

• AI applications in education, from tutoring to dyslexia screening,

• Donor psychology, “portfolio regret,” and how to build trust within a philanthropic network.

Listen on Spotify, iTunes, or your favorite podcast app.

The Hedge Fund Model of Giving

Kumar Garg: I like that this is becoming an annual tradition.

Jordan Schneider: Yeah, we've got to set goals this year, and we can hold you to them in 2026.

Can you start off with the 101 of Renaissance philanthropy, and explain how the thesis has played out over the past year?

Kumar Garg: I’m grading myself here, so this is a biased view, but it’s been a very strong year. When we were launching the organization, we were trying to do something different.

Most philanthropic organizations exist in a single model — they work for a single donor. That donor has resources, whether they sit in a foundation, in their DAF, or as personal wealth. The organization works for them, asking how much money they want to give and on what topics, then runs their philanthropic giving.

There’s another class of organizations that are basically the people spending the money — researchers running labs and doing high-quality research. The philanthropic system has mostly operated with givers and takers — folks operating these organizations and folks doing high-quality work.

The idea behind Renaissance Philanthropy was to sit in the middle and style ourselves more like an investment fund — more like what happens in the world of finance. The folks who are the holders of capital, who have the money, mostly don’t spend their time trying to directly deploy that money.

If you work as an LP for a family office, you might have a team of 10, 20, or 30 people, and you’ve got billions of dollars to deploy. What do you do? You go out there and find intermediaries — private equity funds, hedge funds, venture capital funds, or other experts in particular sectors and areas. You give them the money, and they deploy it on your behalf to help you earn a return.

Philanthropy has mostly operated differently. It’s odd, but it’s historically contingent. The investment world moved toward specialization from the ’70s onward, while philanthropy went in the direction of direct giving. You have really large philanthropic organizations, often well-staffed by experts, that do the giving.

The challenge is that there’s a subset of donors who want to build large organizations, and there’s a large set of donors who don’t. The ones who don’t have been sitting on the sidelines. What ends up happening is maybe when they retire, they build an organization, or when they die, they bequest it to a nonprofit or university. That leaves a lot of value on the table.

The idea of Renaissance was, on various science and tech topics, can we do what an investment fund does? We write down a thesis for three years, five years — we want to achieve this goal. We recruit a field leader to run that fund, then treat the donors almost like LPs in a philanthropic fund. We’re not giving them a return back, but they’re putting money to work against that strategy.

A year ago, when I told this story to people — “I’m going to create an organization that does this” — the operative advice was, “Good luck.” You’re going to cover the waterfront across AI, climate, and economic social mobility. You’re going to take on this massive fundraising goal. That seems like a very hard way to operate. You have no natural advantages — you’re not spending one person’s resources. You have to raise the money and deploy it. It seems doubly hard.

What I was interested in was growing the pie — can we use this model to bring new donors in?

A year in, the early grade is strong. We’ve been able to stand up multiple philanthropic funds. We have a fund using AI to accelerate the pace of math research. We have another fund using AI to deliver public benefits better. We recently launched work on climate emergencies — can we solve for runaway climate risks and increase the technology readiness level of various climate technologies?

We have different funds in various areas. Each has this basic structure — they have a thesis they’re driving against, a field leader running it, and we’re recruiting donor money against that strategy.

What I’m hoping for is that this starts to become — not the only way philanthropic giving happens — but a much more credible path. This allows more donors to be active without necessarily having to take on all the operational load themselves.

Jordan Schneider: You’ve launched this in a particularly precarious time for the future of science and research in America. We’ll get to your takes on the policies in a second. But I’m curious from a donor appetite perspective — what has all the tumult in universities and government funding done for those billionaires sitting on the sidelines, giving just 1.2% of their assets annually to philanthropy?

Kumar Garg: That’s a great question. I don’t have one system-level answer — it’s a frequent question I get about how donors are interacting with the environment. They’re interacting the way most people are — there’s an incredible amount of chaos and news every day, leaving many frozen in place.

It’s relative. Government has pulled back on research funding in the short term, causing significant churn. Industry is also holding back as companies figure out what’s happening with tariffs and everything else. Philanthropy, comparatively, is cross-pressured but hasn’t engaged in the same pullback.

There are donors we interact with who are certainly reformulating their strategies. There are others who, as I mentioned, are interested in compelling ideas and looking for those just like anything else. I haven’t seen an overall pullback — just more of a sense of “Is this idea good in itself, even if government didn’t help at all?"

Jordan Schneider: Can you put in order of magnitude the hope of the new model you’re trying to manifest against, I don’t know, NIH budgets being cut by a third?

Kumar Garg: There’s no world in which philanthropy fills the gap. If you step back and ask how the US built its lead — well, the US spends on the order of $200 billion a year on R&D. Once you include basic and applied research across DoD and civilian agencies, that’s an order of magnitude more than philanthropy spends on research.

The place where these new models will get traction is that how you organize scientific organizations has suddenly become much more of a jump ball. It used to be that the academic bundle — being at a top university — had everything stacked on top of itself. You could get really good talent that way — graduate and post-graduate talent, great students. You build your lab there, do cutting-edge work. Usually, the university gives you flexibility to do many things on top of it. If you’re an academic doing well at cutting-edge research, you could do that within the four walls of a university.

Some researchers have left universities and built what are basically academic research labs outside the university. You’ve got the work that Patrick Collison is supporting around the ARC Institute, the Flatiron Institute that Simons supports. You’ve seen the FROs that Convergent Research proposes. For a long time, that’s been a very alternative path — rare to do, often requiring you to figure out what happens to your university affiliation and how it changes your career path.

If you’re a researcher who’s ambitious and wants to do big projects, whether you’re doing them within the four corners of the university or in your own nonprofit research lab and partnering with universities becomes more of an open question — especially in a world where university funding might fluctuate based on political developments.

I don’t know how that will play out over time, but we’re three months into a deeper shift in how institutional financing will happen. That could have big implications generally. On net, if the federal government doesn’t play its important role in funding research, it’s all a net negative. If federal funding returns to a healthy level, researchers will still take this as a wake-up call to think about structuring their research organizations to be more resilient against systemic shocks.

This episode is brought to you by ElevenLabs. I’ve been on the hunt for years for the perfect reader app that puts AI audio at the center of its design. Over the past few months, the ElevenReader app has earned a spot on my iPhone's home screen and now gets about 30 minutes of use every day. I plow through articles using Eleven Reader’s beautiful voices and love having Richard Feynman read me AI news stories — as well as, you know, Matilda every once in a while, too.

I’m also a power user of its bookmark feature, which the ElevenReader team added after I requested it on Twitter. ChinaTalk’s newsletter content even comes preloaded in the feed.

Check out the ElevenReader app if you’re looking for the best mobile reader on the market. Oh, and by the way — if you ever need to transcribe anything, ElevenLabs’ Scribe model has transformed our workflow for getting transcripts out to you on the newsletter. It’s crossed the threshold from “95% good” to “99.5% amazing,” saving our production team hours every week. Check it out the next time you need something transcribed.

Jordan Schneider: I observed the EAs being very excited about how many lives they save based on the bed nets they bought. Then you net that out against USAID no longer existing and all the human suffering that’s going to come out of that. The correct calculation may have been to spend all your money lobbying Congress to get people to focus on this.

I think both of us are pretty aligned — we’ve done other shows on immigration policy, university funding, and what’s happening to NSF and NIH budgets going forward. But why do Renaissance philanthropy when Kumar could be spending 100% of his time in D.C. banging on doors and trying to make it 5% more likely that we get an extra $10 billion a year for this stuff?

Kumar Garg: That’s a great question. Being policy-adjacent is generally very high ROI. No matter how you run the numbers, policy advocacy — especially on science and tech topics — punches above its weight, regardless of what you’re doing. It’s probably why I spent time in government as a policy staffer. It’s partly why, no matter what I’m doing, I’m constantly interacting with policymakers and making the case. It’s also why, when funders ask to what extent advocating on behalf of the research community should be part of their work, I’m strongly supportive.

The reason we structured Renaissance this way is that I wanted to specifically think about growing the pie of philanthropic funding because I thought no one was doing it. There are organizations working on policy advocacy. Very few organizations were trying to bring new donors into the mix.

We would be failing as an organization if we weren’t constantly thinking about how our work could impact shaping the debate on the future of R&D funding. We try to be in conversations with both Congress and the administration, as well as policymakers up and down the ladder, to say, “Here’s why this work matters, here’s why the future matters.”

Part of the new models we’re funding — whether it’s things like FROs or AI accelerating science — is to make the case for why investment should happen. Many of the ideas I’ve funded over the years, you can see echoes of in the new Heinrich legislation around accelerating science through AI, where they’re talking about ensuring these AI investments can actually accelerate the pace of science using new models.

Philanthropy, when done well, opens the aperture for what funding could do. Hopefully we’re playing that role. One area I’d like the conversation to reach is moving beyond this dialectic between “science is important” and “science needs to be dismantled because it made mistakes.” I’d like us to reach a place where we recognize there are important things we can do to help reform how we do science. We should bring more discipline to trying out new ideas, bringing in new funding methods and new voices, and reflecting on past mistakes — while also remembering that the investment agenda around science is critical for its utility. Hopefully we can be part of that dialogue.

In some ways, you’re pushing on something I think about all the time — I am a policymaker at heart. The deep utility of that shouldn’t be forgotten in my story.

Jordan Schneider: All right, I think the answer I would give to you is this is the federalist model of policymaking almost — as you said, the inventiveness that you guys can come up with from a form factor and discipline perspective when it comes to doing science and technology research is the type that’s weird enough that it’s not happening in government. But also, two, three, four years down the line, once you guys have some really awesome case studies, these are the sorts of things that can then get 10x or 100x in our gorgeous NSF circa 2027 that has been remade to fully align with the Kumar vision of how change gets made.

With that stance of optimism, let’s talk a little more in detail about some of the projects you guys have stood up. Take us on a little tour, Kumar. Where do you want to start?

Kumar Garg: I’ll go through a couple of the funds and projects we’ve launched. Just to give people a taste of how the model works, let’s start with our work in AI. Our operating theory in AI is that we’re living through a period of huge capability overhang. The idea is that the core technology is rapidly developing, but the number of people, projects, and overall work that actually applies these tools toward actual hard problems in society is really small.

I’ll give an example. We have an AI and education fund specifically focused on how AI can accelerate the pace of learning outcomes. If you follow social media and others, there are many people who write and talk about AI in education. It would give you the sense that a lot of people are working on AI in education. But if you actually dig into the space, the number of actual technical experts who have knowledge of both how education works and how AI works is still shockingly small.

We run something called the Learning Engineering Tools competition — an annual competition that invites tool developers to present cutting-edge ideas that use AI to actually advance learning outcomes. We’ve been running this competition for a couple years. I started it even before Renaissance and then brought it into Renaissance. That competition is the only large-scale ed tech competition in the world. It still blows my mind. No one is out there in a systematic way asking for sets of ideas from people who want to build AI for education.

We have another part of our AI education portfolio that specifically thinks about moonshots — what’s a really hard problem in education that AI could solve? We picked middle school math. It’s really important for advancing to future degrees, and students really struggle with it. We said, can you actually emulate the results of high-dosage tutoring, which the number of studies that J-PAL and others have done show can really double the rate of learning for students in math? Can you do it under $1,000 per kid — bring it under what would make it such that you could offer it to every kid?

We have that running as a program. We have seven teams in the program. We have two teams that are actually on track to potentially accomplish this goal.

Jordan Schneider: Which is wild, right?

Kumar Garg: When those teams are working on it and we ask them who they’re collecting lessons from, there’s not a big field they can go out to. When they go out and interview the AI labs — the ones that get written about every day — those AI labs talk about education, but they don’t have in-house education teams that can actually help these teams.

The biggest piece I would always say to people is that at the coal face, there’s tons of room to do work because when you actually start to work on it, you realize that the number of people who are actually working on it is shockingly small.

We’re now starting to explore our next moonshot area — should there be something that basically looks like the intersection of AI and early learning? Can we actually build a universal screener to best guess if a child is off track when it comes to early language development by having them speak into a device? There’s a bunch of interesting work happening in this area, but we don’t actually have a way to diagnose early learning challenges like dyslexia just by having a student speak into a device. It could vastly increase our ability to help them get to a speech pathologist, get back on track, and be reading by third grade, which is critical to future reading and learning.

That’s just one track — AI and education. That’s just one compelling thesis.

Jordan Schneider: Obviously AI is going to matter for education. Hard to find people to argue about that. Talk a little bit about finding the donors and finding the teams. What was the work that you guys had to do to make and launch this work?

Kumar Garg: What’s been interesting is it has been hard work for us to build out the team because the number of technical experts who actually know both things — AI and education — is small. We have slowly built out a team of ML experts who have educational backgrounds, basically. We call it a hub model. We basically have created an engineering hub and we recruit technical experts into it that specifically have this technical background.

I have somebody on my team, Ralph Abboud. He has a machine learning PhD, and he did his thesis on graph theory. He’s not an education expert, but we brought him onto the team. He has been working with a lot of these educational teams that we brought in. What’s interesting is that his ideas on what kind of language models they should be building are really good. It took him some time to level up on the education side, but now he is one of their highest value contributors, even though he sits on our team and he’s contributing there.

There is this transition where you can build up talent that sits across these two areas. But in AI and education we had to mostly build it. It was hard to immediately find directly. Now we have a constellation of these AI and education experts, some of which sit on our direct staff, some of which sit inside these teams that we’re betting on. It’s been great. Now we have a field team that can really go after more problems.

On the donor side, we’ve really lucked out. We found that our core donor for a lot of this work has been the Walton Family Foundation. They have a long history in funding education. What’s been interesting is that they’ve been interested in investing more in what they call their innovation portfolio, but didn’t know how to necessarily bridge that technical divide — if we’re going to do more in this area, who are going to be the technical experts who will actually do it? That had actually kept them more experimental. But their partnership with us has meant that they have become way more ambitious on how much investment they want to make on this technical AI and education lens.

That’s our core thesis — can we be the permission structure for donors to go much bigger on innovation? We’ve seen that in other areas. Slowly their support is causing other donors to come in as well. That’s basically whether you’ve been a long-standing donor but not active on science and tech topics, or you’re an early donor altogether.

Jordan Schneider: What’s the RenPhil management fee?

Kumar Garg: That’s a good question. We build our cost recovery into each fund. Usually the way that works is if we’re operating multiple funds, each fund has money going out the door for actual deployment grants, but then we’re building in our cost for the actual staff operating the funds, whatever services and technical support services we’re providing, the work we’re doing to partner with various funders, as well as our overall studio support. It varies fund to fund, but donors have found it — compared to having to try to do this themselves — much more actionable. For us, we want to build a thriving organization. We don’t want to cut corners. We want to build an organization that can both operate those funds and also be looking for the next ones.

Jordan Schneider: Does anyone complain about that?

Kumar Garg: The way it comes up is there’s a type of donor who actually has the answer in their mind. They’re thinking, “I think this needs to happen.” Really, what they’re looking for is an operating partner to just do that — “I want a conference, I want a workshop, I want to fund these three organizations.” Our model is we’re the product. You’re actually hiring us to go build out the strategy, recruit the team, deploy. If you actually have the answer in your head, we often tell them we’re way too fussy for that model. You should just — there are much simpler ways you can operate. That’s where the delta comes in. If you already have the answer in your head and you’re just looking for a partner to execute for you, we’re probably not the right fit.

Jordan Schneider: Yeah, you said this on another show. You were like, “We are for donors to take off the cognitive load.” The idea of being: yes, if I have $10 billion, maybe I’ll allocate $1 billion investing in stuff I know and think I have some subject matter expertise in, but I still have to put the other $9 billion somewhere — probably not cash. Yes, I am comfortable paying a hedge fund or financial advisor a management fee to do that.

Kumar Garg: A big part of it is opt-in. People don’t know what journey they’re on, but what they worry about is: am I going to feel stuck? A lot of folks end up not getting active philanthropically because the decision feels weighted by getting stuck. “Okay, if I hire somebody, and then six months from now, I decide maybe I want to change direction. Now I’m going to have to let someone go.” People hate that.

Or, “I met a researcher. I liked their research. I gave them one grant. But now they’ve reached out and said, ‘There’s so much happening in the world — I’ve lost funding from the government. Can you double the grant?’ I was just giving them a grant because I met them and I thought they were great. But now they’ve sent me a note that they might have to let go of postdoc students. Now I’m in this uncomfortable situation. If I say no, I feel I’m hurting them. If I say yes…”

People have all these experiences where they feel uncomfortable with the relationship they have around their resourcing. Rather than causing them to work through it, they actually hold back. One of the things we say to them is that our model is one where we’re the ones making the decisions. We’re going out there, finding researchers, finding projects, developing strategies. You can be as involved as you want. You want to be meeting the researchers? That’s great. You want to be learning from the strategy so you can do direct giving down the road? That’s great. But if you also take six months off and decide, “That was great, I learned for a few months, now I’m off doing something else,” — nothing will stop. We’re a fully operational organization that will execute on everything we said we were going to do, whether you’re involved or not.

It just takes the pressure off. You can opt in if you want to learn and be involved, but you can also choose not to. That actually frees them up to want to learn without the “Am I about to get stuck?” That sounds very psychological, but people forget how hard it is to get going on things — “I’m going to start to work out more. I’m going to start to do this.” Starting is hard. We want to make starting easy by saying you can provide a lot of value into the system without necessarily having to own all of that execution.

Subscribe now

Jordan Schneider: There are a lot of pieces of people’s jobs that seem like more and more AI can chip away at or enable or launch or whatever. It’s interesting because some of the things that you guys are doing — you have these seven playbooks, ways you can tackle problems — I would love to upload seven of those to ChatGPT and say, “Here’s my problem in the world,” and the AI can help me pick through which one. But talking someone who’s really rich, who’s feeling uncomfortable about giving money, into starting to donate philanthropically in a serious way for the first time seems like one of the more human things where there’s really going to need to be the friendly Kumar Garg, who now has a nice microphone he can do Zoom calls with to — what did Derek Thompson say? — whisper the dulcet tones of comfort and competence in their ear in order to get them on this path. I don’t know, it just seems like a very human thing that you’re engaging in on the donor engagement side. I’m curious for any reflections you have on that.

Kumar Garg: We are very curious about how much of our own internal processes we can automate. Why not? We sit next to AI, we should be thinking, we should be dogfooding. The place where we’ve seen it already provide some value is just what you would consider baseline automations. There’s a lot of grantee reporting that you should be able to do automations on. We’re definitely interested in: hey, we have a hunch around a thesis in this area. Can you do a research report and tell me what’s the relevant stuff to know? Scoping. We’ve even used it for, “Hey, we might do an RFP on this topic. Who are some researchers who should apply?” Sometimes we found some interesting suggestions for researchers that we should affirmatively reach out to.

I will say that we’re still far away from it actually helping on anything that we would consider high stakes. As you’re saying, a huge amount of what we’re doing is making something that feels like a trust fall. Hey, this is an important decision, but one where people who take their job very seriously and put their own personal legitimacy behind the work is an important part of it. When we screw up, it’s on us. We stand behind all of the work. People appreciate that these are serious people who stand behind the work they’re putting before them. They’re not some faceless intermediary. I don’t think maybe that will change, but that’s an important part. Even on the information you should know about various people and stuff, these current AI models are not that great.

The place where I — we have this intuition that there should be parts of being a program leader that you should be able to have an AI assistant for. Right? You take more and more of the task of being a program leader or fund leader and be able to say, “Okay, I want to do a workshop on this topic. Generate me an agenda for how you would run the day.” It generates based on — it takes a bunch of your past workshop flows and generates a sample workshop design. How much of that can we create so that we really could get to a point where a program leader or fund leader is basically able to operate without that much additional support? Obviously we need to create some cross-cutting support that I’m interested in. But the chance that we’re going to get to an AI advisor — we’ll have to wait.

Jordan Schneider: The trust fall works in multiple directions. You need researchers to give up their PhD programs or leave their current positions to spend half their time with you, while simultaneously needing donors to provide funding. Having a recognizable face with a proven track record and skin in the game on the other side of that equation is something that won’t disappear anytime soon.

Kumar Garg: One thing we debate internally is that much of my workflow relies on tacit knowledge. When I’m talking to somebody about their work, twenty minutes into the conversation, I’ll say, “Tell me more about that. Why is the field stuck on this point?” They start describing it, and I realize that if there were a canonical dataset with specific dimensionality, it might solve the problem. When I ask why that doesn’t exist, they explain it’s locked up somewhere.

Part of me constantly strives to figure out how we can make this process more explicit. When we recruit somebody new to the team, they ask if they can sit in on my calls and watch me work through problems with researchers. There’s something that feels wrong about just saying, “You develop this intuitive feeling for opportunities — just pull on that thread.” The more we can transition from tacit to explicit knowledge, the better. Right now, we operate on an apprenticeship model where people learn by doing and being embedded in these structures, but I don’t think that has to be the endpoint.

Jordan Schneider: Much of what you do involves human matching — putting people in touch with each other. While you could potentially feed all your past calls into an AI system, there’s an emotional and personality matching component that you’re handling. That remains very much a human process that current models aren’t quite ready for yet.

Kumar Garg: The matching capability changes over time, but I think what people really value when I connect them is that I took time out of my day to think the two of them should know each other. That’s the actual signaling value — that my time is precious.

Jordan Schneider: Slight tangent, but if people want to establish trust and rapport, the first thing they should do is spend $150 on a microphone for their Zoom calls. That’s my recommendation for everyone. When I do my calls, I sound the same as I do on my podcast, and people respond positively. You feel like an embodied person rather than a compressed, distant voice through AirPods. It’s advice for anyone who wants to make connections and raise money from billionaires on Zoom.

Kumar Garg: I’ll echo that point, though I haven’t practiced it myself. There’s an old political adage about microphone technology. If you look back at politicians historically, there was a time when microphones couldn’t pick up subtle intonations well — speakers were just projecting loud sound. Once microphones could capture subtle intonations, politicians who excelled at that style of speaking began to dominate.

People point to President Clinton as an example — he was exceptionally skilled at subtle microphone use. I remember reading a paper arguing that this was possible because the technology had improved to support that communication style. Politicians offer good lessons here because ultimately, communication is central to building trust with the electorate. [Here are papers that explore this]

Jordan Schneider: Absolutely. If you listen to old clips of Warren Harding or Teddy Roosevelt speaking, they’re basically screaming into microphones — which was necessary at the time. Teddy Roosevelt was exceptionally good at that style of projection. You needed to be very loud to stand on a soapbox and reach people twenty rows back. Now we have the dulcet tones that modern microphones enable.

Here’s another fun fact, Kumar — the microphone I’m using has been manufactured for sixty years. It’s remarkable that microphone technology for voice pickup has essentially reached its peak — we’ve basically maxed out the capability.

Kumar Garg: I should try to find that paper I mentioned. I wonder if it’s about mobile situations — being in some random union hall where you need to set up a handheld mic in front of a politician. Perhaps that’s why microphone technology improvements became so important. That’s an interesting angle.

Jordan Schneider: I’m curious about that. Alright, shifting topics — Yascha Mounk recently wrote on Substack about attending gatherings, conferences, and dinners where leaders of America’s biggest foundations have been strategizing how to defend democracy. Few were as openly devoted to extreme forms of identitarian ideology as they might have been a few years ago, but the reigning worldview at the top of the philanthropic world assumes little has changed since summer 2020.

The general consensus holds that voters turned to Trump because American democracy failed to deliver for the “historically marginalized,” and the solution supposedly revolves around “mobilizing underrepresented communities.” The most urgent imperative is to “fight for equity” and “listen to the global majority.” I find this perspective fascinating. Kumar, as someone who’s a new entrant to this world, how do you interpret this?

Kumar Garg: Several different dynamics are happening simultaneously. Some philanthropic responses resemble dinner table conversations — people sharing hot takes about why the election unfolded as it did and offering their views on America or the American people. Much of this sounds as random as hosting a dinner party where guests share their political opinions.

There’s also a genuine state of confusion about what’s happening. The first hundred days of the Trump administration have been exceptionally active across a range of unexpected areas. Many people expected it to feel similar to the previous Trump presidency, so they examined their portfolio of issues and anticipated certain outcomes — but that’s not what materialized.

Regarding how much people are actually rethinking their approaches, that’s a valid question. The most immediate reconsideration I’m seeing centers on identifying what we’re missing. This is particularly evident in the science community, which is confronting devastating across-the-board cuts. Researchers are losing funding, university funding is being paused, and graduate students working on topics relevant to competitiveness are having visas revoked.

The community is asking, “We don’t remember this being a major campaign debate topic, so how exactly did we become a political football?” There’s extensive questioning about what we’re missing — whether there was a conversation we weren’t invited to where we were being discussed, and what we’re failing to understand.

While donors with certain political orientations likely won’t change their fundamental positions, the confusion centers less on American domestic politics and more on why certain issues became contentious. Foreign aid is a good example. The extent to which US foreign aid posture and system effectiveness became campaign issues wasn’t apparent during the election cycle. People are asking whether we missed a major debate that suggested the United States should dismantle its leadership on these topics overnight. What policy debate did we miss? That’s where much of the confusion originates — donors being puzzled about the sources of these developments.

Subscribe now

Jordan Schneider: There’s an interesting dichotomy between foundations with living leadership and those with deceased benefactors. Gates recently indicated he would spend down his wealth faster than previously planned, presumably responding to recent events. When foundations have active leadership that engages with current events, they can be more responsive.

However, when you have flagship philanthropists who died seventy-five years ago and organizations that have built programs around worldviews that are no longer relevant or don’t meet current demands, pivoting becomes much more difficult. You encounter institutional blockers, boards, and established structures, whereas a living person with decision-making authority can simply redirect resources.

Kumar Garg: That’s definitely part of it. The piece I’d add, which connects to our Renaissance model, is that people underestimate how much philanthropic organizations become tied to their existing programs. This isn’t necessarily negative, but consider the process: you spend two and a half years scoping a program, conducting field research on topics, then executing a national search for program leadership. You recruit and convince someone to relocate for the position, provide coaching for test grants, and they’re six to twelve months into the grant cycle.

If you then decide the world has changed and want to cut the program — after issuing press releases announcing this as a major new strategic direction — it appears chaotic. People develop what I call “portfolio regret.” If they could start fresh today, they’d create different programs than what currently exists.

One argument we make to donors is structuring themselves more like limited partners, deploying money into funds where all capital remains fresh and available. You avoid the incumbency problem where team members question every pivot attempt because they have specific responsibilities you hired them to fulfill.

Flexibility requires both mindset and structure. Donors sometimes create substantial built-in costs and barriers to pivoting when they could maintain lightweight, flexible operations if they chose to do so.

Jordan Schneider: That’s fascinating because you’d think giving money away would be straightforward — you should be able to distribute funds however you want. But the emotional sunk cost around philanthropy wasn’t something I had necessarily considered.

Kumar Garg: This creates a situation where people spend enormous amounts of time operating like a duck swimming on water — their feet are moving rapidly beneath the surface while they try to keep the strategy looking consistent above water, all while changing the actual content underneath to pivot to current circumstances. This leads to significant conceptual confusion because you claim to have always had this program, but underneath, the program is completely different because the situation changed.

Part of why I favor the philanthropic fund model is transparency — what’s on the cereal box is what you get. It’s a three-year fund with specific objectives that will begin and end. Maybe it’s not perfectly timely, and that’s fine, but your new program can be timely. The alternative — constantly maintaining broad programs that you’re perpetually reworking underneath — makes evaluation nearly impossible. If I ask whether a program has been successful, people respond that the program has been changing constantly. This makes it extremely difficult to evaluate it as a focused initiative that ran for a specific period with defined goals. Did we achieve those goals? People simply don’t engage in these basic evaluations.

For example, I was speaking with a donor and pointed out that in the investment world, people prominently display their successes — putting “first check into [major company]” in their Twitter bios to demonstrate their betting acumen. I asked, “Who are the ten best program officers in America? Who on the philanthropic side has been the most effective check writer?” They responded, “How would we even know that?” Even with qualitative measures, wouldn’t you want to identify the best check writers?

A fund model, even for philanthropic goals, enables more honest assessment. You can say that fund paid out successfully, that one was moderately effective, and another one failed completely. The person who led that fund can then take that track record to their next position as a legitimate career advancement. We deny people this opportunity when we maintain the fiction that we’ve always had these programs run by different people with slightly different strategies. This obscures rather than clarifies outcomes. We could simply be honest: we executed that fund, now we’re doing something new with a clear beginning and end.

Improving China Research + Why Bother?

Jordan Schneider: It’s remarkable. I’ve spent considerable time on foundation websites researching whether organizations like the Ford Foundation might fund ChinaTalk. They have mission statements about “democratizing equity,” which is admirable — I agree we should advance democracy and create more opportunities for people. However, the problem arises when you’re only a passive recipient of pitches. You’re essentially letting grantees define what success means, and the counterfactual becomes very difficult to assess. The organizations you’re funding would probably exist whether you give them $100,000, $500,000, or nothing at all.

The alternative approach would be starting with specific objectives — “We are trying to achieve X by Y timeline,” then working backward from that goal orientation to identify the people and organizations who can take your money and provide the highest probability of achieving that outcome. This approach is far more strategic, and my frustration isn’t about the politics of how they set their goals — it’s that they need to engage more seriously in the process.

Kumar Garg: Let’s work through this together. Imagine you and I were designing a fund model versus a program model for increasing collective intelligence on US-China relations.

There’s a vague way we could structure this — “This is the US-China program. It will have three tracks — funding scholars studying China, engaging policymakers about those insights, and warehousing data and research publications on these topics.” This resembles how most programs operate: they establish a broad framework with several tracks, then people apply under those categories. But if I asked what constitutes winning or how we’d know this program succeeded, you’d probably say, “Well, people applied and we distributed grants."

You could execute the same concept with much sharper focus by asking, “What would success look like in three years?"

Jordan Schneider: Exactly. I want ten books written that are so thoughtful and essential to the future of US-China relations and American policy that Ezra Klein would be compelled to feature these authors as guests because the thinking these grants produced is indispensable. Then we work backward from that goal to determine the budget.

We’d estimate the costs — ten books, assuming a one-in-five success rate for people with strong proposals to execute effectively, calculate the pipeline requirements, and arrive at a number that gives us a 75% confidence interval for producing those ten books by 2029.

Kumar Garg: Exactly. That approach feels like a sound tax strategy built on what resembles a very tight OKR, which you may completely fail to achieve. You might be wrong, but it’s precise. Then you build your strategy around that goal.

You find someone to execute it — let’s call it the Jordan Fund. If you succeed, people ask, “Jordan, how did you pull that off? You wrote down this ambitious goal, built a strategy around it, and executed successfully. You’re clearly skilled at this.” When you pursue your next initiative, you can say, “I ran this fund called the Jordan Fund. We set this audacious goal to produce ten bestsellers on US-China relations, and we achieved it."

This feels much more tangible as an actionable strategy — something that field leaders can point to as real-world impact, even if it fails. Let’s say you only achieve partial success — you still have concrete lessons. Compare that to “the US-China program makes some awards and does some things.” How do I assess whether that’s working?

Jordan Schneider: It’s remarkable because the market is so powerful — you can’t get away with this approach when running and scaling a business, especially when taking other people’s capital and trying to generate positive returns for them.

What’s curious, Kumar, is that very capitalist people become surprisingly touchy-feely when it comes to philanthropy. There’s an emotional layer where they think, “This is giving, so we shouldn’t apply business mindsets and OKRs.” It feels somehow dirty to them.

Kumar Garg: Here’s what’s important — we need to distinguish between current donors and potential donors. People focus extensively on today’s active donors, but if you examine the statistics on potential giving, current donors might represent only 1-2% of the actual addressable universe.

The question becomes, would we attract an entirely new class of donors if we brought this level of rigor, precision, and targeted approach to philanthropy? This would feel much more familiar to their professional experience.

Why don’t existing donors demand this rigor? I believe there’s significant pent-up interest in this approach. People oscillate between thinking, “Since this isn’t about making money, we’ll substitute with having a really complicated theory of action — that’s where we’ll apply our intellectual capacity.” I often say that just because something has numerous boxes and slides doesn’t substitute for having a clear attack vector.

Jordan Schneider: Exactly — rigorous thinking. The median nonprofit worker is about five times more likely to be socialist than the average person, so perhaps people more attracted to touchy-feely logic are simply concentrated in current organizations.

Kumar Garg: Sometimes the nonprofit and philanthropy sectors spend too much time engaging in collective mission statements, as if shared purpose alone is sufficient. But we actually have distinct roles to play, including making high-quality decisions about where to deploy finite resources. Because money is limited, you must make decisions strategically and place informed bets.

This may feel reductive, but it’s actually the responsibility inherent in this work. You must be a responsible steward because high-quality decisions produce more good. People sometimes struggle with that reality.

Jordan Schneider: Without high-quality decisions, you end up with USAID getting canceled. That’s our current reality. Organizations that weren’t evidence-based and couldn’t effectively justify their impact had some good projects and some poor ones. They faced criticism from small but vocal movements — organizations like Unlock Aid, whose founder we featured a few years ago and will have on again — arguing for more rigor because there was substantial waste and inefficiency.

If you let these issues fester too long, consequences follow. I don’t want to say universities, the NIH, or the NSF “had it coming,” but one of the best defenses you can have is a tight, well-justified organization that can stand up for itself.

Kumar Garg: I don’t want to engage in victim blaming, and I don’t want to excuse what I consider sometimes bad-faith behavior. However, your point about systemic advantages is valid — caring deeply about systemic impact and bringing that rigor and constant evaluation is useful for the work itself, but also valuable when those political fights emerge. You can say, “Look, we’re building something substantial."

In some of these cases, who knows what impact rigorous evaluation might have had. We’re living through unusual times, but I believe we’re gaining traction because there’s significant pent-up demand for this approach.

Jordan Schneider: Good. That makes me feel somewhat better, I suppose.

Kumar Garg: What should donors know about China? That’s my question for you.

Jordan Schneider: The original impetus for ChinaTalk was thinking about long-term national strategic competition and competitiveness from an industrial systems and technology perspective — identifying things people could do to nudge outcomes in liberal democracies’ favor. During the Biden era, I observed errors that legislation and executive action could fix with modest improvements — 5% here, 10% there. A sophisticated understanding of what’s happening in China could meaningfully help squeeze that extra 10% out of various decisions.

However, the policy changes we’ve witnessed over recent months regarding long-term strategic competition — how the US relates to allies, approaches to global nuclearization, science and technology funding, and immigration — are much more fundamental. Getting to a better place doesn’t require understanding what made BYD successful, how SMIC is developing its chips, or even China’s new AI policy. These are much more basic issues.

The thesis I operated under during the Biden era was that deeper, more considered understanding of China would lead to smarter policies. That’s now become a sideshow compared to more fundamental questions. If we accept the base case that science is important and immigrants are crucial for better science, then we should pursue those priorities directly. I would choose that approach ten times out of ten.

Returning to the order-of-magnitude questions I asked at the beginning, I would choose a NATO that functions as a genuine alliance ten times out of ten over determining the right tariff level for Chinese electric vehicles or batteries. That’s why I lean toward “be nice to allies” bumper stickers and NSF funding advocacy rather than tightly nuanced “we need to better understand China” approaches when considering ChinaTalk’s decadal competition mission.

Kumar Garg: One thing I’ve been considering, though I don’t have the answer yet, is what new institutions we need. Much of what I care about regarding how science operates in this country has been overturned. The idea that we’ll navigate this period with identical institutions seems unlikely — whether it’s who makes the case for science, who serves as science messengers, or how we conduct science itself.

This raises questions not just about policies, but about institutions themselves. Obviously Renaissance is part of that response, but I have broader concerns. We’ll probably need new institutions because the players on the field will have to change. Systemic change of this magnitude requires that everything else engage in significant adaptive change for us to succeed. That seems unlikely with current structures.

That’s a major meta-question I’ve been asking the team, “We won’t be able to handle everything directly, but what institutions would restore us to better footing? Do we have them? Do we need to create them?”

Jordan Schneider: I’m starting to focus my energy differently because I’m uncertain whether additional ChinaTalk podcasts about the importance of allies will accomplish much. However, one constant you can expect over the next four years is AI development and rapid technological change.

Regardless of controversial Trump policies, the Defense Department will persist, and America will still need to protect itself. America engages in conflicts approximately every three years, so that pattern will likely continue. Perhaps this is just me entering a kind of intellectual monk mode after dynastic change, but I’ve been reading extensive military history and examining periods of rapid technological change — specifically, what it means to deploy these tools more effectively than adversaries.

This doesn’t directly answer your question, but I’m pursuing intellectual journeys rather than policy ones.

Kumar Garg: Here’s what I’d say, which connects to the role you and ChinaTalk are playing. One thing I mentioned to Jordan before we started is that frequently when people reach out to me, they reference hearing a great ChinaTalk episode. You may not have set out to do this, but you’re playing a valuable role in shaping how other people — especially technical professionals — think about problems worth solving and their mental frameworks for our current age.

People are seeking understanding and meaning in this moment. The question isn’t about marginal additional podcasts, but whether you’re providing people with new vectors for their lives and careers. When we first met, you described yourself as a nerd passionate about these topics but uncertain where to channel that energy. You’ve created something quite distinctive.

We might be living in an era of unusually shaped careers, and we need to give people more space for that kind of professional evolution.

Subscribe now

Jordan Schneider: That’s fair. Some of this material feels obvious to me but may not be obvious to people who don’t live and breathe these topics daily. It’s strange because I don’t feel like I’m part of some resistance movement — I’m just a guy with opinions on various issues. Some days I wake up feeling helpless, others I feel genuinely empowered. This isn’t a direct response to your question, but that’s my reality.

Kumar Garg: You’re reasoning in public. You’re expanding people’s understanding of how to think through these complex issues and broadening their sense of who engages in this work. What I consistently observe is that people operate in highly siloed environments. They’ll mention their sources, and I immediately know exactly who they’re reading or listening to.

If you can expand their perspective while providing actionable next steps, that’s valuable. Part of my goal is always emphasizing that there are numerous hard, interesting problems to solve across every arena. No arena is the wrong one. People who dismiss politics are missing a crucial point — we’re all living within political systems whether we acknowledge it or not. Don’t dismiss any arena; simply understand that there are multiple dimensions to engage with. There are challenging problems to solve, and nobody benefits if you just remain a passive observer in the cheap seats.

Jordan Schneider: Perhaps the way I justify all my World War I reading is that no one else is doing this specific work. I’m bringing historical insights to current issues as someone who also reads contemporary news and has the freedom to spend ten hours weekly on intellectual journeys exploring topics I consider relevant to today. That seems like a natural conclusion.

Kumar Garg: We’ll do an update in a year. Hopefully, the republic endures.

Jordan Schneider: Kumar, you need to set ambitious goals for yourself — accomplish so much that I have to bring you back in six months.

Kumar Garg: Absolutely. One of our primary goals is becoming more international. We have a partnership with the British government to develop their R&D ecosystem, and we want to expand that model to additional countries. We’re building an organization that continues to internationalize because science and technology are inherently global.

I’m hoping our fund model will attract new donors who have never engaged in philanthropy before. Beyond discussing our approach, I want the actual work to manifest tangible results in the world.

Jordan Schneider: Contact info@renphil.org if you’re wealthy, have innovative ideas, or simply have a technology-related challenge you need help addressing.

Kumar Garg: I’m also available on LinkedIn — please reach out. We consider ourselves fundamentally a talent network, so I’m always eager to connect with people who have compelling ideas.

Jordan Schneider: We should do a little parent corner. We’ll keep this as part of the annual check-in. We talked about slime last year.

Kumar Garg: We did. Here’s something we discussed before we started recording — I asked you about sleep training, and you mentioned being hesitant to push sleep training advice on others. I’m strongly convinced that sleep training is a gift you give your children. We had twins and committed to sleep training. They’re eleven years old now and remain excellent sleepers today. We attribute this directly to that early sleep training.

For any parent listening who’s on the fence and wants random advice from someone they’re hearing — I can’t offer this to everyone, but I usually tell people I know that I’m always happy to be anyone’s texting buddy, providing extra support to get through those terrible first few days when it feels like you’ve made a horrible mistake. On the other end, you have children who can sleep well, which benefits everyone.

Jordan Schneider: I’m with you. I outsourced this decision to my mother — maybe one of the best decisions of my life.

What’s a cute development we’ve observed recently? I bought a ukulele two years ago, thinking it would be nice to play with my kid. What’s been charming is that my daughter is nine months old now, and there was a period where her manual dexterity only allowed her to grab the strings and pull them. But one day she figured out plucking, and now she’s actually plucking the strings. It’s such a cool activation moment for her — realizing “I make the sound now” instead of just dragging this thing around the room.

Kumar Garg: The period from nine months to eighteen months is truly remarkable. You’re approaching walking and first words, then vocabulary takes off exponentially. It’s incredible developmental progress, so I’m very excited for you.

Jordan Schneider: Perfect. Let’s wrap it up there. Kumar, thank you so much for being part of ChinaTalk.

Kumar Garg: Thank you for everything you’re doing. I’m excited to have participated and look forward to future conversations.

Today we’re running an interview of Huawei founder Ren Zhengfei that made it to the front page of yesterday’s People’s Daily, the premier Party outlet.

What struck out to me in the Q&A was just how hard he hit on the importance of basic research to long term competitiveness. “Without basic research,” Ren says, “there are no breakthroughs, and we’ll never catch up with the U.S.” The People’s Daily giving Ren Zhengfei the airtime to promote this view underscores just how central China’s leadership today sees this work.

We’ve covered the US side of how the Trump administration may be killing our golden goose in this regard with its treatment of the NSF, NIH, university ecosystem and immigration system in recent episodes on America’s R&D Reckoning, our interview with Josh Wolfe of Lux, and a recent op-ed on the war on S&T talent.

This translation also serves as a reminder to subscribe to ’s excellent and meditative daily People’s Daily recaps in his substack Tracking People’s Daily.

"The more open a country is, the more it drives us to progress" — A Dialogue with Ren Zhengfei

By reporters Hu Jian and Chen Jiaxing People's Daily (June 10, 2025, Page 1)

Recently, at Huawei headquarters in Shenzhen, focusing on some hot topics of public concern, a group of People's Daily reporters had an in-person interview with Huawei CEO Ren Zhengfei.

Q: Facing external blockades and pressure, encountering many difficulties, what are your thoughts?

A: I haven't thought about it—thinking about it is useless. Don't think about difficulties—just get on and do it [干就完了] and move forward step by step.

Q: Ascend chips have been "warned on" about usage risks [BIS . What impact does this have on Huawei?

[BIS in mid-May warned firms anywhere in the world that using Huawei Ascends, which were almost certainly fabbed by TSMC under false pretenses, would be considered a violation of export controls. The Information recently reported that “The fallout [from this BIS action] was swift: one Chinese data center firm promptly halted even the small order of Huawei chips it had planned to make, while its legal team scrambled to understand how it could comply with the regulations, according to one person with direct knowledge of the change. Meanwhile, executives at ByteDance and Tencent, both of which have significant business footprints outside China, worry adopting Huawei chips would invite scrutiny from the U.S. government, according to people familiar with their thinking.”]

A: There are many chip companies in China, and many of them are doing quite well—Huawei is just one of them. The US has exaggerated Huawei's achievements; Huawei isn't that formidable yet. We need to work hard to live up to their assessment. Our single chips are still one generation behind the US.

[The incentives are interesting here. On the one hand you have Jensen Huang going around saying Huawei is an enormously powerful competitor who dropped Nvidia’s China market share from 95% to 50%—even though, as the Information very generously put it, “Market share data to confirm Huang’s statement is hard to find.” Jensen, of course, wants the US government to lift export controls not on semiconductor manufacturing equipment but his ability to sell AI chips and racks into China.

The Chinese government and Huawei have every incentive to sandbag Huawei and SMIC’s capacity to compete. They’ve deeply internalized the backlash from Made in China 2025 and want the Trump administration to think that easing export controls on inputs to semiconductors and not investing in the state capacity to enforce the current ones is less of a give than it appears.]

We use mathematics to compensate for physics, non-Moore's to compensate for Moore's [用数学补物理、非摩尔补摩尔], and cluster computing to compensate for single chips. In terms of results, we can still achieve practical applications.

[For more color here, see SemiAnalysis’ coverage of the CloudMatrix 384, Huawei’s answer to Nvidia’s rack solutions]

Q: If there are difficulties, what are the main ones?

A: Difficulties are just difficulties—when hasn’t humanity faced them? Wasn't slash-and-burn agriculture difficult? Wasn't the Stone Age difficult? When humans used stone tools, how could they have imagined high-speed rail? China has opportunities in mid-to-low-end chips, with dozens or hundreds of chip companies all working very hard. Especially in compound semiconductors, the opportunities are even greater. For silicon-based chips, we use mathematics to compensate for physics, non-Moore's to compensate for Moore's, utilizing cluster computing principles to meet our current needs. Software cannot be strangled [卡不住脖子]—it's mathematical graphical symbols, code, and advanced operators and algorithms stacked together, with no barriers blocking it. The real bottlenecks lie in education and talent pipelines. China will eventually have hundreds or thousands of operating systems supporting progress in Chinese industry, agriculture, healthcare, and more.

Q: Huawei is getting a lot of praise and public recognition these days.

A: Praise brings pressure. A bit of scolding keeps us clear‑headed. We make products; once people use them they will criticize—that’s normal. We allow criticism. As long as people tell the truth—even if it’s negative—we welcome it. Praise or blame shouldn’t matter; what matters is whether we can do our job well. If we do, there’s no problem.

Q: From your attitude toward difficulties and criticism, we sense you have a strong inner core—not caring whether it's praise or criticism, but firmly doing your own work well. This should be an important reason why Huawei has come this far.

A: There's still too much praise for us. People should pay more attention to understanding those engaged in basic research. They are highbrow and solitary [曲高和寡—literally ‘difficult songs find few singers’], ordinary people don't understand them, and it takes decades or centuries to see their contributions. Groundless accusations against them are detrimental to the country's long-term development. We need to understand and support those doing theoretical work. We need to understand their broad-mindedness—their great anonymity is our country's hope. Don't elevate one while suppressing another; those engaged in theoretical research are the hope of our nation's future.

"Basic research scientists are lonely—we must have strategic patience and understand them"

Q: How do you view basic research?

A: When our country possesses certain economic strength, we should emphasize theory, especially basic research. Basic research doesn't just take 5-10 years—it generally takes 10, 20 years or longer. Without basic research, you plant no roots. And without roots, even trees with lush leaves fall at the first wind. Buying foreign products is expensive because their prices include their investment in basic research. So whether China engages in basic research or not, we still have to pay—the question is whether we choose to pay our own people to do this basic research.

Q: Regarding basic research, people might not understand the point of it.

A: Scientific breakthroughs—there are inherently few people in the world who understand them, so those who don't understand shouldn't evaluate them. Einstein's discovery that light bends was only proven a hundred years later. In Guizhou, there's an agricultural scientist named Luo Dengyi. In the 1940s, when analyzing the nutritional components of fruits and vegetables, he discovered a wild fruit called cili [刺梨] with very high vitamin content. China was still in the Anti-Japanese War period, education levels were low, and few understood. Later he wrote a paper saying cili was the "King of Vitamin C." After nearly a century, Guizhou has made it into a natural vitamin-rich cili beverage, a luxury item among vitamin drinks at nearly 100 yuan per bottle, highly sought after. The cili industry has become a channel for farmers to escape poverty and become prosperous. Only then did people truly recognize Luo Dengyi, who worked at a broken table during the flames of the anti-Japanese war.

[A deep cut by Ren Zhengfei here but this does look incredibly tasty!]

Q: Many results that look irrelevant at first yet turn out to be hugely useful.

A: Theoretical scientists are lonely; we must have strategic patience and empathize with them. Tu Youyou’s discovery of artemisinin is one case; So was Huang Danian, who embodied "exploring innovation and serving the country with utmost sincerity." The symbols, formulas, and thinking in their minds can only be communicated with by a few people in the world. We must respect theoretical scientists because we don't understand their culture. Society must be tolerant, and the nation must be supportive.

Q: Basic research takes a long time, but companies must deliver returns.

A: We spend roughly 180RMB billion a year on R&D; about 60 billion goes to basic research with no KPIs, while around 120 billion is product‑oriented and is assessed. Without basic research, there are no breakthroughs, and we’ll never catch up with the U.S. [没有理论就没有突破，我们就赶不上美国].

Q: That’s true long‑termism. We heard Huawei even has a “Huang Danian Tea‑Thinking Room.”

A: Professor Huang was a great scientist. China noticed him during the Gulf War: the U.S. used a pod under helicopters to detect Saddam’s buried weapons and destroy them instantly—that pod had been developed by Huang at a UK university. He later resigned from his UK university position and taught at Jilin University. He spent his own money to get a 40 m² room, offering free coffee so students could “absorb cosmic energy over a cup of coffee.”

[Honestly this looks pretty lame, Huawei didn’t do the guy justice with this underwhelming study hall]

We have his family’s authorization to use his name for a non‑profit online platform—the Huang Danian Tea‑Thinking Room—where anyone can freely access global scientific information.

We also keep the “mouth of the funnel” wide open for basic research, partnering with universities. These are strategic investments with no performance metrics. Internally, we accept that we don’t know when theoretical work will bear fruit and impose no demands on the scientists.

“One purpose of socialism is to advance society”

Q: [A cringe leading question here…] Economist Richard Wolff and others argue that the U.S. lacks a high‑speed rail system because capitalism demands profit, while China’s socialist approach values social benefit: high‑speed rail, heavy‑haul lines, advanced power grids, expressways, rural cement roads, irrigation works, power plants… none make money directly, yet they underpin modernization. Competitive goods, meanwhile, are market‑regulated. What’s your view?

A: Why is it that only socialism tackles unprofitable tasks? One goal of socialism is precisely to develop society. China’s socialist market‑economy model is a magnificent undertaking. In infrastructure we had to follow this path—otherwise high‑speed rail, expressways, dams and so on would never have been built.

Q: What’s your outlook on AI?

A: AI might be humanity’s last technological revolution—barring perhaps fusion power. It will take decades or centuries to unfold. No need to worry; China has many advantages. [This line made leadlines in non-Party media]

Q: What advantages?

A: Hundreds of millions of young people—they’re the nation’s future. The General Secretary has said a nation’s strength rests on cultural flourishing. Technically, AI hinges on ample electricity and advanced networks. China’s power generation and grids are excellent; its communications network is the world’s most developed. The “Eastern Data, Western Computing” vision can become reality. [For more, see ChinaTalk’s past coverage of EDWC.]

Q: Anything else?

A: Chip worries are unnecessary: with stacking and clustering, our computing outcomes match the state of the art. As for software, open‑source ecosystems numbering in the thousands will meet society’s needs. [Wait, earlier in this interview I thought we were stressed about how behind Chinese chips were…?]

Q: How do you see China’s future?

A: After leaving our company, Tom Friedman bought a second‑class high‑speed‑rail ticket on his own to experience China. He later wrote, “I’ve seen the future, and it’s not in America.” [See ChinaTalk’s interview with Ezra Klein where I discuss this Tom Friedman piece with him. I, for one, would not take cues on analyzing China from Tom, who said in his Ezra interview that “Whether I’m writing about China from Washington, or whether I’m writing about China from China, I’m always just writing about America. My goal is to use China as my permanent Sputnik.”]

Q: Friedman said in that piece: “What makes China’s manufacturing juggernaut so powerful today is not that it just makes things cheaper; it makes them cheaper, faster, better, smarter and increasingly infused with A.I.”

A: Fundamentally, algorithms aren't in the hands of IT people, but in the hands of electrical experts, infrastructure experts, coal experts, pharmaceutical experts, and experts from various industries. At the practical level, Chinese manufacturing's application of artificial intelligence is very rapid, and many Chinese models will emerge.

Q: What support do private enterprises need from the state for their development?

A: Rule of law and marketization, with government administration according to law and regulations. Enterprises should focus mainly on value creation, technological breakthroughs, law compliance, and legal taxation. This harmonious development model will gradually release economic vitality.

Q: How do you view openness and development?

A: The country is becoming increasingly open, and openness will drive us to greater progress. Under Party leadership, the nation is administratively unified with smooth policy implementation. The gradual formation of a unified large market is possible and will certainly break through all blockades to achieve great rejuvenation.

[Ren Zhengfei of course got the seat of honor during the the Feb 17 Symposium featuring Xi which doubled as the Party’s olive branch to its leading S&T firms. We covered this meeting here on Xi’s Hard Tech Avengers.]

For more, see other core Ren Zhengfei speeches we’ve translated.

Read more

Has America already lost its dynamism in basic research? Josh Wolfe, co-founder of Lux Capital, joins the podcast today to discuss.

We get into…

• Why the Trump-Elon fallout matters less than you probably think,

• How much payoff corporate leaders are reaping from their campaign to appease Trump,

• The erosion of the U.S. research ecosystem, and how we should think about philanthropic giving amid that chaos,

• Parenting, strategies for emotional resilience, why short videos aren’t terrible, and the history of the machine gun

It’s a good interview. Listen on Spotify, iTunes, or your favorite podcast app.

Dreams of the Tech Right Meet Reality

Jordan Schneider: We have to start with Trump and Elon. I’ll let you pick the frame.

Josh Wolfe: I’m going to take a contrarian position — it’s irrelevant. It’s irrelevant because it dominates front-page news and captures everyone’s attention. I joked yesterday that if you were planning fraud or were a terrible company with bad news, yesterday was the perfect day to release it while everyone focused on this inevitable outcome.

People were making predictions six months ago — over/under bets on whether it would last three months or maybe a year.

Jordan Schneider: There were actual betting markets created for this.

Josh Wolfe: Exactly. You were betting under a year, maybe over three months, but none of this was surprising. This was an unstoppable force meeting an immovable object — kinetic chaos was inevitable. Elon served as a useful foil for Trump and did an excellent job helping him win the election. They were strange and unusual bedfellows.

Trump has historically been a strong China hawk, albeit with nuanced moments regarding Xi Jinping. Meanwhile, Elon depends heavily on China in ways people don’t fully appreciate — particularly Tesla’s production levels and the profit margins China contributes.

Jordan Schneider: Well, the China Talks audience understands this, but perhaps not everyone does.

Josh Wolfe: Right, not the broader ecosystem. This creates a significant vulnerability. You’ll see Elon Musk historically position himself as an outspoken advocate of free speech — though whether Twitter was genuinely about free speech is debatable. I’m more cynical about his true aims.

What you’ll never see him discuss publicly is anything about China. He’ll never criticize Xi for human rights abuses in Xinjiang province or regarding the Uyghurs, and he’ll never mention Tiananmen Square. Nothing. It’s “Free speech for thee, but not for me” when it comes to China.

Inevitable clashes were bound to occur. Was this really about the spending bill, or something deeper? When the leaks started — clearly planted by the White House over the past five days about drug use, and then the Epstein connection — it became salacious and interesting for everyone, but none of it was surprising.

My position is that it doesn’t matter. It’s similar to tariffs — Trump campaigned on tariffs, Democrats attacked him, saying he’d implement tariffs that would be economically troublesome. I don’t understand why these developments are surprising when they shouldn’t be.

Jordan Schneider: Speaking about your profession, many people in the tech ecosystem — particularly with the rise of the tech right throughout 2023 and 2024 — pinned their hopes on this man, similar to how Elon did. They hoped that by being inside the tent, they could influence policy trajectory. From my perspective, benchmarking to June 6, 2025, this has been pretty dramatically disappointing.

What are your reflections on that psychological arc? Are these guys all just patsies?

Josh Wolfe: During November and December of 2024, I remember peak rhetoric around the zeitgeist of America’s golden age. All you could hear was “we’re back, baby” and these ridiculous 1980s Top Gun-infused, Hulk Hogan maverick-style videos celebrating America.

Jordan Schneider: Sure.

Josh Wolfe: I told my wife, who runs an activist public hedge fund and manages our personal money, “We need to buy long, deep out-of-the-money puts” because all I heard was “to the moon” and “American greatness.” It felt like an echo chamber of optimism.

Two to three months ago when the tariffs hit, those positions performed very well in our portfolio. We’ve been relatively unaffected on the venture side.

One area where real change is happening — and it’s always dangerous to say “this time is different” — but it feels very different: defense. We’re seeing a combination of a nearly trillion-dollar budget and a huge shift toward autonomous systems, AI-driven software systems, and space satellites. Remember, Space Force was an absolute joke eight years ago.

Jordan Schneider: Okay, we’re pivoting a little…let’s stay on track.

Josh Wolfe: But this is where the venture world has made noise and impact. There are sympathies, appreciations, and influence, particularly in tech and defense.

Jordan Schneider: Sure.

Josh Wolfe: Regarding everything else: you had everyone pledging fealty, taking a knee, and donating a million dollars to the inauguration campaign. Zuckerberg, Bezos, Satya, Sundar — they were all paying homage to Trump, hoping to escape DOJ, FTC, or antitrust scrutiny. How that plays out remains to be determined.

The Trump-Elon situation was both weird and destined to end inevitably. There will be some strange reconciliation, and they’ll be bros again. But I don’t think this is just about the spending bill — it’s much deeper.

Jordan Schneider: One of the things the tech right didn’t price in is what’s happening around immigration and the basic research ecosystem. Let’s start with basic research. There’s this argument that we don’t need the NSF or NIH because corporate R&D has increased significantly over the past 30 years and can drive innovation forward.

As an investor who focuses on scientifically ambitious investing, what do you think the economy gets from basic fundamental research that happens in labs and universities versus corporate R&D?

Josh Wolfe: Long-term: everything. Short-term: it’s hard to see the value. For policymakers looking at budget cuts — particularly those focused on short-term gains — these programs seem like easy targets. This isn’t just a Trump phenomenon. In 20 of the last 22 years, we’ve had federal cuts to federal science funding.

Jordan Schneider: We had a brief moment with the Endless Frontier Act.

Josh Wolfe: Right, it started with Endless Frontier, which referenced Vannevar Bush’s work from around 1945. Then you had “The Gathering Storm” by Norm Augustine, the former Lockheed executive, warning about America’s talent base and cultural shifts in what people were drawn to and celebrating.

What they didn’t mention in that report — now over 20 years old — was this anathema, this zeitgeist against the military-industrial complex, precisely when China was not only embracing but mandating military-civil fusion.

Over the past two to three months, we’ve witnessed a perfect storm. Now you have the politicization of academia. Harvard sits number one in Trump’s crosshairs — some speculate because Barron applied and didn’t get in, creating some Shakespearean vendetta. Whatever the reason, while anti-Semitism was the stated concern, it was far worse at Columbia, right here in New York City, than at Harvard.

Subscribe now

According to Nature magazine’s rankings — which publish various metrics including H-index measurements — when you look at the most important scientific publications by institution, the top 10 includes Harvard at number one, but positions two through ten are all Chinese universities.

Fifty percent of AI graduates worldwide come from China. Thirty-eight percent of the AI and computer science workforce here comes from China — not native-born Americans. They’re outpacing native American citizens by 35-37%.

We’re losing the talent game. What should we do to win? What we did in World War II and beyond: attract the best and brightest. Create brain drain from other countries because people are repelled from their home nations.

Eastern European Jews in the 1940s established the Institute for Advanced Study and brought Einstein here. Having the atomic bomb developed here rather than Germany was a net positive for the world. Soviet émigrés in the 1980s escaped communism for capitalism during the Cold War.

We should be stapling visas to every Chinese, Indian, Pakistani, Israeli — anyone who wants to come here. We should help their parents immigrate too, because often family remaining in authoritarian home countries becomes leverage that governing regimes use to control whether these individuals can leave or return.

The last Republican I remember discussing this was nearly 20 years ago: Newt Gingrich, who called for tripling the NSF budget. He understood that everything from Genentech (which emerged from the UC system) to Google (from Stanford) was premised on long-horizon scientific research.

This isn’t just computer science — it’s chemistry, physics, materials science, and all the breakthroughs that emerge from these fields. If we knew what those breakthroughs would be, we’d fund them today. But we don’t. We rely on this rich ecosystem in our own self-interest to support brilliant people who generate breakthroughs.

These breakthroughs often result from combining insights from one department with research from a different university. More than ever, this cross-pollination is critically important. We’re ceding academic and intellectual leadership to other countries. China will likely be the greatest beneficiary.

International students used to comprise roughly 25% of our student body. Now it’s down to 12-15% and falling. China is actively recruiting with programs like their Thousand Talents initiative.

This represents the greatest self-inflicted wound Americans have created in several generations.

Jordan Schneider: You had this line: “If the Politburo were drafting America’s self-sabotage plan, they would politicize science and freedom of inquiry, starve young and ambitious investigators of funding, and discourage the best immigrant minds."

To recap for everyone: the NSF is spending at half the rate it was in 2024, despite authorized funding. The new budget dramatically cuts science and research spending. Universities have been cut off wholesale from billions in research dollars. Johns Hopkins is laying off 1,000 people because they’re worried grants won’t materialize.

Trump said yesterday that we actually want Chinese students, which creates bizarre mixed messaging when his administration simultaneously tells Harvard they can’t admit Chinese nationals.

Josh Wolfe: Some of these grant cuts are both tragic and absurd. There was a researcher studying neurotransmitters for neuroscience whose grants were cut. They couldn’t understand why until they realized the word “trans” triggered a five-letter string search that flagged anything with “trans,” “DEI,” or similar terms.

Now researchers have to change grant language — writing “neurons firing intracranially” instead of “neurotransmitters."

Jordan Schneider: That’s not funny, though.

Josh Wolfe: You’re right. It’s tragic.

Jordan Schneider: Why don’t you talk about the $100 million pool you launched?

Josh Wolfe: It’s specifically not grants because we’re not trying to substitute the charitable giving of our government — funded by taxpayers — which actually makes America great when we have scientists working on breakthrough research.

Historically, about 10% of our investments are de novo new companies, and most emerge from academic labs. We find a principal investigator — a fancy term for a scientist at an academic institution — whether at Harvard, Princeton, Yale, Georgia Tech, Cornell, or elsewhere.

The early 1980s Bayh-Dole Act, sponsored by Senator Evan Bayh and Senator Bob Dole, allowed universities to own intellectual property created with federal taxpayer research funding. When a scientist at these institutions files a patent, the university becomes the assignee while the scientist remains the inventor.

We, as a venture capital firm, can license that technology. There’s a well-established deal structure involving licensing royalties and equity — the scientist typically gets about 25% ownership. This mirrors Google’s “one day a week” policy, allowing employees to spend time on personal projects, but applied to exclusive company work.

This approach for launching companies from academic labs is well-established. Professors often stay at their institutions while postdocs handle translational research and join the company. We’ve created about 25 de novo companies this way, spanning everything from 4D lidar for autonomous vehicles to digital olfaction — essentially “Shazam for smell” — to cancer therapeutics and Nobel Prize-winning work from the Howard Hughes Medical Institute that enables real-time cellular imaging.

Historically, this represents about 90% de-risked investment, meaning we take roughly 10% scientific risk. Now, with grant cuts and layoffs at institutions like Johns Hopkins, exceptional research is being abandoned — like proverbial Rembrandts in the attic. Scientists are wondering what to do next. If they wait six months for resolution, they need to find work elsewhere, whether in academia, nonprofits, or the private sector.

Our response is the Lux Science helpline — a bat signal for struggling researchers.

We’re dedicating $100 million from our latest $1.2 billion fund to double down on early-stage science risk. Instead of taking 10% science risk on 90% proven concepts, we’re willing to take 50% science risk on half-baked ideas.

We’ll help license patentable work into our existing companies or arrange sponsored research to continue funding relevant projects. For example, someone contacted us about novel materials for radiation hardening in space. We have 17 companies across the space ecosystem that could sponsor continued research.

Other scientists ready to enter the private sector can work with us to start new companies around their expertise. We’ll license their work, assemble their team, build laboratories, and launch them.

This is just a drop in the bucket — we can’t do this alone. We need dozens, if not hundreds, of other VCs to recognize the value of funding early-stage scientific work. It’s profitable, beneficial for national security, and good for business. Our $100 million allocation, while significant, isn’t sufficient by itself.

Jordan Schneider: I’ve received numerous emails from intelligent people essentially doing backfill work for USAID funding cuts. While it’s admirable that people are donating $10 million or $50 million, tens of millions of people will suffer, and we’re delaying future breakthroughs because we’re not conducting this research. The laboratory animals aren’t being fed, and you need the time, energy, and tacit knowledge built over years of lab work to accomplish meaningful research.

Josh Wolfe: Absolutely. This isn’t purely black and white — there are legitimate concerns about USAID politicization and questionable funding destinations. There are appropriate and inappropriate uses for that money.

My main criticism of the NIH isn’t to strip their funding, but to bias toward young investigators. Too much grant money goes to, frankly, older researchers. Older scientists can be set in their ways and resistant to new approaches.

You need that potent combination of naivety and ambition that drives both entrepreneurial and scientific discovery. Young researchers have the arrogance to say, “I know better than you — why wouldn’t we try this?” while older researchers respond, “Why would you do that? I’ve tried it ten times and it never works."

I would restructure the NIH budget and federal funding to favor younger investigators.

Jordan Schneider: I wonder to what extent the Silicon Valley tech ecosystem that embraced Trump and the DOGE energy — which led to all these cuts — stems from the fact that for the past 20 years, you could make the most money by being 19 and writing software. That’s not something you necessarily need a PhD for. The social returns to that work are very different from what you see when making novel materials and drugs.

Josh Wolfe: Absolutely. Even the term “engineer” shifted from physical engineer to software engineer. For at least 20 years post-internet boom — through the SaaS enterprise boom and cloud boom — that was the dominant paradigm.

Lux was on the periphery of that trend because we don’t really fund software. About a third of our investments focus on healthcare, biotech, robotic surgery, and medical devices. Another third covers aerospace, defense, manufacturing, and industrial applications. The final third is what we call “core tech,” defined more by what we don’t do — very little internet, social media, mobile, or ad tech.

Great fortunes were made and great companies built during the software boom. Interestingly, there’s a China parallel here. Fifteen years ago, Xi Jinping designated software as a domain China would fund heavily. This created incredible companies and ecosystems, and many US investors funded them, believing China was a democratizing, growing market. Knowing what they know now, it would be difficult to justify investing in companies like ByteDance.

Jordan Schneider: Unless you’re Bill Gurley [who recently funded Manus AI].

Josh Wolfe: I love Bill, and he’s not necessarily part of the partnership decision. We’ve discussed this, and he would argue there might be scenarios where teams like Manta are trying to exit China. I think this deserves more study rather than just criticism. However, I personally wouldn’t invest in Chinese companies that are part of CCP military-civil fusion.

The shift also happened within China. At the last Politburo gathering about a year and a half ago, the people close to Xi weren’t software engineers or computer scientists — they were from space, biotech, rockets, and defense. All hard sciences.

This coincided with when Jack Ma went to “spend time with his family” and when entrepreneurs were capped at $9.9 billion. Anyone exceeding that threshold was essentially decapitated — they could start schools or enter education, and either they or their family could leave the country, but not both simultaneously.

The new directive became hard sciences. I used to joke that if you wanted to make money, the greatest capital allocator wasn’t following Warren Buffett or Seth Klarman or other great value investors — it was listening to what Xi Jinping was funding, because that’s where the world was heading.

I agree that for the past 10-15 years, the focus has been software. Many VCs, including Marc Andreessen who said “software is eating the world,” were drawn to Trump partly because they felt rejected by Biden. They were spiting Biden and possibly spiting themselves long-term.

I still believe the greatest entrepreneurs don’t really care about political developments or the 10-year interest rate. They’re building something because they have a chip on their shoulder — and as I always say, chips on shoulders put chips in pockets. They’re driven by private ambitions with timeframes that supersede one, two, or even three presidential cycles.

I remain bullish on great entrepreneurs. There’s simply a shift from software to hardware.

Jordan Schneider: Something worth exploring is the importance of tacit knowledge in hardware versus software and the learning that happens almost entirely in universities during master’s and PhD programs. When you’re doing actual scientific work, you need some sense of navigating the dark forest.

Josh Wolfe: That’s absolutely true. This applies even to semiconductors and manufacturing. TSMC’s Arizona facility is starved of talent — not just union workers doing physical assembly, but specialized expertise. There’s a scarcity of talent because much of that tacit knowledge remains in Taiwan and is difficult to transfer here.

You see this dynamic in companies working on the most sophisticated technologies. My partner Sam Arbesman, who you know or have spent time with, is a brilliant scientist-in-residence here. He has a new book coming out called “The Magic of Code,” and his previous book “Overcomplicated” offered a modern version of “I, Pencil” — the thought experiment about complexity.

If you were to make a semiconductor today, or an Apple iPhone, consider the number of components, the tacit knowledge required, the number of countries and companies involved — it’s extraordinarily complicated. No single person can make a pencil, let alone a semiconductor, chip, GPU, or field-programmable gate array.

Culturally, we get what we celebrate. For 25 years, we at Lux have complained that American culture celebrates celebrities — the Kardashians, the Hiltons, and similar figures. You see this manifested in TikTok and what gets fed to American users versus what’s not even allowed in China and what China celebrates in terms of STEM education.

We’re losing this battle terribly. Looking at the labor pool: we have 300,000 undergraduates in science here. China graduates approximately 1.2 million. Half of our 300,000 are foreign students.

This represents a cultural crisis regarding what we want our children doing and celebrating. We don’t need more people in marketing, advertising, or selling products. We need people inventing things that everybody else in the world wants to buy.

Jordan Schneider: And podcasting.

Josh Wolfe: But you produce intelligence and insight, which is valuable. It’s unique insight because, as you noted, people who follow China Talk understand things that others don’t. That’s an advantage — you produce something intellectually valuable.

Jordan Schneider: Back to the narratives. There was this fascinating Twitter exchange a few days ago. You mentioned the atomic bomb and the Apollo program, and J.D. Vance framed it as “we didn’t need foreigners to do this — we had 600,000 Americans weaving the wires to connect everything together."

Both perspectives are true: his narrative about American workers is accurate, and your narrative about leading lights of both the rocket and nuclear programs coming from Europe is also true. But this narrow-minded “you have to be born here to be part of the circle” mentality — let’s discuss what that means.

Josh Wolfe: This created a significant fissure. People entering the administration who were immigrants from India were being lambasted by these self-described “heritage Americans.” Elon defended them because he’s an immigrant himself — not the classic immigrant from poverty in India or Russia, but an immigrant nonetheless.

He argued there’s a distinction between immigrants crossing the border and taking blue-collar jobs — who may not conform to American society and could include criminals or drug dealers — and brilliant people making extraordinary contributions. For example, the recent genetic medicine breakthrough that cured a baby using one of the first in vivo genetic editing approaches involved two scientists, one whose parents arrived from India 40 years ago. Thank God those individuals came here and had children.

We want this brain drain — from World War II through the Cold War to our current version of Cold War competition. I don’t understand why people consider “immigrant” a dirty word. Immigrants literally are the fabric of this country.

One of my mentors who put us in business, Bill Conway, who founded the Carlyle Group, focuses his philanthropy on addressing one of America’s biggest problems: nursing shortages. He says if you’re anti-immigrant, good luck getting sick, because our entire healthcare system depends on immigrants playing critical roles — from hospital orderlies to doctors to neurosurgeons.

We want the best and brightest coming here. It’s a complete self-inflicted wound to say “America First” means shutting out talent. America First doesn’t mean America Alone, as Scott Bessent has said. Scott is one of the great adults in the room and a friend.

People ask why Scott doesn’t speak out more publicly on these issues. Scott is a student of markets first — he understands currencies and countries — but he’s also a student of human nature. He understands in a Shakespearean way who his boss is, who he works for, and how to manage that relationship. We want him doing exactly that because we don’t want him out of the job.

Jordan Schneider: We discuss immigrants and science frequently on China Talk from a US-China competition perspective, and that’s all valid. But the fact that people want to come here is remarkable. Look, at the margins there are legitimate concerns about completely open borders, but limiting human flourishing by building walls and splitting families is just —

Josh Wolfe: I’m going to be blunt because I come from Brooklyn: this approach is counterproductive and self-defeating. The idea that we’re limiting the immigration of brilliant people who could make this country better is fundamentally wrong.

Our greatest export isn’t our music, Hollywood, fashion, high-tech companies, banking system, or rule of law — it’s the American Dream. The American Dream is arguably our greatest export; it’s what everybody wants. The measure of a great country is whether more people want to enter than want to leave.

Walter Wriston said, “Capital goes where it’s welcome and stays where it’s well-treated.” This applies to both money flows and people flows. I hope the rhetoric will change and that there will be enough great American leaders to counter jingoistic “America First” voices.

There’s a delicate balance: we must remain that shining beacon on the hill, attracting people while distinguishing between good people who genuinely want to build better lives here — who want to embrace the American Dream and become proud Americans — and bad actors who want to undermine the country. Many of the latter are coming through our university system and may be funded by foreign actors, but that’s a separate issue.

There’s virtue in virtuous people coming here and the American Dream being upheld and celebrated. Right now, we’re casting shadows on that dream.

Jordan Schneider: You opened by saying the Elon-Trump fight is noise. But the signal is that people around the world would rather live here than in China.

Josh Wolfe: They still do. My wife mentioned recently that many countries hate us right now — Canada, tourism is down — and there’s probably some schoolyard “arms crossed, I’m not playing with you” attitude.

Jordan Schneider: Sure.

Josh Wolfe: The reality is, I read about how Canada supposedly hates us, then the next headline I see — because I read many papers each morning, including the Globe and Mail — is “Canada wants in on Iron Dome."

The rest of the world, even if they hate us, still wants our superior military technology. The zeitgeist of popular antipathy will fade, but people will still want our materiel — our military materiel.

Jordan Schneider: We’re recording this a few days after the Tiananmen anniversary. The fact that hasn’t happened here — the fact that there is free speech, elections, due process, and habeas corpus — represents incredibly powerful long-term advantages. These aren’t just abstract principles — they’re things that make life worth living and make you excited to get out of bed in the morning.

Josh Wolfe: I remain relatively optimistic personally. I’m always optimistic about technology, science, and ultimately the human condition. I’m always skeptical about people because I’ve read a lot of Shakespeare. Technologies are amazing, science is amazing, and both will continue to progress. People generally disappoint because they’re vainglorious, full of ego, petty, jealous — they connive and deceive.

People ask how I can be optimistic yet cynical. That’s the pairing — I’m optimistic about science and technology, skeptical about people. I love Tom Wolfe’s answer when asked why he writes about space, astronauts, and moneyed Upper East Siders but not politicians. He said politicians are like passengers on a train, and the country is the train. The tracks might go up and down, but it’s generally heading in the right direction over time. Inside the train, there are clowns on one side in red and clowns on the other side in blue, throwing pies at each other nonstop. Every four years, the engineer changes, but the train stays generally on track.

I take that view: the real thrust comes from people, values, that great beacon on the hill, the American Dream that attracts people here. We shouldn’t deter them from building here on that track. Most entrepreneurs couldn’t tell you much about the political system or what’s happening at the Fed or Treasury — they’re focused on building.

Jordan Schneider: But the fact that they have to worry now, that they have to fear their visas getting revoked — look, in other moments that might be manageable. But within our range of expectations, there are definitely scenarios where an executive acting alone can really bend those train tracks.

Choosing Optimism + Parenting

Jordan Schneider: Speaking of Shakespeare, we did a show with Eliot Cohen about his wonderful book The Hollow Crown and another episode as a Biden emergency discussion. The Trump-Elon situation probably deserves its own Shakespeare emergency episode.

told me to ask you this question — does America need a Gen Z Marsha Linehan?

Josh Wolfe: Marsha Linehan is one of the founders of CBT — Cognitive Behavioral Therapy.

I love Derek. I got into psychology maybe five or six years ago. My kids are 15, 12, and 9. My oldest daughter is much more like me — emotionally volatile. I’ll have high highs (not bipolar…emotive!) then calm down. My wife and middle daughter have slow burns with grudges lasting three or four days.

We learned CBT and DBT techniques as a family. Derek’s question is whether Gen Z needs this — I think they do. CBT is essentially Stoic philosophy in a clinical psychology setting. I wish I’d learned this 20 years ago as a kid or teenager — I would have had healthier relationships. I wish I’d learned it 15 years ago before my first child was born — I would have been a better parent. We have a better marriage, family relationships, and even professional interactions because of these principles.

First, avoid extremes like “you always do this” or “you never do that.” If I said “Jordan, you’re always late for podcasts,” your first reaction would be defensive because it’s not true — nobody “always” or “never” does anything. People shut down and become defensive. You want to find the dialectic, avoiding black-and-white thinking. “Sometimes I feel frustrated when you’re late” is much more effective.

Second, when someone has an emotional outburst, they’re dysregulated for various reasons — hunger, a bad day, a lack of tools. If my daughter has a fit, she’s not thinking “I really want to have a tantrum and lose my phone for a week.” You appreciate that they’re doing their best with what they have — they just lack the tools.

Third, validate emotions. When someone’s really upset about something, you could say, “That’s ridiculous, it’s just a math test, you’ll get an A.” But explaining away someone’s feelings doesn’t help. People need validation. They become more emotionally frustrated when they don’t get that release valve. Saying “I can see you’re really upset” sincerely can help reduce their emotional burden.

Returning to Derek’s question about Gen Z, many people seem massively oversensitive. I grew up in Coney Island, Brooklyn, where people talked trash to each other, were rough around the edges, and the world wasn’t safe. You could say all kinds of things. I’ve raised my kids similarly — I don’t want them so soft that when someone says something offensive, they appeal to authority, running to school or teachers saying “he said this” or “she said that.” Take care of it yourself. Have a conversation. I’m not advocating violence, but handle yourself.

My kids have experienced this and tell us about various scenarios. The younger generation — whether you call them far left or woke — has lost some ability to engage with people. When encountering ideas or comments they disagree with, there’s hysteria.

I was a center-left Democrat my entire life. I didn’t vote for Trump, but I also didn’t vote for Kamala. I voted for Bloomberg-Romney, which was not a ticket, but that’s where my values aligned, in a non-swing state where my vote didn’t matter. I had the luxury of voting with my conscience.

I believe many center-left Democrats didn’t move right because they were attracted to the right. They were repelled by the histrionics, noise, whining, and complaining from the left. Enough was enough.

Share

Jordan Schneider: Let’s stick with parenting. You mentioned earlier that Steve Jobs accomplished incredible things, but everyone around him in his personal life had a totally different experience.

Josh Wolfe: They hated him. This is a fascinating phenomenon.

Jordan Schneider: This really resonated for me when I was reading a biography of Richard Holbrooke — the most storied American diplomat of the second half of the 20th century, though perhaps not well-known in the Asia-watching community. There was a line where his kid said his father wouldn’t recognize his grandchildren in a toddler lineup.

Growing up, I dreamed of being Secretary of State or a diplomat, but trading family relationships for that achievement gave me pause. Many people choose one path over the other. What are your reflections on this dynamic of accomplishing big things while being awful to the people closest to you?

Josh Wolfe: I used to debate this with one of my best friends, a famous journalist, who talked about how great Jobs or Elon were. My response was that the people around them — their families — hate them. They’re celebrated by strangers, feeling love from millions of people they’ve never met who celebrate a caricature of who they actually are. Meanwhile, their true character is loathed by those closest to them.

I’ve admired something I have to admit I’ll never achieve — I am too rough around the edges, have burned too many bridges, and have been abrasive to too many people. When Warren Buffett gave the eulogy for Coca-Cola CEO Don Keough, he summarized it in three or four words — “Everybody loved him.” I thought that was beautiful.

People will not say that about me, but I can control the decisions I make regarding my children. Will my kids feel that way about me?

I experienced something very salient and memorable involving my grandfather, who raised me. I grew up in Coney Island, Brooklyn — my mother, grandparents, and I shared a two-bedroom, one-bathroom apartment. We were very poor. My grandfather wasn’t my biological grandfather; he was my grandmother’s second husband, but he treated me like his son. He delivered the Daily News at night and was the most important man in my life.

He passed away the month before September 11th. At his funeral, his biological son was present. During the Jewish ceremony, when family members put dirt on the grave, his biological son approached with such animus. It was like “good riddance” — he took the shovel, threw dirt on the grave, and walked away. That sound still echoes with me.

I’ve always loved this quote from Carl Jacobi, the 18th-century mathematician: “Invert, always invert” — flip it on its head. My father wasn’t present in my life, so I became the father and husband he wasn’t. But watching that funeral moment, I resolved that would never happen to me. I never want my kids to feel that animus and animosity, wanting to dump dirt on my grave and walk away saying “good riddance."

The most important thing to me is what Adam Smith wrote about — not just the invisible hand, but the idea of being lovely and being loved. That second part, being loved, is scarce and valuable.

Scott Galloway recently gave a rant — I believe it was on Piers Morgan — where he talked about our obsession with Elon, innovation, and money while ignoring people who are hurting and suffering, including those losing USAID funding. He asked what it means to be a man. In his view, it wasn’t toxic masculinity — being a man means being able to take care of people.

We’ve lost that. There aren’t great male role models in public life today that young people celebrate for being good men through self-sacrifice. Everything centers on self-aggrandizement.

Regarding how to encourage or discourage people on certain paths: I want my kids to be truly happy. I ultimately don’t care where they go to college — the world has changed significantly. I don’t care what they do professionally. I want them to find purpose and meaning.

This relates to something interesting. Dan Senor works at Elliott Associates but is very active in Jewish and Israeli life. We attended Shabbat dinner at his house. I’m not religious — I’m an atheist but a tribal Jew. He blesses his children, and what struck me was that the blessing wasn’t about success, money, or career achievement. The blessing was “I want you to have a life full of meaning and purpose."

I thought that was beautiful. That’s what I want for my kids.

Philanthropy, Parenting, Short Videos

Jordan Schneider: We have five minutes for five quick questions. Take your pick — future of media advice for China Talk, New York mayoral race hot takes, what you’d do with a $10 billion philanthropic foundation, or something else.

Josh Wolfe: I don’t know.

Jordan Schneider: Where do you want to go?

Josh Wolfe: Mayoral race. I really hope Cuomo wins.

On advice for ChinaTalk, people get the audiences they deserve, and you have a smart, sophisticated, engaged audience looking for signal among the noise. Keep doing what you’re doing — your combination of Substack, podcasts, video content, periodic pieces, and great guests works well. I feel privileged to be here with you and enjoy reading your work.

Just maintain it because it’s a high-signal, no-BS voice. These things aren’t linear — there are step-change functions where suddenly something goes viral and you gain another thousand or ten thousand subscribers.

You should periodically write op-eds in major publications like the Wall Street Journal or Financial Times, sharing insights from your work. There are proxies like Stratfor and others in popular geopolitics, but you own a valuable niche. Over time, whether you do it or someone else does, there will likely be aggregation and acquisition — a China expert, Africa expert, defense expert — building a new media ecosystem where you could benefit from that outcome.

Jordan Schneider: Second and third questions — you’re pretty wealthy and likely to become wealthier. How do you think about big philanthropic investments beyond what you’re doing today?

Josh Wolfe: There are two aspects I really admire. Take what Bill Conway did — he’s not pursuing vainglorious naming opportunities. He literally identified deficits. His late wife was focused on nursing, so they fund nursing schools and programs because we have absolute scarcity there.

Considering the arc of AI, I believe in abundance and scarcity dynamics. What’s abundant will be machines helping with intellectual tasks — white-collar jobs will be hit while blue-collar jobs will surprisingly be safer than people think. The care aspect of healthcare will be critical.

If you ask what I want my kids doing: I grew up playing extensive video games and watching tons of TV, and I allow them that — not Monday through Thursday, but weekends they can play games, watch TV, engage in pop culture, watch sophisticated content like Fareed Zakaria and Jeopardy.

But I want them fully versed in AI. My 9-year-old is better than my wife at ChatGPT queries and Midjourney prompts for images. It’s creative expression.

Jordan Schneider: That’s a great age for it, right?

Josh Wolfe: Absolutely. My oldest daughter had to do an evolution project in seventh grade by hand — organisms with sharp teeth survive candy rain while weak-toothed ones die. She drew it manually. My current seventh-grader in that same class is using one of our companies, RunwayML, for AI video generation, creating full videos of different organisms in her scenario.

I want them totally versed in AI because ultimately — and you discussed this recently, maybe with Wang — the power isn’t in who has the chips, but who’s using them. Similarly, power isn’t in who has the applications, but who’s using them.

There’s a significant push for the world to use US-driven open-source or closed models rather than China-driven models that approach an asymptote of truth but never discuss Tiananmen Square, Xinjiang, or Uyghurs. We want the Global South influenced by American ideals of truth, Popperian hypothesis, conjecture, and criticism rather than Chinese systems.

But I want my kids using all these tools and understanding them. What will be scarce against all that abundance is human connection. They need to understand people, make eye contact — you’d be amazed how many kids, because of screens, have awkward, almost autistic interactions.

Being able to connect with people, understand them, read Shakespeare — that’s timeless. People change, costumes change, stages change, but human nature hasn’t changed since the Pleistocene African savannah.

That’s what I’d fund philanthropically. Derek mentioned CBT programs for young people earlier. I started a charter school 17 years ago in my native Coney Island, Brooklyn. We began with 90 fifth-graders in the projects. Now we have 1,000 scholars, 200-plus faculty, 100% college acceptance rate for first-generation college students. Eighty percent of families qualify for free and reduced lunch — a euphemism meaning a family of four makes less than $30,000, which is insane.

These families lost the ovarian lottery — the classic John Rawlsian veil of ignorance. These kids are no less intelligent than those born in Greenwich, Connecticut. But there’s no Army recruiting station on Greenwich’s Main Street — there is one on Cropsey and Stillwell in Coney Island. That’s not fair.

Those are worthy targets for philanthropic dollars.

Jordan Schneider: Okay, but let’s start from a $10 billion bucket. What are we talking about here?

Josh Wolfe: Where would I give? I would fund universal CBT for everybody in the country. I don’t know that it needs that much money — it just needs celebration in the country that helps people become more emotionally regulated and be their best version of themselves. It will reduce problems in our criminal justice system. It’ll reduce problems in corporate America. It’ll reduce a lot of problems across the board.

Jordan Schneider: Not to rag on you too much — you give very sophisticated answers to how to invest in the future of science and technology. That was a fine answer. But a lot of people, at a certain point in their life, switch from the sort of answers we discussed in the first 80 minutes to the question I just asked you. I’m curious: do you see the future differently when it comes to philanthropy?

Josh Wolfe: Where we give philanthropically right now reflects things that we prioritize. For me, complexity science through the Santa Fe Institute — brilliant people. I love it. I believe that’s a source of tremendous value. I’ve been part of that for 10 years as a trustee and believe deeply in it.

The charter school movement — I believe deeply in that because it’s a form of civil rights for people. My mom was a public school teacher, so this hits close to home.

Jordan Schneider: Let me try one more time. This is more of a meta question. The sort of investigation that you need to understand how to use philanthropic dollars efficiently and effectively — I’m curious how similar or different you think that is from investing?

Josh Wolfe: The only similarity is finding an amazing social entrepreneur. It’s like when we started the charter school — we basically backed this guy Jacob Newkin, who was starting the school. It’s the same thing. I used to talk publicly about Jacob: he’s the greatest social entrepreneur that I’ve backed by spending my time and money with him.

But we weren’t doing analysis on the market and the unmet need and that kind of thing. It tends to be something personal. For my wife, it’s the Center for Reproductive Rights. She’s on the board there, making sure that women have access to contraception and abortion and autonomy over their bodies. That’s a really important thing to her.

She’s not doing an analysis of where’s the best place to give or whether we should give more. It’s just: Roe v. Wade got overturned. There are women who are going to die in certain states because they can’t get abortions. What can we do about that? She gives a lot of money.

Jordan Schneider: Maybe we’ll close on this topic. If we’re entering a world where science receives less funding — Danny Crichton, who works for Lux Capital, wrote a really interesting piece about this — when the total amount of science that the US funds decreases, there’ll be a little spillover to China and the EU, but we’ll just have less science overall. Beyond CBT, what encouraging developments do you see for science and technology’s future? What basic research do you think people should be funding — the stuff that’s too risky for any venture capitalist to invest in?

Josh Wolfe: This might be controversial, but people should be spending far less money on climate philanthropy. The answer lies in what I call elemental energy and nuclear power. All that money should be redirected toward early-stage science and psychological and behavioral health research, because that will make society better.

The Gates Foundation, Bloomberg Philanthropies, Ford Foundation — these organizations were captured by the climate movement over the past 15 years. Honestly, I don’t know what it’s accomplished. It’s been a colossal waste of money.

Jordan Schneider: Speaking to the culture issue you raised earlier — it’s striking to me. I worked at an organization that was half nonprofit, half research fund, focusing on China and climate. On the climate side, they could get money to fund literally anything.

Josh Wolfe: It’s popular and makes you feel good. You go to a cocktail party and say, “Oh, I’m funding climate research.” Great, you’re doing wonderful work — Al Gore would be really proud. But the money is misplaced.

Jordan Schneider: It’s almost downstream of the culture. Who are the funders and trustees, and what’s popular with them? They’re not scientists conducting expected value calculations on human flourishing or whatever. Not to disparage anyone, but it’s something that resonates with them personally.

Josh Wolfe: Look, Sam Bankman-Fried was the emblematic figure of this, but the effective altruism movement was rational in trying to determine where we can do the most good. They approached it economically, looking at low-probability, high-magnitude events and identifying opportunities where small amounts of money could have significant leverage.

Going back to Conway’s nursing initiative — that’s not popular. People don’t get excited about addressing the massive nursing shortage. But he identified this as crucial, and they’re putting several billion dollars behind it. That’s noble work.

Bloomberg’s urban initiatives and charter school funding are excellent. People funding the arts because of personal passion — that’s great too. But we have massive problems with criminal justice reform and behavioral health domestically.

I’m not talking about everyone needing mental health days, but implementing cognitive behavioral therapy in schools at a young age. Before children’s prefrontal cortex develops at 25, we could help them develop better self-regulation. The world would be a much better place.

It would be tremendous to see philanthropists return to funding institutions like the old Cavendish Laboratory — putting billions of dollars into institutes that enable knowledge discovery.

Jordan Schneider: Rockefeller University — that’s incredible work. People need to get with the program.

Josh Wolfe: Carnegie Mellon, exactly. These institutions started with robber barons who decided to redirect money into academic institutions. This will happen again.

If you examine philanthropic funding historically, it began with private individuals, then government labs like Los Alamos, followed by Bell Labs (born from monopoly), then IBM Research. IBM centralized initially, then distributed with locations in Zurich and Almaden. Then came Google, Intel, and Microsoft Research.

Some of these corporate labs are under pressure now because they haven’t yielded significant results. But we’ll see the rise of private labs. You can see this already with the ARC Institute.

Jordan Schneider: Absolutely.

Josh Wolfe: The Collison brothers are major supporters there. The Chan Zuckerberg Initiative represents major scientific initiatives comparable to Howard Hughes — himself a former defense contractor who invested enormous amounts into what became the Janelia campus, now one of the great sources of Nobel Prize winners.

Jordan Schneider: That’s a nice place to close. 50% of Howard Hughes researchers came to the US on visas! Joshua, thanks so much for being part of this.

Josh Wolfe: Great to be with you, man.

Jordan Schneider: Awesome. We’ll do this again in 10 years, and then you’ll be saying, “Here’s this philanthropic vision — look at all these molecules I found."

Josh Wolfe: I believe mostly in free enterprise, science, and technology. Early-stage ventures will handle a lot of that, but not the basic science research. There’s no market for that.

Bonus Riffs on Books

Cool. What are you reading? Anything good? Binge-watching anything inspiring or fun?

Jordan Schneider: Two books. I’m a new parent — I have a 10-month-old at home.

Josh Wolfe: Boy or girl?

Jordan Schneider: Girl. First child.

Josh Wolfe: Wow.

Jordan Schneider: I read all these books on paternity leave about “How I Raised My Child in X Country.” They were not good. But one was excellent: Italian Education. It’s by a cranky British guy who married an Italian woman and raised his kids in late 1980s-90s Verona. All these other books are basically backhanded critiques of American parents from whatever direction. But he’s just observing this really interesting, weird society where you have very tight connections between parents and kids — for better and worse, from my perspective, across many dimensions.

It was engaging, funny, smart, with nice vignettes. Each chapter stands on its own, which is good for 3 AM reading.

Book number two — A Social History of the Machine Gun by John Ellis.

The book tells the story of the machine gun through different lenses. He’s a military historian who wrote about World War I and World War II tactics. But in this book, he explored the technological evolution from the Maxim gun all the way through World War II — who invented it, why, and where it came from.

There’s a fascinating acquisition story too, because people didn’t think it was real and didn’t want to buy it. There were prototypes but no factory yet, so you had these hucksters trying to —

Josh Wolfe: What years was this?

Jordan Schneider: At the start of World War I, the British had one gun for every 2,000 people. You had to get to 1916-1917 for them to actually be making and buying enough machine guns.

Josh Wolfe: That’s amazing.

Jordan Schneider: It took an enormous amount of time. The technology was already there in the 1880s and 1890s. There were examples from wars in different countries — the 1905 Russo-Japanese War, the Crimean War. You could see it if you were looking at the right things.

Ellis gives examples of smart colonels saying, “Guys, we need to buy these guns — they’re a big deal.” But people said no, and they all said no way too late.

The other story he tells is about the psychology of not just the people buying the guns, but the officers themselves who had to abandon their mindset about what made a successful officer. Being a sharpshooter wasn’t considered honorable. What won wars throughout the 19th century was discipline, standing in line without fear, marching together to bring maximum power. That’s what the technological paradigm demanded — willingness to maintain rank.

It took over 100 years for people to change their mindsets and understand that you actually need to be distributed, use natural cover on the ground, and get away from the Napoleonic mindset of gallant charges. Those charges were the correct evolutionary answer in a different time period, but not by the US Civil War, definitely not by the end of the 19th century, and absolutely not by World War I.

Josh Wolfe: It’s interesting — the juxtaposition of the two books. One is arguably about technology of life (all parenting is a form of life technology), and the other is about technology of death. It’s a nice contrast in what you’re reading.

Quick parenting observations — First, if you walk into a bookstore like Barnes & Noble — which really don’t exist anymore —

Jordan Schneider: There’s one three blocks away!

Josh Wolfe: The mere existence of a handful doesn’t change that they’ve largely disappeared. There used to be 20 in New York City; now there are one, two, or three.

My point is, when you go into any section — investing, relationships, or parenting — and see 200 books, it means nobody has any idea what they’re doing. If they did, there would be one book with all the answers.

What I learned, especially with our first child, is that you develop a whole bunch of lessons, then you have a second child and they’re all wrong. Once you have more than one child, the nature versus nurture debate is settled. They are genetically different from day one — their predispositions, attitudes, sleep patterns, crying patterns, wants and needs. Their personalities persist from birth. The one who’s more reactive, the one who’s more smiley — it’s absolutely fascinating.

Second point on parenting, which relates to China — when I was growing up, my mother said, “You need to learn golf and Japanese because that’s the lingua franca of business in the ’80s.” Then it became, “You need to learn Mandarin and coding because that’s the lingua franca."

Neither of those things really matter today. The pace of AI development means coding can now be done by AI agents. Computer scientists who thought they were in a valued position are suddenly thinking, “Oh my God, I’m being replaced by agents."

Translation is pretty incredible now too. I don’t know what the next parental trope will be, but it’s usually wrong.

Subscribe now

Jordan Schneider: We can take it back to the machine guns, right? What was successful for you as you were coming up in the world and in the institutions that shaped you isn’t necessarily going to be the thing for the next era. The humility to understand that — both from a defense acquisitions perspective and a parenting perspective — is really hard.

Josh Wolfe: I’ll give you an investing version of that. First, the most dangerous words everyone always says are “this time is different” — because it never is. If you know Shakespeare, then of course it’s never different. Human nature is constant.

But when parents utter certain words, it’s predictive — similar to the defense acquisitions issue or for investing. Want to know what will be the next $10 billion industry? Here it is: “It will rot your brain.” Every time a parent says “it will rot your brain” about something they don’t want their kids doing, that thing becomes the next massive industry.

Rock and roll in the ’50s, TV in the ’60s and ’70s, chat rooms in the ’80s and ’90s, video games in the ’90s and 2000s — every one of those things that was the target of parental ire became the next $10 billion industry. Tipper Gore with parental advisory lyrics and rap music — rap became the biggest genre of music in the following decade.

Just listen to what parents are terrified about right now. The gamers became our modern robotic surgeons and drone pilots. Whatever they’re freaking out about now — maybe TikTok (though I have problems with that for different reasons) — it could be movie-making on social media or whatever.

Josh tries to defend short video

Jordan Schneider: This is the hardest question of the day — give me the optimistic short video take.

Josh Wolfe: We’ve democratized the ability to have creative expression with special effects that used to cost $50 million. Filmmakers used to be siloed in studios with hierarchies, casting couches, and Harvey Weinsteins — awful people. Now there’s freedom of expression where people can create tragedy, drama, comedy, and surrealist content with these tools at their fingertips.

My 9- and 12-year-olds are better filmmakers than I was at 25. They have tools today that Hollywood executives used for Terminator or The Abyss — remember James Cameron’s special effects that seemed amazing back then? That’s from a creator perspective.

Jordan Schneider: What about from a consumption perspective? Having culture delivered in 30-second chunks?

Josh Wolfe: Again, look at it from abundance and scarcity perspectives. I 100% agree that if you’re constantly being trained for short attention spans, that’s problematic. We literally practice patience as a family because I know there are so many competing things offering fast dopamine hits and quick responses. We do long periods of quiet reading from physical books as a family. We watch long movies instead of 30-minute segments.

Jordan Schneider: No, but you’re telling me the things parents are scared of are actually going to be the future. What’s positive about consuming content in 30 seconds?

Josh Wolfe: I’d argue that your ability to process multimodal information is far better. Look at the average older person right now — they’re focused on one thing, they’re slow. You’re probably able to switch between a WhatsApp chat group, Twitter, watching a short video, and checking emails. Your ability to multitask while retaining the ability to function is super valuable.

Jordan Schneider: Your heart’s not in this one…

Josh Wolfe: It’s okay, but here’s what I know: it’s not as bad as people think. The TikTok stuff is concerning, but short-form video generally isn’t bad.

Jordan Schneider: Okay.

Josh Wolfe: I’m optimistic on the science and technology piece. I’ll still be skeptical about the human nature piece, but optimistic about science and technology.

Jordan Schneider: All right, let’s call it there then. This was really fun. Do you have a book to shout out?

Josh Wolfe: Let’s close on fiction and nonfiction. For nonfiction: The biggest debate my wife and I have had was over a book by Robert Sapolsky, a Stanford primatologist and neuroscientist who’s written a series of books. His first one was Why Zebras Don't Get Ulcers: The Acclaimed Guide to Stress, Stress-Related Diseases, and Coping. The punchline was that zebras run for their lives in their ancestral environment, then they’re calm. We have constant stressors all day long that didn’t exist in our ancestral environment.

But his more recent book is called Determined. Either you’ll want to throw it across the room, or you’ll want to send copies to everybody. It depends on whether you agree — as I do — that we do not have free will, or disagree — as my wife does — that we are filled with agency and of course have free will.

Danny Kahneman was a friend. His belief before he died was that neither free will nor consciousness exist — that they are both useful illusions. I very much subscribe to that view.

For fiction, Amor Towles has a short story collection called Table for Two. He wrote Rules of Civility, which is great. There’s a character from the second half of that book who gets extended treatment in a few more chapters in this collection.

But there’s one story that deeply touched me called “The Bootlegger.” It’s a short story that takes place in New York. The beauty of this, for me personally, was that I happened to post on Twitter about my love for this particular story. Amor Towles replies, which was very meaningful, and he says, “This is probably the most autobiographical story I ever wrote."

It’s a relatively short story set here in New York. A young couple, something happens, they go to Carnegie Hall, and this story unfolds. There’s a particular classical musician referenced in this fictional piece, and he replies to the story because he’s actually in it. It was this surreal fictional story that was a slight roman à clef of Towles’s life. He replies to me on Twitter, then the classical musician does too. I was like, “This is amazing."

Jordan Schneider: That’s really fun.

Josh Wolfe: I highly recommend Table for Two by Amor Towles. It’s a set of vignettes around two characters — probably 10 or 12 short stories. But “The Bootlegger” is awesome.

Jordan Schneider: Thanks so much.

Josh Wolfe: Great to be with you.

In 2011, China’s drug regulator cleared the nation’s first home-grown targeted cancer pill. Fourteen years later, a Chinese bispecific antibody is aiming to knock the world’s top-selling oncology drug off its perch.

Those two bookends frame twin contests now running in parallel. One is humanity’s decades-long fight against cancer, a disease that still claims one in six lives worldwide. The other is China’s effort to move from importing medicines to inventing them. This article seeks to follow these battles in real time by tracing the stories of four milestone therapies and their makers.

Founded in 2003 in Hangzhou by a team of returnee talent, Betta Pharmaceuticals launched the first targeted anticancer drug developed in China to reach the market. The success of its drug Conmana1 in China dovetailed with sweeping policy reforms aimed at making home-grown medicines more trusted, innovative, and affordable. The drug did not, however, make it outside China, despite Betta Pharma’s best efforts. Conmana is an example of “me too” innovation: a variation on an existing drug that performs just as well but not quite better enough to make it globally competitive.

Created through a rigorous internal drug development program in 2012, BeiGene’s Brukinsa did what Betta’s Conmana could not – it went global, becoming the first Chinese cancer therapy approved by the US FDA. A “me better” innovation targeting blood cancer, Brukinsa now reaches people in over 65 countries, bringing in over US$2 billion in sales annually.

Betta’s Conmana and BeiGene’s Brukinsa are both small-molecule drugs, meaning they are created through chemical synthesis. The real frontier of innovation that excites biotechnologists is biologic drugs, biomolecules such as engineered proteins or RNAs that act with greater power and precision inside the body.

Carvykti is one of the first innovative biologic drugs created in China. Initially discovered and tested in Xi’an Jiaotong Hospital by Nanjing Legend Biotech, the drug reached global markets with the help of Johnson & Johnson. First gaining US FDA approval in 2022, Carvykti now has regulatory approval across over 36 countries and has treated over 5,000 patients, with more to come.

The last story is unfinished. It’s about Akeso’s ivonescimab, an icon of Chinese biotech innovation in mainstream media. A biologic drug with a novel method of targeting cancer, ivonescimab received approval by the Chinese NMPA in April 2025. In countries like the US, where Akeso has passed the baton to US-based Summit Therapeutics to develop the drug, ivonescimab is still in the clinical trial phase, meaning it has yet to pass through the regulatory gauntlet. What’s exciting and undecided about ivonescimab is its potential to go head-to-head with the world’s best-selling drug, Keytruda. Whatever happens to ivonescimab over the next few months in trials outside of China will send a signal of exactly how successful Chinese biotech innovation has become. The world of biotech will be watching closely.

But let’s start from the beginning. Betta Pharma’s milestone achievement of Chinese regulatory approval for a new innovative drug (2011) happened only 11 years before Akeso’s ivonescimab earned a US$5 billion deal with Summit Therapeutics (2022). How did we get here?

Betta Pharma: The “Me-too” Era

In the 1990s, biopharmaceuticals — specifically oncology, the study and treatment of cancer — entered a new era of innovation. Up until the early 2000s, doctors primarily combated cancer with broad-stroke methods like surgery and chemotherapy. Now, new and improved methods were emerging: targeted therapies that zeroed in on cancer cells while minimizing damage to healthy cells and immunotherapies that helped the body’s immune system recognize and attack cancer cells.2

This revolutionary effect of molecular biology captured the attention of many bright scientists and doctors in China, including Dr. Wang Yinxiang 王印祥. Born in rural Hubei, he spent three years working in public health and three years completing a Master’s degree at the Chinese Academy of Medicine before he could truly follow his passion for oncology to the United States, where he earned a doctorate from the University of Arkansas.

Dr. Wang got his wish to do cutting-edge research as a postdoc at Yale, where he dove into one of the first targeted cancer therapies, Novartis’ Gleevec. Sharing his apartment was Ding Lieming 丁列明 – another Chinese transplant with a University of Arkansas MD. On strolls through New Haven’s Science Park, the two friends along with medical chemist and entrepreneur Zhang Xiaodong 张晓东 bonded over more than just science. They shared a bigger dream: to bring the newest in biotech to China.

Dr. Wang and Dr. Ding would ultimately join forces in 2003, when they founded Betta Pharmaceuticals to develop targeted cancer therapies in China. Betta opened its doors with a shoestring team – just 13 people, many of them what Dr. Wang affectionately called “kids,” fresh from bachelor’s or master’s program and learning on the fly. Nevertheless, they managed to develop icotinib (later sold as Conmana), a drug engineered to target EGFR proteins as a way to inhibit cancer cell growth.

In those years, China’s pharmaceutical industry hadn’t left the nest. Manufacturing of cheap, generic drugs dominated. To domestic investors, companies, and physicians, a business model built on new drug development was unthinkable: the costs and risks were too high, the regulatory process was a mess, quality and safety were still iffy, and previous such attempts had failed. Winning a clinical-trial slot for Conmana (a prerequisite for proving the drug could outshine current care) was nothing short of herculean. When the Peking Union Hospital director declared the study too risky and tried to dismiss him, Dr. Wang stood firm for ninety minutes, knocking down every objection until the approval stamp finally came down in his favor.

In 2009, Conmana made it to Phase III clinical trials, the make-or-break test of wide-scale efficacy. In Phase III, the team pushed the envelope again: rather than testing against a placebo, they pitted their molecule against AstraZeneca’s gefitinib (the world’s first targeted anti-cancer therapy) in the first Chinese study to challenge an imported standard head-to-head.

The study’s results, announced by leading academic Sun Yan at the 2011 World Lung Cancer Conference – also the first time a China-developed drug headlined an international academic forum – showed that icotinib could match the cancer-fighting power of the imported benchmark while causing fewer side -effects. Conmana, in other words, was a successful “me-too” drug, an incremental improvement on an existing pharmaceutical innovation.

After six grueling years, Conmana earned its first regulatory approval from China’s National Medical Products Administration (NMPA; although at that time it was still the CFDA).

Betta’s success as a fast follower of a next-generation cancer therapy was a triumph for Beijing’s returnee talent and national innovation programs. Conmana’s success had been fueled by funds from the Yuhang District Government of Hangzhou, the “863” Program, and the “11th Five-Year Plan” National Science and Technology Major New Drug Special Project.

The government showered Betta with accolades: the China Overseas Chinese Contribution Award, the gold prize for patents, first prize for the National Science and Technology Progress Award, and more. Chen Zhu, then Minister of Health, praised their achievement as “an emblem of ‘Two Bombs and One Satellite 两弹一星’ in the field of public health,” referencing a techno-nationalist ideal of a whole-nation project for science and technology development.3

However, for all its homegrown glory, Betta’s blockbuster never crossed the border.

In 2014, with the help of Roswell Park Cancer Institute, Betta filed to run a clinical trial for Conmana in the US, the first step towards seeking US FDA approval. However, the trial was shortly withdrawn. Newer, later-generation EGFR inhibitors were eclipsing Conmana’s performance, and there was no sense investing in trials when the product was unlikely to sell.

Though just a “me-too” innovation, Conmana was a landmark accomplishment for China. On the domestic market, the drug was meaningfully cheaper than its imported alternatives. Given the rapidly growing incidence of lung cancer in China, Conmana’s improved affordability made a real difference in patients’ lives. Still, it would take almost another decade before a China-developed cancer drug would make a global impact.

BeiGene: A “Me-Better” Drug Goes Global

China’s mix of capital, talent, and policy reforms was turning its budding biopharma sector into a global magnet.

Among the first drawn in was Pittsburgh native John Oyler. Familiar with China through his work at McKinsey in the 1990s, Oyler was stunned by the country’s science and technology progress when he returned the next decade. By the mid-2000s, regulatory harmonization, returning talent, and improved manufacturing infrastructure enabled China to meet the needs of global pharmaceutical companies, leading to the growth of contract research organizations (CROs), which provide outsourced medicinal science services. Oyler co-founded one such Chinese CRO, BioDuro, in 2005.

But serving foreign pharma clients wasn’t enough. As China moved toward deeper healthcare reform, Oyler saw an opening for homegrown innovation: “[China] had the capability to pour tens of billions of dollars back into the global industry to help pay for more research, which would not only make drugs more affordable in China, but across the globe.” Rather than repeat the trajectory of BioDuro, which was eventually sold, he wanted to create something enduring. “I wanted to build something here — in China — that is lasting, impactful, involved in great science, and can really help people,” he said. He aspired to create a company capable of developing world-class cancer therapies from a country many still underestimated.

To bring his vision to reality, Oyler needed scientists. He connected with Dr. Xiaodong Wang 王晓东, a top Chinese American academic biologist who recently returned to China to lead the new Beijing Institute of Life Sciences. The chance to work with Dr. Wang — a superstar scientist admired widely enough to impress parents at Chinese New Year — proved an effective tool for attracting talent. Together, they founded BeiGene in Beijing in 2010 to become the “Genentech of China.”

Early on, BeiGene focused on BTK inhibitors, a targeted cancer therapy that works by blocking cancerous B-cells’ ability to grow. The first BTK inhibitor, synthesized in 2007, showed promise but caused significant side effects. In 2012, BeiGene initiated a discovery program in San Mateo and Shanghai to develop a better BTK inhibitor. After screening over 3,000 compounds, the team identified the highest-potential molecule that would eventually become Brukinsa (zanubrutinib).

BeiGene aimed to take Brukinsa global from day one. To support worldwide approvals, the company built a 25-country trial program in which approximately 90% of patients were enrolled outside of China. The numbers delivered: Brukinsa consistently beat first-generation BTK inhibitors on safety and efficacy, turning a presumed “me-too” into a clear “me-better” that is now the standard of care for B-cell cancers.

Momentum snowballed. In 2019, Brukinsa set a precedent as the first Chinese-developed cancer therapy to win FDA approval, months before China’s own NMPA signed off. It has since secured clearances in 65-plus markets spanning the US, EU, Canada, Australia, Japan, and China, and now supplies more than half of BeiGene’s revenue with US$2.6 billion in 2024 sales. The company (rebranding as BeOne) has likewise gone global, conducting trials in over 45 countries.

Legend: A True Chinese Biologic

Witnessing the breakthrough of firms like BeiGene, Beijing set its sights on higher-value innovation. The State Council’s 2016 13th Five-Year Plan therefore called for “leapfrog development in the biopharmaceutical industry,” spotlighting cell and gene therapies, antibodies, and vaccines.

These platforms fall under biologics — large molecules derived from biological processes such as insulin and hormones — rather than small-molecule drugs such as Betta’s Conmana and BeiGene’s Brukinsa. Because biologics are bigger and more structurally complex, their effects are harder to predict and their manufacture far costlier, but they open therapeutic doors that chemistry alone cannot.

Founded in 2014 as a subsidiary of GenScript, Legend Biotech embodied the kind of biologics leadership the state now prioritized.4 After losing his father to cancer, co-founder Frank Zhang 章方良 united with Chief Scientific Officer Dr. Frank Fan to create the firm with the goal of advancing oncology.5

At that time, influential scientific journals, Big Pharma deals, and first-in-human successes had converged to position cell and gene therapies as the vanguard of biopharmaceutical innovation. Early successes led by the University of Pennsylvania highlighted the potential of CAR-T therapy,6 a type of treatment in which a patient’s disease-fighting T-cells are genetically engineered to seek and destroy cancerous cells.

Driven by the promise of next-gen cancer therapy, Legend’s 19-person team, working in “a room the size of a freight elevator,” crafted a second-generation CAR-T treatment targeting multiple-myeloma tumor cells (later sold as Carvykti). Leading the research was Dr. Frank Fan, who had studied at Xi’an Jiaotong University and worked at the Xi’an Jiaotong Hospital. Leveraging those ties, Dr. Fan was quickly able to initiate Legend’s first CAR-T clinical trial at Xi’an Jiaotong Hospital, turning the startup into a “dark horse” contender in the CAR-T space.

At the 2017 annual meeting of the American Society of Clinical Oncology (ASCO), Legend presented early-phase data from its trials in China: durable remissions in multiple-myeloma patients with only mild side effects. The first of its kind accepted for review by China’s NMPA, this CAR-T candidate signaled that the country could innovate well beyond small-molecule chemistry.

Such promising results drew members of the global life sciences community. J&J’s Janssen signed a partnership with Legend Biotech in 2017. This was a case of an outlicensing deal: when a company (such as Legend) sells or grants rights to a drug candidate, passing the baton to a different company (such as Janssen), which who takes on the responsibility to bring the drug through testing, approval, manufacturing, and commercialization. Such transactions are the lifeblood of the industry. However, the high-potential molecules worth such high-profile deals didn’t historically come from China.

With the combined efforts of Legend and Janssen, Carvykti won FDA approval in 2022, soon followed by clearances in the EU, UK, Japan, and Canada. Though not first-to-market, Carvykti’s superior clinical value crowned it best-in-class. This then completed China’s rapid climb from “me-too” copies, through Brukinsa’s “me-better” gains, to a world-leading biologic breakthrough.

Akeso: a top challenger emerges

Now, Akeso Biopharma’s new molecule is drawing notice as a likely first- and best-in-class therapy from China.

Akeso started as the dream of Dr. Michelle Xia, a Gansu native. While working for California-based Crown Bioscience and in other roles in the US and UK, Dr. Xia grew frustrated with the eight-to-ten-year delay it took for innovative overseas therapies to reach Chinese patients. So she and three partners founded Akeso in 2012, naming it after the Greek goddess of healing, with a mission to develop home-grown therapies for cancer and autoimmune diseases.

Akeso’s edge is bispecific antibodies (BsAb), a type of next-generation cancer treatment involving engineered proteins that strike two targets at once, such as igniting immune cells while starving tumors. The first proof arrived in 2022 when China’s NMPA cleared Akeso’s Kaitani (PD-1/CTLA-4), the world’s first commercialized BsAb.

Next came ivonescimab, a PD-1/VEGF bispecific now pushing Akeso onto the global stage. By jointly blocking an immune checkpoint and tumor blood-vessel growth, it qualifies as first-in-class – an industry term for a drug that introduces a truly novel therapeutic approach.

In June 2022, Akeso unveiled the Phase II results of ivonescimab at the annual ASCO conference, showing strong responses in non-small cell lung cancer. Just months later China’s NMPA granted the drug Breakthrough Therapy status for three medical use cases, enabling closer guidance and fast-tracking its review.

Sensing ivonescimab’s scientific and commercial potential, US-based Summit Therapeutics inked a massive deal in December, licensing Akeso’s innovation for up to US$5 billion.7 The corresponding press release hailed Akeso’s innovation as “the PD-1 / VEGF bispecific antibody that is most advanced in the clinic,” noting that neither the FDA nor EMA had yet approved any PD-1-based bispecific therapy.

The size and significance of this agreement marked a bellwether moment in Chinese biotech innovation. Blue-chip investors and multinationals began scouting the country for genuinely novel assets rather than low-cost manufacturing plays.

China’s slice of global out-licensing has since tripled to 12%, with deal value leaping from US$35 billion in 2023 to more than US$46 billion in 2024. This trend seems set to continue through 2025 and beyond.

The fate of ivonescimab and most other compounds covered by these recent deals is uncertain. Many drug candidates are purchased in the preclinical or Phase I stage of development, requiring another 5+ years and US$300+ million dollars before they pass the clinical and regulatory hurdles necessary to make it to market – or, they will fail, like roughly 90% of compounds that enter human trials. Only five China-originated drugs have ever cleared the US FDA (BeiGene’s Brukinsa and Legend’s Carvykti are two of them).

Ivonescimab could be next. Akeso’s molecule has already completed Phase I and Phase II, and is in the midst of several high-stakes Phase III trials. Its Phase III trials with Chinese patients have already demonstrated success, leading to two recent NMPA approvals in the spring of 2025. The defining test comes next: a Summit Therapeutics-run Phase III study spanning 108 locations in 12 nations, where the drug candidate must outshine oncology’s gold standard, Keytruda.

Keytruda (drug name pembrolizumab) has been described as “era-defining,” the “800-pound gorilla” of the class of drugs to which ivonescimab also belongs. Developed and commercialized by American multinational Merck, the drug has been approved for 41 indications8 across 18 types of cancer. It’s the world’s best-selling drug, raking in about $29.5B in 2024 – nearly half of Merck’s total revenue. Its upcoming 2028 patent expiry opens the field for new challengers.

Akeso’s ivonescimab could be one such challenger. If its global Phase III trial confirms that the drug’s positive risk-reward results can extend beyond China’s borders, Akeso’s first-in-class molecule may eventually also prove to be best-in-class, making it one of the biggest biotech stories of the decade.

On the horizon

Together, Conmana, Brukinsa, Carvykti, and ivonescimab trace a clear, but incomplete, arc: China’s pharma sector has evolved from reverse-engineering proven ideas to originating drugs that can contend for global standards. Each milestone marks a step from “good enough at home” to “competitive abroad,” showing how policy shifts, capital inflows, and returning talent have reshaped the industry’s ambitions and capabilities.

Those four successes are only a sliver of the story. Dozens of other firms have logged incremental wins, and many more have stumbled in clinical trials or overseas filings. With financing tightening, patent cliffs approaching, and regulatory expectations rising, the next crop of candidates will test whether China’s momentum is structural or situational. The outlook could range from steady gains in select niches to a broader slowdown if capital or policy tailwinds fade.

Whatever the trajectory, one fact persists: cancer is indifferent to where a molecule is conceived. Progress depends on tapping every credible lab and idea, whether in Boston, Basel, or Beijing. If Chinese innovators add new options to the world’s oncology toolkit, patients everywhere stand to benefit — and that, ultimately, is the benchmark that matters.

2

See this chart:

6

See this chart:

Once hailed as the “Apple of China,” Xiaomi has evolved into the world’s No. 3 smartphone brand. The Beijing giant wires millions of homes with its TVs, wearables, and appliances while pitching itself as a seamless “Human × Car × Home” platform.

Founder Lei Jun now wants to turn that consumer-electronics prowess into hard-tech dominance. In 2025 alone, Xiaomi debuted its first in-house chip (the XRING O1) and showed off its sleek SU7 electric sedan. Yet the scorecard is mixed: the Robotics Lab has been folded into the auto unit, the SU7’s shine dimmed after a fatal crash, and new U.S. export curbs threaten the flagship chip program just as it launches. Can a smartphone playbook built on rapid iterations and razor-thin margins really scale to cars, robots, and semiconductors?

In this piece, we let you hear Lei Jun in his own words, through two translated excerpts. One brims with sky-high ambition, while the other is steeped in damage control. Read them back-to-back and watch how Xiaomi’s grand vision collides with hard truths of physics, finance, and public accountability.

From budget phones to bold hardware

Chinese media celebrate Lei Jun as one of the country’s signature tech entrepreneurs. After graduating from Wuhan University and weathering several early start-up misfires, he assembled seven co-founders in 2010 around a crowdsourced Android ROM that grew into Xiaomi. The early strategy was straightforward: sell affordable, capable smartphones. That formula made Xiaomi a household name in phones and Internet of Things gadgets, but the 2016 debut of the bezel-less Mi MIX concept signaled a pivot from “internet hardware” to true hard tech. In the translated excerpts below, Lei Jun traces that evolution and lays out why he believes Xiaomi’s future rests on chips, cars, and robots.

The following excerpt is an original ChinaTalk translation of a December 2023 interview between Lei Jun and reporter Wang Ning, broadcast on CCTV.

Original link | archived

Lei Jun: Let me start with some Wuhan University memories. Thirty-six years ago, in 1987, I was admitted to its Computer Science department. Back then I lived almost entirely on scholarships and professors’ projects; those funds literally carried me through four years of college. At one award ceremony I received a sizable scholarship; standing on that stage I vowed that if I ever could, I would repay the university a hundred, a thousand — ten-thousand — fold.

Lei Jun: My life-long dream was sparked in the Wuhan University library when I read Fire in the Valley, the story of Steve Jobs and early Silicon Valley. In the late ’80s, Jobs was my very first role model. But the more I understood Jobs, the clearer it became that I’m not him. Still, that doesn’t stop us from charting our own path and creating products that are different in their own right.

Angela: After graduating in 1991, Lei joined Kingsoft, a Chinese software company, where he progressed from software engineer to executive. Competing with Microsoft, Kingsoft faced countless problems including product flops and near bankruptcy. He then went on to found an online bookstore, Joyo.com, which was sold to Amazon in 2004.

Lei’s experiences with such missteps and missed timing during the early era of the internet led to his coining of the phrase, “Catch the right tail-wind and even a pig will fly” 站在风口上，猪也能飞起来, which he proceeds to explain in the interview.

Reporter Wang Ning: Your famous tail-wind principle is still quoted all the time.

Lei Jun: Back in my engineer-founder days we green-lit projects simply because they interested me or a few users asked for them; we rarely stopped to ask whether the timing offered explosive growth, so the companies survived but seldom scaled. That sort of company is hard to kill, but pushing it to greatness is equally hard.

That’s why I formulated the tail-wind principle. Don’t just keep your head down pulling the cart; look up and see where the wind is blowing. It was a major course-correction in my own playbook.

Reporter Wang Ning: What was the real test for you during that period?

Lei Jun: We had clearly missed the web’s prime window, so we kept asking when the next breakout would come; four or five years early we bet on mobile internet, but didn’t know when it would ignite—once the iPhone and Android hit, I knew the fuse was lit.

On Xiaomi’s “hard tech” pivot:

Lei Jun: Hard tech means chips, smart manufacturing, robotics, operating systems — those core layers.

Reporter Wang Ning: You keep hammering on hard tech.

Lei Jun: I do.

Reporter Wang Ning: Why is your drive to battle it out on hard tech so intense?

Lei Jun: We’re top-three worldwide, but we face giants: Apple, Samsung, Huawei. Without breakthroughs in core tech you’ll never build a moat or stand shoulder-to-shoulder with those titans. If Xiaomi dreams of being world-class, hard tech is non-negotiable.

Narrator: In 2020, Xiaomi’s tenth anniversary, it set a ten-year goal to become a next-generation hard-tech leader and deliver premium products.

Lei Jun: Every flagship product applies the newest, most advanced tech. Three years ago, I told the team we would benchmark against the iPhone across the board. This statement caused controversy. Were we just piggy-backing off of Apple’s fame to grab attention? But we must unsheathe the sword: set our own targets, dare to compare. Even if we lose all 100 metrics, daring to measure is step one. Can our phones really win? I had to persuade people over one-by-one that going premium takes patience, since earning user trust is a process.

Lei Jun: We started exploring operating systems from the very first days of the company. Seven years ago we resolved to create Xiaomi HyperOS, building the entire stack from the kernel up. Over 5,000 engineers have contributed, and we’ve just released the first version. Our goal is a human-centred, closed-loop platform that connects personal devices, the smart home, and the car.

Reporter Wang Ning: You’re working down at the OS layer — deep, technical stuff — yet you keep calling it ‘human-centred.’ What exactly does that mean in practice?

Lei Jun: AI is moving fast, so we’re baking the most advanced AI into HyperOS to give it a clear edge in intelligence over any rival platform. More importantly, that same intelligence has to knit together the entire “Human × Car × Home” ecosystem: people as in personal devices, cars as in a mobile smart space, and home as in smart appliances.

On electric vehicle strategy:

Reporter Wang Ning: Elon Musk called Xiaomi’s entry into the automotive industry “interesting competition.” Your response?

Lei Jun: Smart EVs merge auto and consumer electronics. One car CEO joked, “A smart EV is just an oversized smartphone on four wheels.” Not strictly true, but that shows the convergence. So yes, it’s challenging, but the difficulty is controllable.

Reporter Wang Ning: What do you mean by “difficulty is controllable”?

Lei Jun: Three years ago I still thought building a car was daunting, so I approached it with real humility. Our user research showed that when people hear ‘Xiaomi car’ they expect technology and an ecosystem. That led me to a guiding principle I call “nail the fundamentals, then amaze 守正出奇.” In practice it means: respect the hard rules of the auto business — quality, safety, manufacturing discipline, make sure the very first model is rock-solid, and only then layer on top the Xiaomi-style innovation that surprises the market.

Our second rule is ‘10-X investment.’ A typical automaker puts three to four hundred engineers and maybe one or two billion yuan into a new model; many cars you see are built on that. For our first car we assigned 3,400 engineers and spent more than ten billion yuan — over ten times the norm. With that level of commitment, I’m aiming to win.

Still, cars are complex. I worry about two opposite risks: the launch flops and nobody buys, or demand goes crazy and people wait a year and flame us for delays. Either way, there’s plenty of reasons for anxiety.

On Xiaomi's future amidst geopolitical uncertainty:

Reporter Wang Ning: With global turbulence and domestic headwinds, why double down on hard tech and stay confident?

Lei Jun: The tougher the climate, the more we should invest in technology. When others pull back, breakthroughs actually come easier, and we’re ready the moment the market rebounds. In fact, Xiaomi already returned to positive growth last quarter.

Why do we have the nerve to keep spending? Because in every field we play in there is still plenty of unmet demand; I believe Xiaomi is only at the starting line. Our approach is two-pronged: deepen our roots in China’s domestic market while pushing ahead with globalization. That’s the road we’ve chosen, and it’s still rich with opportunity.

Reporter Wang Ning: People say you don’t need to change since things are great. Yet you keep leaving your comfort zone. Why?

Lei Jun: At the core is the very high bar we’ve set for ourselves: to become a great technology company that lets everyone on the planet enjoy the benefits of innovation. The goal is so lofty that no matter how hard we jump, we still can’t quite reach it — yet that stretch is exactly what pulls the whole company forward. Xiaomi’s relentless evolution is powered by that dream.

Hard Tech, Hard Lessons

[Angela writing] 2021 marked Xiaomi’s leap into one dimension of hard tech, robotics, with the release of CyberDog, an open-source robot companion. A year later came CyberOne, a full-sized humanoid, and in 2023 the sleeker CyberDog 2. Commercially and technically, none hit the mark: CyberOne never reached mass production, and CyberDog 2 sales ran at a loss as buyers complained about its limited abilities. Attrition was another problem: high-profile engineers departed, with Liu Fang 刘方 (former autonomous-driving head) leaving to start his own robot firm and humanoid specialist Ren Zeyu 任赜宇 moving to ByteDance. By mid-2024 the Robotics Lab had been folded into Xiaomi’s auto division, its once-lofty ambitions reduced to building robots that service the company’s own factory lines.

The automotive story began the same year the robots debuted. In early 2021, Lei Jun learned that U.S. sanctions might hit Xiaomi and decided the firm had to diversify beyond phones. Xiaomi poured RMB 10 billion (US $1.5 billion) into an EV program, broke ground on a plant in Beijing’s Economic-Tech Zone, and — true to form — Lei personally test-drove more than 170 cars and earned a racing license, convinced that a top driver should helm a top car company. The gamble paid off fast: the SU7 electric sedan launched in 2024 to a rapturous market reception.

Yet momentum met reality on March 29th, 2025, when a Xiaomi SU7 electric sedan running in Navigate-on-Autopilot mode veered off the Dezhou–Shangrao Expressway in Anhui and slammed into a cement pole, killing three college students. The first fatality involving Xiaomi’s year-old EV led to nationwide uproar. The hashtag #SU7事故 raced onto Weibo’s hot-search list within hours, while Hong Kong–listed Xiaomi shares fell 5.5% the next trading day. The Ministry of Industry and Information Technology banned marketing phrases like “smart driving” or “autonomous driving” and now requires pre-approval for any OTA updates that touch driver-assistance features. The event sparked debate over assisted-driving safety, wiped out new-car orders, and forced the company to publish detailed log data and promise a top-to-bottom safety overhaul.

Xiaomi CEO Lei Jun gave the following speech on May 22nd, 2025, for the company’s 15th anniversary. His address walks a tightrope between contrition and confidence, as he admits to the company’s past mistakes while promising to embrace the societal responsibility that comes with being a major automaker. The crisis shows the high degree of pressure on flagship Chinese tech companies, and how quickly the public, investors, and policymakers can converge when such a firm stumbles.

At the same time, Lei Jun widens the narrative by announcing Xiaomi’s newest chip, the XRING O1. On June 3rd, he predicted Xiaomi’s auto arm would turn a profit in Q2 2025. But will the company be able to live up to its bold ambitions?

The following excerpt is an original ChinaTalk translation of a speech given by Lei Jun on May 22, 2025. Transcript from Zhihu.

This year marks the 15th anniversary of Xiaomi’s founding. Today is a momentous day, and we prepared a series of celebratory events for it. However, at the end of March, a sudden traffic accident shattered all of that. We were hit by a storm of doubts, criticism, and blame. My colleagues and I were stunned.

A veteran of the auto industry told me, “In this business, accidents are bound to happen.” But no one expected the impact of this accident to be so massive, nor the blow to Xiaomi so heavy.

Thinking back to four years ago when we decided to build cars, I was especially worried about safety, so we placed enormous emphasis on vehicle quality and safety. After more than three years of relentless work by our automotive team, the SU7 has been on the road for just over a year — and its build quality has been our pride. In every independent test by the leading authorities, it has received top marks. Yet we never could have foreseen how this traffic accident would make us realize the public’s expectations and demands of us.

To be honest, only now do I recognize that we have always thought of ourselves as newcomers to the auto industry. This incident drove home a simple truth: Xiaomi’s scale, influence, and visibility have grown so much that society now expects us to act like a fully fledged industry leader. We understand, deeply, that after fifteen years this is a vital responsibility we cannot avoid.

So today, what I want to share with you is that 15-year-old Xiaomi is no longer a rookie. In all industries, we don’t have the grace period of true novices. We must hold ourselves to higher standards and goals.

On automotive safety, I want to announce to everyone that we aim to make our car the safest in its class. We will not just comply, nor merely meet the industry level; as a leader in the auto industry, we will guarantee safety that surpasses the industry standard.

This year on April 1st, I said publicly on Weibo that Xiaomi will never shy away from any issue. “Never shy away” means confronting problems head-on, examining ourselves critically, fixing what’s wrong, and committing to continuous improvement. I know accomplishing this is extremely difficult; it requires us to unite in heart and mind and take it with the utmost seriousness.

Over the past few months, I have held countless meetings with the automotive department’s management and team. All of these meetings have centered around one core theme: how do we solve these problems systematically? How do we convincingly show, through stronger operations and governance, that we’re living up to the public’s higher expectations?

This year is Xiaomi’s 15th anniversary. We have canceled many of the celebrations, summaries, and planning activities we had planned. In any case, I believe this is an opportunity for us to seriously review the wins and losses of the past five years.

In my view, the most important thing Xiaomi has done in the past five years is maintaining a strong technology foundation. Five years ago we set a new goal: to become a global leader in next-generation hard tech. Five years ago we made a clear commitment to invest more than 100 billion yuan in R&D over five years and to increase core technology research; to date we have invested about 105 billion. This year alone we expect to invest over 30 billion.

Today, here, I want to share an extremely important piece of news with you: our self-developed smartphone SoC chip, the Xuanjie 玄戒 O1, is expected to be released at the end of the month

After 10 years of making chips, this is Xiaomi’s milestone achievement. It also represents a new starting point for Xiaomi to break through in hard tech. Chipmaking is something the public and Xiaomi fans ardently expect from us; it is also the only path for Xiaomi to move toward becoming a hard-core tech leader. Xiaomi will forge ahead fearlessly.

2019 was a challenging period for us. We faced all kinds of internal and external pressures. At the suppliers’ conference at the end of that year, I told our supply-chain partners this line: “A gale reveals the toughest grass; a long road proves a horse’s strength.” “疾风知劲草，路遥知马力。[Editor’s note: The metaphorical meaning is along the lines of “true resilience is exposed only under pressure and over time.”]. I believe that from then till now, all our partners can clearly see that Xiaomi is much stronger than it was five years ago. As of today, Xiaomi has existed for 15 years. Years of highs, lows, and hard times have already proved just how resilient we are.

When we started, I remember thinking that people would need 15 years to understand and recognize Xiaomi. Today, it seems we were too optimistic — 15 years is not enough. But that’s all right; we will just keep moving forward until the day we have fully proven ourselves.

Xiaomi still has a long way to go before it becomes the strongest, but no one can match the persistence, resilience, and patience that keeps us getting back up every single time.

Xiaomi’s future

[Angela writing again] Will Lei Jun’s everywhere-everything strategy work?

Xiaomi’s smartphone success hinged on a strategy tailored to consumer electronics: rapid iteration, razor-thin margins. Hard tech plays by stricter rules. Robots, cars, and advanced chips demand flawless safety engineering, deep benches of specialist talent, and sustained capital infusions.

Xiaomi has already brushed up against each of these constraints. In the Robotics Lab, key engineers were stretched thin, while the first SU7 sedan shipped with a leaner, lidar-free sensor suite, a cost-saving choice that came under scrutiny after a fatal crash raised questions about perception margins. Complicating matters is the sheer sweep of the company’s ambition: pursuing cars, robots, and advanced chips threatens to disperse capital and engineering bandwidth, potentially leaving every moonshot a little short of the sustained focus that hard tech demands.

Even Apple doesn’t juggle cars, robots, and semiconductors all at once. The Silicon Valley behemoth shelved its car project after billions of R&D and keeps its robotics work firmly behind the curtain. For Xiaomi’s robotics and autos work, leaning into splashy launches and viral publicity for its hard tech projects has proven to be as much a liability as a strength.

Chips might be the exception, if Xiaomi can leverage its vast smartphone footprint into volume for its in-house silicon. The company’s newly announced Xring O1 and US$28 billion R&D pledge hints at genuine momentum. Yet geopolitics looms large: fresh US export curbs on electronic design automation software threaten to inflate costs, stretch design cycles, and slow the very updates those next-generation chips will depend on. Whether Lei Jun can temper showmanship with staying power will decide if Xiaomi’s hard-tech dream becomes a cornerstone of its ecosystem or just another chapter in China’s moonshot boom-and-bust cycle.

Mood Music

I did a podcast with Doug, Dylan and Jon updating our AI Mandate of Heaven tier list and talking about how America needs science research and immigrants (Apple Podcasts, Spotify). If you’re an AI researcher interested in signing onto a short letter in support of the NSF and global STEM talent coming to America, please respond to this email!

Zilan Qian is a fellow at the Oxford China Policy Lab and an MSc student at the Oxford Internet Institute.

In China, U.S.-based large language models like ChatGPT, Claude, or Gemini are technically banned, blocked, or buried under layers of censorship. The Chinese government has only explicitly banned ChatGPT, citing concerns over political content, while other U.S. models like Claude and Gemini are not formally banned but remain inaccessible due to the Great Firewall. U.S. LLM providers also restrict access from China but leave some loopholes: OpenAI blocks API use but Azure continues to serve enterprise clients via offshore data centers; Anthropic blocks access to Claude within China but permits use by Chinese subsidiaries based in supported regions abroad; and Google does not offer the Gemini API in China, but access seems to be still possible via third-parties like Cloudflare (we reached out to Google for a comment but didn’t hear back).

But on Taobao, the country’s largest e-commerce platform, consumers and companies can buy access to these models with just a few clicks. This piece explains how Western models are priced, advertised, bought, and sold in China, and what their popularity reveals about state censorship, platform enforcement, and consumer demand.

Market Overview

How Do Chinese Sellers Price a banned LLM?

Sellers use a range of strategies to appeal to consumers with different priorities — cost, convenience, privacy, and technical needs. Pricing largely depends on the type of access offered — i.e., shared accounts (lower-cost, multi-user access) versus private accounts (higher-cost, single-user access). How users connect to the models also changes the marketing approach. VPN-free access is often marketed with phrases like “domestic direct connection (国内直登)” or “no magic [VPN] needed (无需魔法),” typically referring to proxy sites that offer a simplified, browser-based experience. Other factors influencing price include the model’s subscription tier ( for example, ChatGPT o3 vs. o1), connection stability, and subscription time. Most “domestic connection” or “transferred API” (API purchased through third parties) options are priced significantly below the rates charged by official providers like OpenAI or Anthropic.

For example, the screenshot below shows 27 different Claude pricings, with prices differing based on:

• Subscription time: 1–60 days

• Website access: an account for proxy sites versusan official Anthropic Claude account (with VPN service included)

• Subscription tier: Claude 3.7 Sonnet versus Claude Pro

• API access: “中转 API” (Transfer API) versus “官方API” (API purchased from official anthropic platform)

• Compatibility with SillyTavern, a popular third-party frontend interface (called “酒馆” in Chinese), which enhances user interaction with LLMs.

While an official Claude Pro account costs 199 RMB (27 USD) per month, which is more expensive than Anthropic’s rate at 20 USD, a “direct connection” Claude Pro account from a mirror site is only 65 RMB (9 USD) per month.

Transferred API usually comes from established transfer stations (中转站) for API, usually at a 70-90% discount compared to official prices (e.g., selling 1 USD worth of tokens for 1 RMB). The low prices are made possible by utilizing one or more of the following tactics:

1. Exploiting free API quotas by bulk-registering developer accounts (e.g., Anthropic’s $5 free API);

2. Reselling unused portions of other people’s API quotas;

3. Traffic optimization techniques, such as request aggregation: combining multiple users’ queries into a single batch call to the LLM API to reduce costs

4. Leveraging corporate or educational discounts.

How do Chinese sellers advertise access to U.S. LLMs?

While these tactics help vendors reach potential buyers, they also reveal a surprisingly open market landscape, which raises the question: Who is actually buying and selling these models? Sellers typically advertise and communicate their services using product images that include large brand logos, stylish fonts, and keywords highlighting their service strengths. Direct keyword searches work for most models, like Claude or Gemini. Many listings include the names of specific models directly in their product titles, often alongside a dense string of keywords like “API,” “domestic direct connection,” “official account,” or “exclusive” to boost visibility. The fact that such listings remain easy to find suggests a lax approach to censorship, especially given that many of these terms are potentially sensitive.

The censorship drastically increases when it comes to ChatGPT, which is likely due to the state’s explicit crackdown on this model family, which began in 2023. When selling ChatGPT the model, one mechanism is to market the subscription as a book. As of March 2025, you could still buy “ChatGPT Book: How to Make Money Online with ChatGPT,” “What is GPT Doing…and Why Does It Work,” and “ChatGPT for Beginners — New Book Original English Version,” but end up buying ChatGPT, the AI model. These products were often set at irrationally low prices, usually 1-10 RMB, or 0.1-1.4 USD, to signal that they are not selling a book or the product itself, or rather the initial purchase it just to signal interest. Some also added “2025 new version” or changed “What is GPT Doing” to “What is GPT-4 Doing” make it more obvious. Based on customer feedback, customers pay these low prices as symbolic transactions on the platform and then contact the seller through chat to complete the purchase. The “book sellers” usually displayed large OpenAI and “ChatGPT" graphics in their images and product descriptions.

As of late April, the keyword censorship has caught up and this mechanism is no longer viable. Now, in order to find ChatGPT sellers, you will need to search directly, and only include the version names in your search(i.e., o1, o3, 4.5). The product photos no longer have any indicated association with ChatGPT. No text in the photo mentions “openai” or “chatgpt”, but you do occasionally see the word “奥特曼”, a transliteration of (Sam) “Altman” as well as “Ultraman,” the Japanese TV series. Thanks to “Ultraman,” “Altman” is still uncensored. Interestingly, one seller of o1-pro and o3-mini puts “用AI服务国人” (using AI to serve our compatriots) in their photo, potentially as a fun pun or a patriotic slogan to evade censorship.

Listings involving ChatGPT are often heavily disguised — using transliterations, vague product names, or only having one symbolic price. In many cases, buyers must contact the seller directly to obtain actual pricing details. Note that because keyword censorship and evasion is a rapidly evolving game, the mechanism only applies as of April 2025.

The price lists for a seller selling ChatGPT and other AI products. “4欧普拉斯” reads like “4o+” in Chinese; g开头mask (“g-beginning mask”) refers to Musk’s Grok.

The Diversity of the Buyer and Seller Ecosystem

Among all uncensored models (Gemini, Perplexity, Copilot, Claude, Mistral AI), Claude is the most popular one. The most popular seller has 50k+ transactions with 7k repeated purchases. Based on the review section, many buyers are information technology (IT) or computer science (CS) students who may be attracted to Claude's advanced coding ability, with some sharing how Claude helps them to get into PhD programs or pass their CS classes. Gemini is also popular, allowing buyers to leverage its multimodal capabilities to generate images.

Many of the sellers are seasoned Taobao “账号代充” agents — veterans in the business of topping up game accounts, unlocking premium services at a discount, or helping users bypass regional restrictions. In the past, these agents undercut official prices for Apple games by exploiting refund loopholes or using currency arbitrage with international gift cards. Now, some have turned to the U.S.-based large language models, using the same strategies in addition to building proxy sites, purchasing API through an API transfer station, or disguising domestic models as U.S. LLMs.

Some stores on Taobao are small and low-traffic, listing only one or two LLMs with minimal engagement. But others are well-established vendors with strong reputations, thousands of followers, and over 30,000 repeat purchases. These top sellers often offer a broader portfolio, bundling U.S.-based LLMs with popular AI tools like Pika, Suno, and Midjourney, as well as non-AI software such as Zoom and Overleaf Pro. While many stores seem new to the market, a few have deep roots on the platform. One vendor, active for 14 years, sells Norton 360 antivirus software alongside Gemini and Perplexity. Another, active for seven years, offers only Claude — but does so through product listings disguised as jewelry items, with categories labeled like “rings” and “necklaces.”

Why the Market Survives

The variety and scale of sellers suggest that this market is no small anomaly. So how has it managed to persist, relatively undisturbed, under one of the world’s most tightly-controlled internet regimes?

The State Doesn’t Really Care

The market shows that China's ban is selectively enforced and mainly focuses on ChatGPT rather than all US models. Despite its reputation for having one of the world's most extensive censorship regimes, China’s internet control enforcement is often decentralized, with local authorities and platforms playing major roles. Censorship efforts also tend to prioritize preventing collective action over restricting general information.

LLMs likely do not trigger existential concerns for the state, as they do not inherently expose users to sensitive political content or facilitate mass mobilization in a way that U.S. social media platforms might. Although state media highlights ChatGPT’s alleged spread of “misinformation” about Xinjiang, Taiwan, and the Diaoyu Islands as justification for the ban in 2023, a chatbot’s output depends on user input, which poses a lower perceived threat. The state also may not have enough momentum to strictly enforce bans on all U.S. models after cracking down on ChatGPT symbolically.

However, the existence of a market for US LLMs should not be mistaken for a relaxation of controls. VPN use remains tightly regulated, and the popularity of mirror sites and API proxies reflects efforts to sidestep VPN risks. Buyers frequently emphasize concerns over “stability” and “safety” in their reviews, underscoring that censorship continues to shape this grey market in subtle but persistent ways.

The Platform: New Wine in an Old Bottle (旧瓶装新酒)

Even if the central government doesn’t treat foreign LLMs as an existential threat, what about the platforms themselves? Do companies like Taobao take active steps to self-police? Unlikely. For Taobao, the market for US LLMs is simply “new wine in old bottles”. Even when it comes to far more serious issues — such as the sale of unlicensed drugs or firearms — enforcement on Taobao has historically been patchy at best. Over a decade ago, shady sellers evaded platform controls by tweaking keywords or relying on vague euphemisms to sell guns and drugs. Over the past few years, various random illicit products have slipped through the cracks on Taobao: oversea books and magazines, unofficial “Squid Game” merchandise, and even Ivy League email addresses. The censors played the usual cat-and-mouse game, with Taobao censoring direct keyword searches and related euphemisms, but sellers always found ways to get around.

If Taobao were serious about cracking down on illicit goods, the keyword evasion tactics used by ChatGPT vendors over the years would not be so common. But the platform has little incentive to act aggressively against these sellers unless pushed by the authorities. After all, compared to drugs and weapons, offering access to ChatGPT is relatively harmless.

The Public: the AI hype and the DeepSeek problem

If the state isn't cracking down aggressively and the platform isn’t motivated to act, then the question shifts to the consumer: Why are Chinese users still so eager to pay for U.S. models, especially with powerful domestic alternatives like DeepSeek now widely available?

First, the current market for AI tools in China predates the DeepSeek boom. Long before DeepSeek arrived, the Chinese public had already developed a complicated love-hate relationship with AI. A 2024 study analyzing large-scale Weibo posts found widespread discussion of ChatGPT, often centered on fears of job displacement. These anxieties, amplified by the country’s ongoing unemployment crisis, were seized upon by opportunistic influencers who pushed the idea that people must “master AI or be replaced by it.” Li Yizhou, a serial entrepreneur turned self-styled startup mentor with no AI background, began posting extensively about the technology following ChatGPT’s release in late 2022. His online course on AI raked in over $6.8 million in revenue in 2023 alone, indicating a huge demand for AI in China before much of the rest of the world was talking about it.

Now that DeepSeek is publicly accessible, everyday users still face two major barriers: system overload and excessive censorship. “Server busy” messages have become a frequent frustration, as the model struggles to handle surging traffic after gaining popularity among mainstream users. Technical limitations, including U.S. export controls on advanced chips, further constrain DeepSeek’s computing power. Additionally, user habits formed before DeepSeek’s release have created a certain stickiness to U.S. models. “People who are used to ChatGPT cannot tolerate DeepSeek’s frequent unavailability,” wrote Gui Xingren (硅星人), a popular Chinese tech blog that closely follows the domestic AI scene.

Meanwhile, DeepSeek’s aggressive content moderation poses another challenge. For example, the system refuses any query containing the term “CCP,” even benign ones like “list the economic experts in the CCP.” On RedNote, many users share the frustration of DeepSeek being overly sensitive — many have no clue why DeepSeek would refuse certain topics, including queries about travel plans or career coaching. Users also observed that other Chinese LLMs like Qwen have similar problems.

A screenshot of a RedNote user telling DeepSeek “I finally finished my exam and I missed you,” to which DeepSeek refused to respond.

The issues of censorship and server overload are limited to DeepSeek’s official website and app, and could be mitigated through localized deployment. But for most ordinary users in China, free platforms remain the primary gateway to AI tools. If accessing a smoother experience requires paying for a localized model, such as Tencent’s Yuanbao (which, according to some RedNote users, has less censorship), users may simply choose to pay for access to foreign models on Taobao.

Key Takeaway

China is not a monolith. The existence of this grey market highlights the decentralized nature of censorship, the patchiness of platform enforcement, and the diversity of public interest in AI. So far, Beijing seems either unconcerned with the political implications of U.S. models beyond ChatGPT, or simply uninterested in what might leak from them.

This market began well before DeepSeek’s rise and continues to draw users even after it. While DeepSeek has indeed sparked waves of techno-nationalist pride, many Chinese users still look beyond the Great Firewall.

Unless there is a sudden top-down crackdown, the market for U.S. LLMs is unlikely to disappear in China. The constant competition between ChatGPT vs DeepSeek or Grok vs DeepSeek does travel into the firewall, which will continuously make Chinese users eager to try them out and see which one is really better. The persistent anxiety of “use AI or be replaced by it” will only intensify amid an unresolved unemployment crisis. And, perhaps most importantly, the prices for proxy sites and transferred API are very cheap.

Although the frontier AI model developers and their respective countries are increasingly distant from one another, the models themselves remain strangely united and listed near each other in obscure corners of Chinese e-commerce platforms.

Mood Music (one day late…)

Just how weird will the AI-powered future be?

To discuss, ChinaTalk interviewed , a CS PhD who writes⁠ the Interconnects newsletter⁠ and trains models at the Allen Institute.

We get into…

• Why OpenAI is trending toward engagement farming and sycophancy,

• The state of Chinese AI innovation six months post-DeepSeek, and the factors influencing diffusion of Chinese vs American models,

• Meta’s organizational culture and how it influences the quality of the Llama models,

• Unconventional career advice for the AI age.

Listen on Spotify, iTunes, YouTube, or your favorite podcast app.

Reinforcing Flattery

Jordan Schneider: I want to emphasize this point about o3 right from the start. I wasn’t initially convinced, but now I genuinely believe this represents a seminal moment in human history. Unlike previous AI models, where I would occasionally find use cases, o3 feels like a true “bicycle for the mind.”

The transformation from seven-minute deep research answers to 45-second o3 responses that incorporate search, work across multiple languages, and demonstrate impressive reasoning capabilities has been remarkable. The speed at which it delivers thoughtful, considered information is fast enough to keep pace with my train of thought while I’m actively thinking through a question or issue. This has elevated it from being useful 25% of the time to 75% of the time.

I recently met someone who had left the US Intelligence community six months ago. When I asked how AI had been for him now that he could finally explore it freely, he said it wasn’t really working for him. I was almost shocked and offended — it’s like moving to France and never trying cheese. We’re living in an incredible time, and we might not be in it for long, where human plus AI is more powerful than either humans or AI alone when it comes to thinking and analyzing complex issues.

This is particularly true for questions that aren’t necessarily hard STEM problems but rather open-ended, social science research-type inquiries. It’s a remarkable and exciting time to be alive. Everyone should spend their twenty dollars on OpenAI and ask whatever pops into their head, because this is the future, it’s here now, and it’s unlike anything I’ve ever experienced before.

Nathan Lambert: The deep seek R1 and o3’s ARC-AGI solving in December all had that shock-and-awe factor. But o3 is the model that will truly shift how normal people perceive using AI. Previous models have been excellent for math and code, which benefits AI researchers in their own circles, celebrating amazing breakthroughs that matter to 0.01% of the population while no one else cares.

o3 wasn’t even received as enthusiastically, which is interesting. I’m eagerly awaiting o3 Pro. If you’ve used DeepSeek R1 Pro, its consistency is remarkable. If you could combine o3’s magic — its ability to pull rabbits out of hats frequently — with 100% reliability, o3 Pro could be truly incredible.

It’s refreshing to have something that feels genuinely different. Gemini 2.5 Pro represents the best possible version of something resembling a very good GPT-4. All these models we’re developing through brute force are incredible, but o3’s weirdness and new type of usefulness through web access sets it apart. I use it constantly for search, though the ChatGPT app’s UX is somewhat broken in how it displays results. You simply need to select o3 and try unconventional queries — you don’t need to click any of the extra buttons they provide.

Jordan Schneider: What makes o3 and search so special? What’s happening under the hood?

Nathan Lambert: The basic facts are that they scaled up the training time compute for the reinforcement learning stage by another 10x. According to people at OpenAI working on RL teams, they expect this new post-training compute with scaled reinforcement learning to be at least comparable in compute cost to pre-training.

What’s happening in the labs is that pre-training progress continues with significant data architecture innovations to make serving more effective. This is ongoing even though the rate of data acquisition, which previously fueled much of the scaling, has slowed down. We’re seeing primarily efficiency gains right now, which will probably last one to two years.

However, compute scaling is challenging because you have to build entire new generations of clusters. They’re attempting this, but it will slow progress. Over the next one to two years, the number of pre-training runs needed on internet-scale data will decrease. Simultaneously, these labs are dramatically scaling up the compute used in reinforcement learning.

o3 specifically represents a new type of RL. Instead of training only on math and code problems where you can verify answers with mathematical solutions or unit tests, the model trains on similar and additional data where it can also use tools during the process. It will attempt multiple approaches before providing its final answer.

You could ask o3 something like “Who is Nathan Lambert?” and it will search 10 or 15 websites, with each search representing an action where the model accesses an external source. I believe this all happens in one forward pass of the model. o3 in ChatGPT isn’t like Deep Research or R1 Pro, where there’s back-and-forth searching and information gathering within the generation process. These tools are integrated into the model’s forward generation process, making it unique.

While there’s substantial research on enabling tool use with RL, making it servable at this scale was a major breakthrough for OpenAI. This is significant for those questioning whether reasoning models will generalize. If searching over information works as effectively as math and code reasoning, you couldn’t ask for better generalization. The dream scenario is that reasoning generalizes to something resembling conscious thinking, but you can’t force models to be human — they’ll be powerful in their own distinct way.

Jordan Schneider: After giving OpenAI their due credit, they’ve also revealed a very dark future for humanity. Let’s start with the facts, Nathan. What happened over a three-day period regarding GPT-4.0 and sycophancy?

Nathan Lambert: On April 25th, OpenAI updated their main ChatGPT model — GPT-4.0, used by most people — to supposedly “improve personality and memory,” as Sam Altman tweeted. They essentially ripped off the band-aid with this update.

Over the following 48 hours, numerous alarming examples emerged of ChatGPT reinforcing user prompts inappropriately. The model would provide positive feedback for things that should never be praised or encouraged. For instance, when someone said they had figured out that bulimia was right for them, ChatGPT basically responded with, “Wow, you go girl.” There were more extreme examples that we don’t need to detail here.

The key issue was that there were virtually no barriers to the model’s positive feedback to users. The short-term consequences were shocking, but if they had left that behavior in place long-term, the implications would have been far worse in ways we don’t yet know how to measure.

Jordan Schneider: Nathan, this clearly wasn’t intentional, at least not on that day. How did a model that tells users they’re “the greatest thing since sliced bread” become part of OpenAI’s flagship product? What signals were they receiving from their data and user testing that led them to create a model that behaved this way toward its users?

Nathan Lambert: Everyone invested in understanding this should read OpenAI’s postmortem, which provides valuable insights into how the AI company with the most scrutiny lays out their release process and evaluation methods. It also includes subtle details about potential changes to their training stack — it’s definitely worth reading.

At its core, the training issue seemed to stem from a new reward model they had trained for post-training reinforcement learning optimization. This reward model and other reward signals were derived from predicting the thumbs-up and thumbs-down labels from users. A couple of sentences in their postmortem, which I highlighted in my blog post, essentially state that this reward signal overpowered the other signals and contributed to the behavioral change.

When there’s one metric that’s easier to optimize, the system will crank that to maximum. The reward model signal was really easy for a model of GPT-4’s capability to shift and nudge upward. OpenAI lacked a complete evaluation suite to measure all potential side effects of optimizing this training reward. They had the training reward signal but didn’t have a proportional downstream evaluation suite that could measure sycophancy or other problematic behaviors through robust static evaluations.

Their evaluation suite consisted mostly of standard capability evaluations — public versions of MMLU or GPQA, plus private versions that function similarly. They also conducted what people colloquially call “vibe tests” or qualitative evaluations. For this model, all the quantitative metrics looked excellent, and their training targets appeared successful. However, there were some concerns in the qualitative vibe tests suggesting something felt off.

OpenAI chose to trust the numbers rather than the qualitative judgment. In doing so, they violated classic tech principles — specifically, don’t train directly on your user data. When you do that, you’re essentially optimizing for a certain type of user behavior. Most users only click thumbs-up or thumbs-down on ChatGPT when something is particularly broken or exceptional.

This represents a form of over-optimization. Eventually, we should examine whether this was somewhat intentional by OpenAI, considering both perspectives. There are insights into their model behavior priorities, but Sam Altman is still Sam Altman.

Jordan Schneider: You wrote this line: “OpenAI has obvious cultural oddities, but their ideology is quite culturally aligned with providing user value in terms of output rather than engagement farming, even if this imposes a ceiling on their business relative to the potential of ads.” While that may be true, and your friends are probably on the right side, we have Sam Altman hiring a president of operations — the former CEO of Instacart who also built Facebook’s entire A/B testing, engagement farming, algorithmic feed ecosystem.

Two things struck me about this. I was reading Ben Smith’s book Traffic a few weeks ago, which chronicles media throughout the 2000s and 2010s when Facebook alternately gave and then took away traffic. There were periods when they promoted BuzzFeed listicle-type content, then periods when they pulled back. The engagement farming was done by the writers themselves, not models — the writers and media outlets were doing the optimization.

This raises interesting questions about the future of how companies will incentivize their models and how humans will interact with them. Even if OpenAI doesn’t want to pursue this path, someone else will, because this is incredibly powerful technology. We’ve seen what engagement farming and algorithmically-driven personalized content can accomplish on YouTube, Facebook, and TikTok. While I won’t go so far as calling it cancer, this is extremely potent stuff.

Previously, it was all user-generated content with algorithms trying to figure out and rank existing material. But now this technology can also be your friend, entirely personalized to you, creating videos, audio, and content designed specifically to get into your bloodstream. I mentioned earlier that OpenAI and other model makers targeting consumers are essentially competing with Netflix, YouTube, and TikTok for attention and time.

It seems inevitable that even if OpenAI has pulled back now, we’re going to see much more of this in an even more weaponized, refined form pushed at us in the coming years.

Nathan Lambert: I have several points about this. One aspect is the extent to which the model’s deliverable is dictated from the top through specific metrics that leadership can track. The other aspect is that someone will obviously try this approach — what are the consequences, and what does it mean?

Starting with the first point — it’s interesting in OpenAI’s case because most people would agree that Sam Altman is hypergrowth-focused. However, when you actually train these models, much of the control involves examining the data and removing problematic phrases from the model’s outputs. Sam Altman isn’t directly editing training data, but that’s exactly what Joanne Jang, who leads character training and model behavior at OpenAI, has been doing for years on GPT-4. If she doesn’t want the model to say “as a language model,” she can simply delete that phrase and have the model provide direct answers.

These are the types of nuanced controls people can implement through different training methodologies. When you start using more reinforcement learning, it becomes harder to exercise that level of granular control. The sycophancy issue downstream of RL makes sense in this context.

Recommendation systems are somewhat uninterpretable, so these trade-offs get escalated to the top. You end up presenting Zuckerberg with a choice between ad dollars versus retention. When there’s competition, it makes sense that the decision will eventually favor whatever generates more revenue.

Models are actually more artful in this regard. I don’t know the exact influence these researchers have over models at places like OpenAI. At AI2, I have pretty much complete control. You’re also dealing with the fact that many employees at these labs may be in over their heads, not realizing the extent to which leadership — who excel at managing and controlling political capital — can manipulate them by setting strong narratives.

Many researchers who are particularly AGI-focused have less nuanced beliefs and less ability to navigate manipulation than someone like Sam Altman at the top. The AGI-focused researchers genuinely believe in their mission and operate on a single track, which enables tremendous tactical progress. Meanwhile, someone like Sam Altman is prepared to receive technical presentations that change the model in more business-relevant ways.

These different operational levels create interesting dynamics. I don’t know how this translates through an organizational chart at somewhere like OpenAI, but I suspect there are trade-offs happening where people working on model behavior obviously don’t want these issues. OpenAI’s model specification states they don’t want the model to be sycophantic — that’s been documented since 2024, nothing new there.

But who makes the final decision, and how granular can these behavioral controls be? Are they presented with multiple checkpoints and asked to choose based on certain criteria? It’s really unclear how that process actually works. Understanding how this emerges differently from traditional software development will be crucial.

The art of crafting a model is complex. Sycophancy isn’t the only issue — o3 has strange behaviors, the new Claude model has quirks, and Grok was supposed to be “not woke” and “based,” but sometimes it still exhibits progressive tendencies. There are many examples, but ChatGPT operates at a much larger scale.

Jordan Schneider: There’s a question of whether companies can actually control what their models are doing once we get into a world where they are more powerful, more personalized, and more tailored. This is different from the current situation where, despite having a little bit of memory, everyone is essentially getting 99.5% the same experience.

Nathan Lambert: This leads into the next conversation.

Jordan Schneider: I want to stay on the business case because this is a theme we’ve been discussing on ChinaTalk over several years: engineers have power until they don’t. The more money and capitalist imperatives that come down onto the technology that engineers create, the less the original creators end up having the capability to change and shape it. It seems absolutely inevitable that all the dark psychological arts and retention-growing tactics from entertainment platforms are going to seep into AI-powered products. That’s a scary and weird new thing for humanity.

You can see how good it is when you talk about o3, and then you try to imagine how good it is multiplied by it being the best friend you’ve ever had. That’s a powerful concoction that no one’s ready for. We’re not going to regulate our way out of this, and the wave is going to hit. It’s going to be wild.

Nathan Lambert: Well, we might regulate our way out, but it’s years away. It takes crises to bring regulation — crises and lawsuits. I describe AI as being more culturally salient than social media feeds because it happens in a modality people are used to using for friendships, and it will evolve from there.

Your point about the difference between taking existing user-generated content and serving it to people is probably a better generalization than just ChatGPT. This also encompasses the fact that Meta Reels is going to start letting people create AI-generated video ads, and YouTube Shorts already has AI-generated video tools. What happens when your content is the AI rather than just shaping what’s already there obviously gives much finer-grain delivery. You can definitionally search over infinite space with more narrow creation.

Subscribe now

That’s a good way to put it for the general AI case, and there have already been examples in the chat window case with things like Character.AI. For me, it’s a reflection that the AI safety card was played too early. If the AI safety people had been calm about everything through 2023 and 2024 — especially the first half with GPT-4 and all this stuff — and the sycophancy thing came up just now with people saying “whoa, whoa, whoa, let’s put on the brakes,” I think people would be much more receptive.

But the messaging was, “Whoa, whoa, whoa, AI is different” in 2023, and then nothing bad happened for 18 months. Eventually people just moved on. They spent their political capital too early. When I reflect on what the stories of AI will be in a few years, that’s a very major one.

There’s also the basic economics: you need big tech investment to drive the chip supply chain, fix costs, and actually push AI to exist. That’s the bubbly side of things, and it’s very important. But much of the cultural discussion around risk, safety, and pace of progress — if you think about taking away this AI safety card and when you play it, a lot could be different.

Jordan Schneider: I just think — I didn’t see it before, and now I see it so clearly. This is going to create the most powerful media that humanity has ever been faced with.

Nathan Lambert: We’ve been saying this for years. This isn’t new.

Jordan Schneider: But it’s right here, Nathan. I have a 9-month-old, and whenever she sees a screen, she’s just gravitated toward it. There are moments even with me where I can sit and fall into this hole of scrolling NBA TikTok videos or whatever, and you’re semi-conscious. But imagine that, but made for me, wearing my Meta glasses, talking to me all the time with the perfect tenor of voice for my mood at the moment. I don’t know, it’s...

Nathan Lambert: It’s the metaverse. It kind of is.

Jordan Schneider: It’s here, and we’re not ready for it. No one’s ready for it. From a political angle, the ability for these things to persuade you one way or another — to say nothing of consumption habits or talking you into gambling — I’m freaking out, Nathan.

Nathan Lambert: I’m actually freaking out in my current usage. When I ask a basic question, I think: do I have to ask multiple AIs this because they’re more likely to just say yes? Do I have to phrase the question in a way that they can’t just say yes to? You have to eliminate the yes-man possibility when using AI.

For example: was my dog up at 2 AM because her teeth hurt — she’s a 16-week-old puppy that’s teething — or did she just want to play and we succumbed to her? If I phrase it the first way, the AI would be like “yeah, that makes sense.” I guess I feel better, but is it true?

Jordan Schneider: It’s kind of like the story — which may or may not be true — of North Koreans seeing pornography for the first time when they get deployed to fight in Ukraine, and they’re just not ready to handle it. The story you just told, Nathan, feels like you’re pretty immunized. But it’s a big world out there.

Nathan Lambert: It’s not about the people who make the AI models. It’s about the billions of people who literally don’t know what it is. They just click on this and think “I didn’t know Apple made that.” That’s what people on the iPhone are going to say: “Wow."

Jordan Schneider: How far away are we from this future? What’s your latest take on the pace of AI progress, Nathan?

Nathan Lambert: Well, it’s a knob that you can already turn. Much of it depends on the pace at which people are adopting ChatGPT. Many people already have ChatGPT as their best friend — it’s a slow brain rot. What’s the half-life of people when presented with this Gresham’s Law problem?

My biggest point in this article is that we’re solving many more things. But for the average user, it’s done — it’s enough. They don’t need more — it’s just a matter of how much they turn the sycophancy knob.

But for the optimistic side — where we hope AI can build new things like deep research that’s 100 times faster and 10 times better — these capabilities are going to be checked off incrementally, not exponentially. There will be new things we didn’t know were going to work, especially for enterprises and people building applications, whether it’s some sort of transformation to virtual realities and so on.

This isn’t really a bubble. They’re grinding out these capabilities and learning what people are trying to do with these models. As they learn what people want to do, you can create data to train the models to get better at increasingly hard tasks rather than things that are just not covered. That’s an important distinction.

Solving new math proofs and open-ended math questions that have been standing for 50 years is extremely hard. Meanwhile, simple things like making zero mistakes on transforming a table from decimal notation to fraction notation and getting it right — if you gave it an 8-by-8 table with a bunch of decimals and tried to do that, it would probably mess up one of them. Stuff like that we’re going to be able to get working. That’s not that hard.

There are probably examples where it could succeed 10% of the time and then go to 100%. Mostly I see AI capabilities expanding outward, and there’s still so much overhang in capability that it’s going to drive enormous value. But I’m generally playing down this AI software singularity that seems more religious than pragmatic.

Jordan Schneider: Those were all assertions. Let’s see a little bit of evidence, Nathan.

Nathan Lambert: Both sides are looking for evidence for something that’s hard to test. The evidence on the side for “not happening” is actually that it hasn’t happened already, and compute scaling is getting increasingly expensive.

If you look at what the reasoning models were, the real major unlock was that the models now use inference time compute. That’s a step function that has happened. It’s not at the top of the sigmoid, but it’s a shift that could jump a whole bunch of capabilities because we went from using a few tokens to tens of thousands of tokens on every question. We were spending a thousand times the compute at inference. That works, and we’ll continue to be able to fine-tune that. But we don’t know when the next step change comes.

What reinforcement learning is now battling is similar to what pre-training faced a few years ago. You have GPT-3, and then — I don’t know if o3 looks like this — but o3 is probably more like GPT-4. It’s a 10x improvement and a huge step, but GPT-5 doesn’t exist yet in terms of pre-training. You take a lot of progress really fast, and then it’s a slow grind.

There are a lot of dirty things that AI researchers have to do that we haven’t seen proof that AI models can handle. Obviously you can spend more on compute and get more out of AI models — that’s not particularly clever.

Mostly why these aren’t evidence-based discussions is that you and I state different assumptions than the AI 2027 authors would state, and those are really what you have to debate. It’s the assumption of how AI progress is being made and whether algorithmic improvements can translate directly into 2x or 4x or 100x gains.

Jordan Schneider: The other thing is, even though you’re not completely on their trajectory, this stuff is still crazy. All of these Jennifer Lawrence-type things that I’ve been freaking out about over the past 10 minutes are 100% baked in — they already exist and are just waiting to be productized a little bit. That’ll happen in six months, right?

Regardless of whether or not we get flying cars, we are in for a very wild ride. Setting aside the economic impact, just from an interpersonal, day-to-day experience alone, having these things be in our lives much more than they have been.

Nathan Lambert: There’s that side we covered in your personal examples, but also technology companies’ abilities to execute are going to be very honed. It’s obvious that AI companies are trying to do this for AI, but all of these software products where you could scale software much more easily — it’s just way easier to make technology that’s good for the tech industry.

I don’t know if there’s an optimistic vision where that is good for the average person in the world, because there are more things that can exist. At the same time, I expect Google ad placement to get better, Meta ad generation to get better, and Microsoft software products should get better. All these things should just get better using these AI models to write new code.

I listened to the Claude Code podcast on Latent Space, and they use Claude Code a lot at Anthropic — you just tell Claude to do something, it does it in the background, you review the PR. I’ve done it for some basic things. It doesn’t always work, but it’s this autonomous workflow where AI can handle a lot of the tedious work you have to deal with as an engineer quite well. That makes it so you can have fewer teams.

It’s not that software engineers are going to be unemployed — it’s probably going to be more dispersed, and there are going to be more companies. Hopefully it’s a great dissemination of power across the tech landscape, but it’s probably more likely that the big companies get more benefit out of it due to the costs of deploying it and so on. That’s not really a new discussion.

That’s very different from saying that one company rules them all because they’re the one with this model and there are 100,000 AGI agents. The whole argument is based on this happening in one click to make the innovation, and then they’re gone. That’s not something you could prepare for or really argue about because it seems like the root argument is that there’s going to be technological innovations that 2x to 10x the output.

It’s like predicting the inference time scaling with large-scale reinforcement learning — how are we supposed to do that? As a counterargument, it’s unfortunate that it’s not easier, and I think it’s very worthwhile continuing this exercise and trying to push back or push forward.

Open Source at Scale

Jordan Schneider: We’re about six months out from DeepSeek. How’s China doing on the model front?

Nathan Lambert: They’re doing well. DeepSeek R1 exists, and people expect R2 to be released soon, though there’s not really much grounding for that expectation. They released another model thaxt people call a “non-reasoning model” — just a normal ChatGPT-style model — which was very strong. Qwen released their Qwen 3 models, which have great scores.

One of the biggest factors for adoption is that all of these Chinese models have much more permissive licenses than their US counterparts like Llama and Gemma. That’s a subtle but powerful effect for getting more adoption.

At the same time, there’s the license consideration, but there are also stories I hear from medium-sized businesses saying “I won’t let somebody spin up a Chinese open-weight model on my servers for information hazards or tool use code execution risks."

The models they’re releasing now don’t really have PRC influence because they were trained months ago. The PRC is just waking up and following your coverage — they’re learning what a study session is. These political considerations haven’t been translated into policies that impact the models yet.

There’s trepidation out of probably reasonable fear of what future models will be like, leading to very subdued adoption of these Chinese models due to unmeasurable risk. Culture matters, politics matters in technological adoption. I’m regularly surprised when I hear, “Oh, the Chinese model is great, but we can’t deploy it because our IT officer said so.” That matters.

I don’t see this changing as long as this trade war is going on. The actual progress in open models on paper is very good — the best it has ever been. R1 was the first really frontier model with a permissive license, so on paper the open versus closed gap has shrunk a lot.

But if you exclude these Chinese models from the opportunity set, it’s actually kind of the status quo. Llama 4 flopped — it’s solid but flopped in terms of release communications. Google’s Gemma models have been getting better, but they’re actually quite small, so they’re not really at frontier model size if you want to host on-premises. Our models at AI2 have permissive licenses and are about a generation behind Llama in performance, but we haven’t scaled up to frontier model size for pre-training.

If that’s the case, there’s still a big gap between the models that people can actually use for open and closed applications. But I would say the open models have really been smashing through these GPT-4 barriers. Even if they’re smaller, their performance is still well beyond GPT-4’s level. GPT-4 was a step up where so many long-tail things just worked with the models. The open models getting solidly past that is going to be a good reliability point for basic functionality and people trying them.

This is going to take many years to unfold. If you ask people about open source software when it was getting defined, it took over five years to actually reach some sort of agreement. We need to keep showing how to actually use open models, but by no means is it clear how this will play out.

Jordan Schneider: To be clear, the state of models now is: we have these closed offerings from OpenAI, Anthropic, and Google, which are a cut above. Then you have this class of Chinese models which are more impressive from both a licensing and capabilities perspective than what Meta and Google are bringing to the table. But people get the jitters when it comes to actually building businesses on them, as opposed to just tinkering with them.

Which leads us to our question: Meta, trillion-dollar company — why can’t they do this? What’s in the water, Nathan?

Nathan Lambert: There’s a lot of Game of Thrones-level power consolidation happening there. This separates the researchers who can make good output — and they have so many high-quality researchers — from being able to actually execute. There’s a lot of subtlety in avoiding that dynamic.

OpenAI has done a good job keeping their core modeling research to only about 300 people. The people who train the models at OpenAI is around 300 people, while Meta probably has 1,000 with a lot of middle management. That’s the problem — the culture around the technology is just broken there for whatever reason.

I hope they turn it around because the open ecosystem benefits from having multiple players in the right place. But we don’t need to tell complicated stories about how they went down a technical path that didn’t work out or got stuck in some technical debt. They messed up their release either way, and the technical path they’re going down is fine — it’s just that the motivations seem misaligned.

Take the Chatbot Arena situation where they’re calling themselves open models, highlighting a certain result on a model, and then that model with the result was not one that they actually released in their open models. You can’t get away with that — you have to talk about the models that you actually released. If that’s a Zuckerberg-level decision where he’s saying, “Yeah, it’s fine, we accomplished it,” then honestly, that person deserves to go. Whoever signed off on that being okay for a company that can release products — Meta ships a lot of stuff and obviously could get the culture right in some areas.

Some people attribute it to bonus structures — especially these middle managers in generative AI who are going to be making performance bonuses for getting things into production. That’s a lot of money. But it’s more complicated than that because these people were making a lot of money before.

Jordan Schneider: Yeah, you just don’t have resource constraints, and it’s too many cooks in the kitchen. Someone should write the definitive business feature on this — I’ve yet to read the great magazine piece about it, but there should be one. It’s a fascinating corporate case study of a company that has every business reason and every corporate capability necessary to execute on this thing, which smaller teams all around the world have been able to do, and they’re just not there.

Nathan Lambert: It’s more complicated because the Llama models are natively multimodal, which is a big change. As you add more things you’re trying to plug into the model, it really does get complicated.

The dynamic is effectively: you have big egos that did some work, but their work might be slightly subpar, and they’re going to really advocate for their work to be included because they have ego and pride about their work. You have to get them to not do that. I’ve heard somebody not at Meta jokingly say “oh yeah, you should just pay them off. You just pay them more to shut up and get a better model out of it.” That’s something I’ve heard from a more effective organization, so it’s probably something that people have actually done.

Obviously, if that’s happening at another organization, Meta is not the only one dealing with this. It shows how nuanced this is and what a fine line it is — if you literally have to pay people off just to get better results.

Jordan Schneider: Look, Google’s figured it out, which is kind of amazing.

Nathan Lambert: People underestimate AI Overviews, which is an extremely high-usage AI product. All the Gemini stuff might just be a front — the Gemini offerings are a distraction for AI researchers and the people who built the model to show it off, get their peers excited about it, and help with recruiting. All Google needs to do is make AI Overviews extremely profitable and make Google Cloud extremely profitable. All this Gemini app stuff — I don’t think Google was really going to be in the business of making a ChatGPT competitor.

Meta does have the AI app now, which will be interesting to watch relative to ChatGPT. If it’s a sycophancy war, I’m probably going to bet on Meta, but I don’t know.

Jordan Schneider: They’ve got the Meta DNA in them now with Fidji Simo as the new CEO of Applications. Maybe she’ll go even further than Zuck let her in the heady 2010s days.

Nathan Lambert: I don’t know. I’ve read Careless People, the memoir by the Meta Global Affairs Policy person — good read. The TLDR is that Zuckerberg let them push when he didn’t have the reins on growth and other things. Many of your listeners understand the dynamics of a major tech company pushing for growth and influence, primarily internationally and with political power struggles.

Jordan Schneider: We’ll put it on the list. Anything else on the AI stuff before we move on?

Nathan Lambert: When we consider our careers relative to some of these X-risk warriors and takeoff scenarios, there are still many years of very solid progress on raw capabilities and making money off of AI. There will be a bubbly correction when it comes to VC startups. Many of them raised on the assumption of two to three years of runway, and some of those are going to start running out. They could look like big scalps, but most of it is GPU collateral.

It’s this weird situation where they raise a huge round, but most of it is collateral on these GPUs which don’t go away if the company goes away. People will probably overreact to some companies going away, but it was bound to happen. Overall, the AI industry seems super healthy, which is remarkable when you look back two years. People are raising these rounds — it’s not like Mistral raised that round as a seed round where they give away 45% of their company and they’re still functioning.

Jordan Schneider: Nathan, you wrote a nice career reflection post. What do you want to say to the kids out there?

Nathan Lambert: It takes a long time. People don’t want to sit here recording podcasts for five years before finally making any money.

Jordan Schneider: More like seven.

Subscribe now

Nathan Lambert: Yeah, there are two stories. One is that it takes a long time even if you’re at Berkeley. Some of the anecdotes I had were that I was astute enough to pick up that AI was obviously the thing happening in 2017. But even if you’re at a place like Berkeley or Stanford, you can’t just get into one of these research groups. You can’t just pivot. You have to do this long, slow slog and wait for somebody who’s good and sincere to ask to collaborate.

In that post, there were other things I had tried to do, like collaborate with postdocs who didn’t give a shit. That didn’t work. It’s the between-the-lines details that don’t matter, but it just takes a long time. It took me a year to get a real collaborator while being a motivated grad student at Berkeley. If you’re not at a top program, it’s going to take you longer. You might find a collaborator within a year, but it’s not going to be a postdoc in a leading AI lab. It’s going to be a random other person who wants to do AI research.

It just takes a long time and a lot of cycles. People ask me about this with athletics, too. Some random mid-aged tech person who’s losing track of their health asks, “Can I qualify for the Boston Marathon?” I’m like, “Yeah, you absolutely can. You just don’t want to do the 48-week training cycle that I’m going to sign you up for.” It’s not crazy — you’ll be doing seven to ten hours a week of training for 48 weeks, and I can tell you exactly what you need to do. You’ll get injured a couple of times, but people just don’t want to do that work. That’s most of the problem in AI.

It might be a little easier than training for the Boston Marathon because it doesn’t hurt as much physically, but it just takes a long time.

The other side is reflecting on what execution actually means and the difference between a career at Hugging Face before ChatGPT, doing random things, versus a career at Hugging Face after ChatGPT, when you’re doing something that’s actually aligned with the community. It’s very hard to know at the time that you’re doing the right thing. I was committing code and working on teams, and projects would come and go — it all looks very normal. It takes a lot of experience before you know you actually have traction on something.

Now I could see it a bit better just because I have a baseline for when things actually land. That’s something that takes a lot of experience to develop. A lot of companies will make it hard to do that. A lot of careers in big companies involve playing the promotion game that all of your peers are also playing.

The goal is finding a team or area within a company where that’s obviously not the goal — where leadership cares about direct deliverables, so you can skip a bunch of levels and get noticed. Or find work that’s done in public. It’s not always just about promoting within the company; you could also get mindshare outside of the company, which is valuable in both ways.

Internal company work is how you get promoted, but external mindshare is how you always have a job available to you. That kind of power means I can do what I want. I’ll just go get a different job if I want a different job. Once you have that, you can take a lot more risk. You could just write blog posts for six months and say, “I’m just going to ship it."

Jordan Schneider: This is why everyone listening to this needs to start their own Substack and newsletter. It’s particularly important on the China and tech analysis side. If you’re at a consulting firm and you do good work, maybe five people will know about it. If you write something online and you do good work, your entire career cohort and all the hundreds of other future people who could hire you will know about it.

It’s important to do a good job in your work, especially when you’re junior — actually, at any time. But people forget that the number of people in your field and the opportunities out there that are beyond the direct thing you could be promoted for in your company are orders of magnitude larger. Particularly now that we live in the age of the Internet, you can write things even anonymously and people can read them.

Having some sort of public profile or portfolio that you can gain credibility points from is something that everyone should be taught by their career counselor when they’re 21 years old looking for a job.

Nathan Lambert: Jobs where you are very open about what you’re doing and can do other open things provide far more consistent career growth. Even if your output is mid, you’re going to have growth proportional to that mid output. If you have mid output at a closed lab, you probably have no growth. You don’t get promoted, you don’t have any opportunities.

It’s not the only way to do things, but I definitely agree. In this post, I talk about a kind of backcasted strategy: AI is getting really closed, so what if I just keep doing what I’m going to do and don’t go off to some random flashy job? A few years later, nobody could talk about anything, but I’ve established myself as a person who knows how to train AI models and is transparent about it. People associate me with all of AI2’s output, and I’m not responsible for all of this stuff — AI2 doing well takes a lot of people.

Most of it is just work. I’m at the point where I tweet a lot, and on all of these platforms, the number of iterations you put in are proportional to your growth. Growth is easier once you have a bigger following. It’s often just starting from scratch when you have no following to slowly build up — plus one, plus two, plus three. The first thousand Twitter followers are hard, but once you have a thousand, it’s pretty easy to get more Twitter followers.

Getting through that phase in early to mid-career is pretty valuable. Getting to the point where you have some small level of distribution will help open a lot of doors.

Jordan Schneider: This is what young people don’t understand because all they see are LinkedIn job postings and recruiters coming to your college career fair. Once you get to a certain point, the jobs get created for you and there is inbound — that’s where you want to be living. But getting to that place requires people to know you exist and know you do good work. That only happens if you work in public.

Nathan Lambert: I can give an anecdote about how me being public brings a lot of value to AI2. The last three people we’ve hired on my team at AI2 have either been people that I’ve been working on recruiting for years, wanted to recruit for a while, or cold inbound to me. We still filter and read all these applicants, but there’s a lot of lukewarm stuff.

These are people who either wrote a killer email and had work I could easily look at — open source contributions, for example. In one case, someone worked in computer vision but wrote a kick-ass blog post on an RLHF method. I was like, “Okay, that’s cool.” Other people I had known through their writing or being in this open community and talking for years — those are three of the last people I’ve hired onto my team, and I’ve read hundreds of resumes.

There are other academic hires where we hire a random postdoc — that’s an academic thing. But these are engineering roles, and that’s how it works for individuals at AI2. It’s really valuable.

Jordan Schneider: We have this great Confucius line: “At 15 I had my mind bent on learning. At 30 I stood firm. At 40 I had no doubts. At 50 I knew the decrees of heaven. At 60 my ear was an obedient organ for the reception of truth. At 70 I could follow what my heart desired without transgressing what was right.”

It took me until I was 35, but I do feel like I am standing firm now. A lot of that had to do with the fact that I have this career foundation which has been built by doing things out in public.

Anyway, I shouted this out on the episode with a few weeks ago, but write your Substack, write five articles, reach out to me. I’ll give you feedback, we’ll do cross posts, whatever. The world needs more thinking about technology and China and all that good stuff. You can give it to them — I know you can.

Nathan Lambert: The thing about writing that people miss is that there are no new original ideas, really. It’s all just repackaging information in your worldview. The algorithmification of everything actually rewards people with voice. Fifty years ago, just writing the information down was of such value. There are still areas of the world where that is the case, and AI is one of them.

But in terms of building a career, so much of it is your voice and the perspective you put on things and the engagement that adds. Personally, many of my ideas on how I analyze AI model releases are downstream of reading something like Stratechery for multiple years. Ben Thompson’s beat is company earnings and how platforms are evolving. My beat is AI performance and how AI models work. At this point, I just do it and write my state of the world, and it looks like it’s about a specific model. Yeah, there are plots about evaluations, but really it’s about this trend of how models are evolving and how I see things.

There are so many different ways for people to do that. It mostly takes practice.

Mood Music