Can China use military force to achieve its political goals, without triggering nuclear war? To find out, ChinaTalk interviewed Fiona Cunningham, a professor of Political Science at the University of Pennsylvania and author of the new book, Under the Nuclear Shadow: China’s Information-Age Weapons in International Security.

Co-hosting today is Michael Horowitz, Penn professor who served in Biden’s Department of Defense.

We discuss…

• How to use open source PLA documents to conduct deep research,

• The evolution of Chinese defense strategy, including the impact of the third Taiwan Strait crisis,

• Nuclear modernization and China’s “no first use” policy,

• How the PLA makes decisions, including why they chose to develop cyber capabilities, anti-satellite weapons, and hypersonic missiles over proposed alternatives.

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

PLA Sleuthing

Jordan Schneider: Fiona, why are PLA studies cool and important?

Fiona Cunningham: One of the biggest issues of the day is the prospects for conflict between the United States and China. How would such a conflict unfold if it were to occur? What can both sides do to improve their chances of not only succeeding in a conflict but preventing it from taking place? US allies, countries in the region, and stakeholders worldwide want to understand these dynamics.

The PLA is also an important domestic actor in China. It plays a big role in the party’s history and serves as the ultimate backstop that keeps the Chinese Communist Party in power because it is the armed wing of the Communist Party — not a state military. This arrangement would be similar to if the Republican Party or the Democratic Party in the United States had exclusive control of the US Military.

For these reasons, we need to understand everything possible about how the PLA operates, its weapons procurement, war planning processes, and its assessment of US and other countries’ actions.

Mike Horowitz: From a US national defense perspective, this is the most important part of the most important country in the world.

Fiona Cunningham: Mike says it better than I do.

Mike Horowitz: Not better, just shorter.

Jordan Schneider: There are obvious epistemic challenges when studying an adversary communist system’s military, particularly using open-source materials. Nevertheless, you wrote 300 detailed pages with over 200 interviews and an enormous amount of documents about Chinese domestic debate and doctrinal evolution regarding nuclear war — how to prevent and avoid it. When people say, “How can you ever know anything about the PLA?” what’s your retort besides waving your book in their face?

Fiona Cunningham: Most militaries, including the PLA, must communicate internally. They teach their students and share information within the organization. At least ten years ago, you could walk into certain bookstores in China and find books with ISBNs that revealed what the PLA was researching and teaching its students about almost every aspect of warfare, not just the areas I researched for my book.

Paradoxically, with nuclear deterrence and other capabilities China uses for deterrence, they need to communicate certain aspects of these capabilities. The objective is to avoid using them while using the threat of their use to shape adversaries’ behavior. You must disclose certain information about your capabilities, usage plans, and force organization to make those threats credible. Otherwise, you won’t get the deterrent value from the resources spent to develop them.

I acknowledge there are certain gaps that open-source researchers like me can never fill. What’s in China’s actual war plans? What are the exact specifications of their weapon systems? What are the precise details of certain PLA organizations?

There are also fundamentally unknowable elements: what was in Chinese decision-makers’ minds when they made certain choices? What choices will they make in a crisis over Taiwan? How will the plans I tried to understand through doctrinal debates actually function during wartime? Some things you can know, some things you can learn from open sources, and some things remain unknowable even with the best intelligence.

Jordan Schneider: Mike, would you like to add to that?

Mike Horowitz: This is tremendously challenging research. From a research design perspective, you face enormous uncertainty about China’s military. There’s a premium on gathering as much information as possible, especially given the challenges involved in obtaining that information.

There’s a justified premium placed on those talented scholars who, despite all constraints, can find primary source documents and information. Even if incomplete, even if different from what you would get researching the UK’s or US military, you can still gather enough information to make substantive inferences about how China’s military might behave.

Jordan Schneider: The PLA does have to talk to itself, but it doesn’t necessarily have to talk to Fiona Cunningham.

Mike Horowitz: It certainly doesn’t talk to me. Actually, I was at a conference right after I left the administration and saw somebody from China, which was probably the first time I’d seen somebody from China in years due to COVID and then being in the Defense Department. The person approached me, and I said, “Hi, it’s nice to meet you. My name is Michael Horowitz.” They replied, “Hi, I know who you are. We read all of your things.” I thought, “Well, that’s scary,” and wanted to back slowly away, much like the Homer Simpson meme.

Fiona Cunningham: Michael speaks the truth. I remember picking up a set of journals in one of the university libraries while looking through recent PLA research. There were many articles on lethal autonomous weapons and military AI applications. Many contained English language footnotes citing Michael Horowitz in the Texas National Security Review. I realized I had colleagues influencing how the PLA thinks about important questions at the cutting edge of warfare. They don’t cite me, though.

Jordan Schneider: Let’s stay on that topic. It’s fair to say that Chinese defense analysts constitute a much larger industry studying America than Western defense analysts studying China. Thoughts on that?

Fiona Cunningham: Several factors contribute to this imbalance. First, the US military is recognized by the PLA and most militaries worldwide as the premier fighting force. If you’re trying to learn from the best, you’ll dedicate a lot of resources to studying what makes the US military so effective, including how America leverages its technological and other advantages.

For China, the US is not just a potential opponent in their most important conflict scenario but also the model of military excellence. They have two compelling reasons to study the United States. Michael can probably elaborate, but I believe the United States learns less from China as an example of how to do things differently and is more interested in studying the PLA to understand how one of its two most serious potential opponents might fight in the future.

Mike Horowitz: That’s certainly correct, but the US does spend enormous time and resources studying everything we can possibly find about China’s military, though from a different perspective than learning lessons.

One challenge is that China hasn’t fought a war since the late 1970s. If we’re considering what lessons the US would learn from China in a military sense, these wouldn’t be battlefield lessons since there’s no recent battlefield data. The lessons would instead concern capability development, acquisition approaches, or force structure concepts — if you could access their doctrinal manuals.

This differs from how the world is currently studying Russia and Ukraine, intensely examining everything about battlefield lessons and what they mean for the future of warfare. The US is trying to glean everything possible about China’s military, but it’s learning lessons in a fundamentally different way.

Fiona Cunningham: The language barrier is also significant. In China, most people learn English from elementary school, making it a very common second language. This removes a major barrier for Chinese analysts studying the US military.

In the United States, or my native Australia, studying Chinese isn’t everyone’s first choice for a second language because of its difficulty. There are fewer college or high school students learning Chinese now than 10-15 years ago, as the incentives have shifted from business to national security.

Beyond language, there’s the issue of available materials. The United States publishes extensive open-source information about its military thinking. There are types of materials I would love to see in Chinese about the PLA that simply aren’t available, while equivalent US materials are accessible to Chinese researchers. I’d particularly value joint publications that would give me confidence that the doctrinal debates and materials I’ve studied accurately reflect actual PLA planning.

Jordan Schneider: Let’s start with the period from the 1960s to 1984 very briefly. China experienced the Sino-Soviet split, fought some border conflicts with the Soviet Union, and nearly engaged in nuclear war. That remained the main focus for a long time — defense in depth, learning from the Soviets.

Fast forward to 1984, when Deng realized Gorbachev was serious and wouldn’t invade China anytime soon. They could reprioritize defense relative to other national priorities. Take us from there through the period when global nuclear disarmament was being considered, up to 1995-1996.

Fiona Cunningham: This is a fascinating period for Chinese defense strategy, particularly regarding nuclear weapons’ role in China’s national defense.

Starting in the mid-1980s, China downgraded the possibility of fighting a major war against one of the superpowers. This coincided with China’s economic opening and reform gathering momentum, prompting the government to redirect resources from national defense to economic development.

China also reassessed what conflicts it might face. Prior to 1988, they prepared for general war with a superpower potentially invading China. After 1988, they shifted focus to “local wars” related to territorial disputes on China’s periphery. Taiwan was only one of several potential conflicts, with uncertainty about whether the US would become involved.

Two interesting developments emerged during this period. First, China debated the role of nuclear weapons in local wars. In conventional invasions, the role of nuclear weapons seemed clearer, even though China didn’t plan to use them first to deter conventional attacks. But their utility in local conflicts was less obvious.

Harvard professor Alastair Iain Johnston wrote a landmark article about the concept of “limited nuclear deterrence,” which was being debated within the PLA in the late 1980s through the 1990s. This concept involved acquiring tactical nuclear weapons to create more options on the escalation ladder during conventional conflicts.

Since Johnston’s influential article, more sources have become available, revealing another perspective in China’s nuclear strategy debate — waiting to see what the Soviets and Americans would do, as global nuclear disarmament might make additional nuclear investments unnecessary. By 1992, China’s leaders concluded they still needed nuclear weapons because the US and Soviets wouldn’t eliminate theirs completely, but China didn’t require large numbers.

Another interesting development arose from financial pressures on China’s military, which turned to exports to sustain its defense industrial base. They developed conventional short-range ballistic missiles intended for Middle Eastern markets. When the US pressured China not to export these weapons, China’s previously nuclear-only missile forces saw an opportunity to find a role in local wars by using short-range ballistic missiles to threaten Taiwan. This was the contingent origin of China’s conventional missile force, which presents a big challenge for the US and its allies today.

Jordan Schneider: Qian Xuesen 钱学森, who most listeners will be familiar with, deserves a mention here. He was forced out of America during the McCarthy era and became the father of China’s missile program. He remained skeptical in the 80s. You have a great quote where he warned that reports heralding a “post-nuclear era” were “deceiving people and they are all false.” The man still wasn’t buying into the new world order decades after seeing the West for what he believed it really was.

Another thing worth emphasizing is the level of military downsizing during this period. They shed almost a million people from the PLA — a dramatic reorientation for any military.

Fiona Cunningham: I don’t cover this in tremendous detail in my book because others have examined it thoroughly, but China’s conventional military modernization accelerated after the Gulf War, when China developed a clearer understanding of what they would need to do.

Mike Horowitz: They observed what the US accomplished.

Fiona Cunningham: Exactly — they saw what the US did and realized that to fight future conventional wars, they needed to develop similar capabilities. This became a decades-long project for the PLA. The decision to change conventional military strategy to enable China to fight “local wars under high-technology conditions” — their strategic guideline — was inspired by watching US operations rather than perceiving a direct threat from the United States. Taylor Fravel’s book Active Defense does an excellent job of explaining that decision. It was a very influential book while I was writing my dissertation.

Jordan Schneider: Mike, can you give us the 101 on why the 1991 Gulf War was so mind-blowing to so many people?

Mike Horowitz: The 1991 Gulf War blew everyone’s mind because it revealed the “second offset” on the public stage. All those developments in stealth technology, precision strike, and advanced weapons — things we take for granted today — made their dramatic debut.

I was in middle school when that war happened and remember seeing images of precision strikes on green-screen displays — missiles hitting specific buildings in Iraq. It seemed like magic, the ability of the US military to strike targets so precisely. This capability shocked the rest of the world.

The Soviets, who were becoming Russians at that point, understood the concept but had been unable to execute it. This was fundamentally different from how people thought wars would happen toward the end of the Cold War. It demonstrated sheer technological superiority by the United States and served as a wake-up call to the PLA.

Jordan Schneider: It’s also important to mention that the US even surprised itself with its effectiveness. There were advanced projections about the casualties required to conquer Iraq. The Senate, when Congress was debating whether to authorize war, anticipated 50,000 American casualties. It turned out to be only in the three or four figures.

Mike Horowitz: You’re referring to 2003. I’m talking about 1990-91. However, in both invasions, the projected dangers to US soldiers were dramatically overestimated. In retrospect, much of that was due to operational art (force employment and the effective use of military power) and how much the American military excelled. But equally critical were defense technology breakthroughs and the ability of the American military to integrate them in ways that shocked the rest of the world.

Jordan Schneider: China realized they needed a new playbook, understanding it would be a decades-long effort to approach the capabilities of the world’s superpower in the early 90s. Fiona, take us to the 1995-96 Taiwan Strait Crisis. What triggered it, and how did it affect China’s conception of what it needed to be nationally secure?

Fiona Cunningham: The Taiwan Strait Crisis had two peaks. The first occurred in mid-1995, with another flare-up in early 1996. The initial trigger was the United States issuing a visa to Taiwan’s first democratically elected president, Lee Teng-hui, whom China viewed as agitating for independence. China saw this visa as representing American support creeping toward Taiwanese independence.

The second peak came in March 1996 when President Lee sought re-election. China attempted to influence the outcome of that election. On both occasions, China conducted military exercises across and around Taiwan, including launching some of the short-range ballistic missiles they had fortuitously acquired earlier.

Mike Horowitz: “Fortuitously acquired” sounds like a euphemism for China beginning a military buildup after the Cold War.

Fiona Cunningham: They “fortuitously acquired” these capabilities because China’s defense industry had been instructed to generate revenue. When the US blocked the export of these weapons, they became available for domestic use.

There’s an interesting sequence of events in the late 1980s. The rocket force, which at that time only operated nuclear weapons, was called into meetings with China’s leaders who asked, “What role will you play in a local war?” They responded, “We have a great idea — we’ll be armed with conventional missiles.” They put forward this proposal in the late 80s largely for organizational survival purposes. They needed a way to remain relevant in a changed environment where most of China’s potential conflicts wouldn’t involve nuclear weapons. This demonstrates classic organizational incentives for military branches to seek new roles when the threat environment changes.

Between the exercises in 1995 following Lee’s Cornell visit and those in 1996 aimed at influencing Taiwan’s presidential election, China began its five-year defense plan. In this plan, conventional missiles and what they called “Shāshǒujiǎn” 杀手锏 (Assassin’s Mace, “trump card”) weapons, primarily missile systems, received prominent attention.

A series of leadership meetings occurred from late 1995 onward after the first set of exercises. From these meetings, we can surmise that China’s leaders began to see what the rocket force had recognized in the early 1990s — these missiles could be powerful tools for intimidating adversaries and addressing a new strategic reality. If another Taiwan Strait crisis occurred — as it did in 1996 — China wouldn’t have Gulf War-equivalent conventional military capabilities to counter the United States. This would make it much more difficult for China to use force if they felt their red lines regarding Taiwan were crossed.

These missiles provided coercive leverage — a way to threaten escalation against a powerful nuclear and conventional adversary when China had few other options. Interestingly, China’s leaders determined that threatening nuclear first use, the other obvious option, was unacceptable to them.

Jordan Schneider: Let’s stay on that topic. What is the “no first use” policy? Where does it come from? How does it constrain Chinese doctrinal thinking?

Fiona Cunningham: I’ll make my best case for why I believe it operates as a constraint on the PRC even today, five or six years into a nuclear buildup.

The “no first use” policy originated in a statement accompanying China’s first nuclear weapons test in 1964, where China pledged it would not be the first to use nuclear weapons. They stated they would only use nuclear weapons if first attacked with nuclear weapons by another country. This policy converted Mao’s views about nuclear weapons into a formal stance that later became the strategy given to the Second Artillery (China’s missile forces) when they began formulating how to implement China’s nuclear weapons strategy almost a decade later.

Before 1964, many statements from Mao and other Chinese leaders indicated they needed conventional weapons for conventional conflicts and nuclear weapons to deter nuclear weapons use. Several reasons explain why China adopted this policy during the Cold War. It differentiated China from the nuclear superpowers, perhaps making its nuclear weapons less of a challenge to the Soviet Union and the United States. It also related to China’s geography — they didn’t need nuclear first use against a conventional invasion because China’s size would exhaust any invader. China could survive a conventional conflict, but a nuclear conflict presented a different scenario.

It’s worth noting that China’s nuclear policy isn’t just a military matter — it’s a civilian policy given to the military by top leaders. Military leaders cannot change China’s “no first use” policy. Because it originated with Mao and Deng, it became orthodoxy that’s difficult for even civilian leaders to change.

Looking at available doctrinal materials — though unfortunately we lack probative information for the last two decades — it’s clear the “no first use” policy constrains how China plans to use its nuclear weapons. However, there have been debates about changing it or placing conditions on it, and questions remain within China about whether other countries see it as credible.

Mike Horowitz: All that can be true. You’ve persuaded me that China’s military believes it is constrained by the “no first use” policy and that civilians must make those changes. In a crisis, depending on the stakes — particularly if Xi thought regime survival was at risk — there could be incentives for China to strike first with nuclear weapons. The question is, to what extent would the “no first use” policy constrain China’s military in a conflict where the civilians who can make policy changes are actively engaged?

Fiona Cunningham: Several points are worth noting. First, if you plan to use your capabilities in accordance with a policy like “no first use,” your ability as a rocket force to develop operations involving first-use options becomes constrained. That doesn’t mean it’s impossible, but...

Mike Horowitz: It’s not as simple as just saying “launch the missile."

Fiona Cunningham: Exactly. It would mean launching the missile outside established protocols or training. It’s possible, but would represent a big departure from doctrine.

In my book, I detail an intense debate over adding conditions to China’s “no first use” policy in the early 2000s. These debates are fascinating because we see writing in leaked Chinese missile force teaching materials discussing lowering the nuclear threshold and nuclear deterrence signaling operations. What they don’t include, however, is a nuclear first-use campaign that would follow if an adversary didn’t back down after these signals.

Mike Horowitz: That’s interesting.

Fiona Cunningham: I found a dissertation written by a missile force officer — now I believe a deputy base commander — who essentially asks what a first-use campaign would actually look like. One option he discusses is detonating a nuclear weapon in space if a conflict over Taiwan isn’t going well. Interestingly, China subsequently developed non-nuclear anti-satellite weapons.

However, official materials produced by the Second Artillery don’t describe what warning shots or nuclear first-use actions would follow these threats. We lack visibility into current debates, though other scholars have found evidence of the Rocket Force now exercising for “launch on warning.” In my decade of interviews in China about nuclear strategy issues, I noticed a change in tone between 2014 and 2016, with some experts beginning to question whether launch on warning violates a “no first use” policy.

Jordan Schneider: What is “launch on warning,” Fiona?

Fiona Cunningham: In its simplest form, launch on warning means launching once you receive warning that your adversary has launched nuclear weapons toward you. Rather than waiting until those weapons detonate on your territory, you choose to launch your nuclear weapons during the period between your adversary’s launch and the detonation on your territory.

Mike Horowitz: Can we discuss this in relation to China’s current nuclear modernization? What you’ve described presents a paradox. Launch on warning is generally considered when you’re worried about your nuclear forces being decapitated — when an enemy’s first strike might destroy your ability to retaliate. If you have secure second-strike capabilities, you can absorb a strike and still respond. Since China has been rapidly modernizing its nuclear arsenal, that should make them more secure rather than less. Why would the PLA be thinking more about launch on warning now as their nuclear capabilities become more sophisticated?

Fiona Cunningham: It depends on which capabilities deliver your secure second strike. If you rely on road-mobile land-based missiles or submarine-based ballistic missiles — nuclear missiles mounted on trucks driving around China’s vast territory, perhaps hiding in caves — you’re depending on mobility and concealment to avoid detection by adversary satellites.

If, instead, your path to survivability is through having systems that are easier for adversaries to locate, the mindset about securing a second strike differs. If your plan involves many fixed silo-based missiles that your adversary can see, you’re relying on the fact that they can’t destroy all of them due to their quantity. However, those silos contribute little to second-strike capability unless you plan to launch the missiles before enemy weapons hit them.

The launch-on-warning approach relates partly to which capabilities comprise your secure second strike, but also concerns deterrence immediacy. Consider the effectiveness of deterrence if you wait weeks to pull road-mobile missiles from caves before retaliating against Washington, D.C.

Mike Horowitz: You maintain a hair-trigger posture to increase the credibility of your threat from a signaling perspective and increase the probability of successful deterrence.

Fiona Cunningham: What surprised me in post-book interviews about China’s nuclear modernization is that Chinese experts now acknowledge the US’s improved capabilities to detect road-mobile missiles. Examining the sensors the United States plans to deploy in space, Chinese strategists reasonably worry that the US will better locate capabilities they previously considered secure and survivable.

Jordan Schneider: America had approximately 20,000 nuclear weapons toward the Cold War’s end. Many have been decommissioned, but we still maintain high four-figure numbers. China started with double-digit quantities and is now approaching an arsenal of 400-500. This presents a challenge when you lack nuclear submarines, bombers, and ICBMs to ensure national destruction capability. The concept of playing three-card monte with tens of millions of lives by driving missile trucks around rural China is simultaneously absurd and terrifying.

Mike Horowitz: Road-mobile weapons are extremely difficult to detect. This explains why every country that has faced the United States since the beginning of the Cold War — whether Iraq during the Gulf War, the Soviets, China, or potentially North Korea — has considered mobility as a solution to overwhelming American firepower.

Fiona Cunningham: Such an approach wouldn’t work in the American hinterland, but China presents a different scenario.

Mike Horowitz: The regulatory environment in the US would make that impossible.

Jordan Schneider: Perhaps this could become our trucking industry’s future. Once autonomous vehicles replace truckers, the only humans on the road might be those randomizing their drives around Nebraska.

Fiona, let’s close the loop on Chinese nuclear military modernization.

Fiona Cunningham: My book examines how nations pressure adversaries in conventional wars by threatening escalation. There’s considerable discussion about whether China’s nuclear modernization aims to create this option with its nuclear forces, especially as it develops more theater-range precision options — capabilities approaching what would be needed for credible nuclear first-use threats.

However, my conclusion is that China hasn’t necessarily decided that its nuclear modernization will replace non-nuclear capabilities as its primary source of coercive leverage in conventional conflicts. I see the drivers of China’s nuclear modernization as primarily about achieving a more robust second strike against the United States.

There’s also a political leverage component — China wants enhanced capabilities because they believe this will make the United States behave more prudently. This doesn’t necessarily involve precise calculations about nuclear exchanges or specific force posture changes to enable credible nuclear first-use threats.

My recent interviews in China suggest a psychological and political leverage rationale for modernization that doesn’t necessarily translate to posture changes that make first use more feasible or improving China’s position in a nuclear exchange. Different schools of thought within China point to different modernization rationales, but first use doesn’t clearly emerge as one of them.

Jordan Schneider: The idea that it’s worth investing to ensure nobody in the Pentagon believes they could execute a first strike on China without America suffering consequences seems reasonable. If I were a captain in the Rocket Force, I’d consider that a worthy investment. But how far that extends and what it means from a readiness perspective remains one of those “unknown unknowns” we discussed earlier.

Mike Horowitz: This might transition us to discussing conventional capabilities. What’s remarkable about China’s military development over the last decade is that they’ve rapidly modernized in every area. There was a time when we could say, “China is prioritizing this capability over that one.” Fiona’s book brilliantly discusses this earlier period when China was making some of those choices, particularly regarding conventional missiles. Now, however, the story of the last decade is essentially “everything everywhere all at once” from a Chinese military modernization perspective.

Modernizing Warfare

Jordan Schneider: I love how your book opens in the early 90s, when China essentially said, “We don’t have computers, so we’re resilient to cyber attacks.” Then we witness the informatization of warfare and doctrinal development regarding offensive cyber capabilities to threaten targets when nuclear warfare threats aren’t credible.

Fiona, give us a brief history of cyber weapons in China. How has the PLA’s thinking about their utility evolved?

Fiona Cunningham: The major turning point for offensive cyber capabilities in China — capabilities they considered using for coercion — came after the 1999 US accidental bombing of China’s embassy in Belgrade. From China’s perspective, this incident was anything but accidental. There’s a notable quote from Jiang Zemin where he basically says, “I’m really indignant. This is not a trivial event. This is a big deal, and the Chinese people cannot be bullied."

Jordan Schneider: This presents an irony, right? In 1991, everyone marveled, “These Americans can hit a window from 3,000 miles away.” Then, suddenly, we accidentally bombed the Chinese embassy. Where do you stand on the conspiracy theories about this incident, Fiona?

Fiona Cunningham: My perspective is that even the best organizations can make mistakes, including the US military. I think China’s reasoning was, “This is a highly advanced military — we don’t believe they could have made such a mistake.” Additionally, once a top leader decides something wasn’t a mistake, that becomes the official position, even if contrary evidence emerges later. It becomes difficult to undo those narratives.

Following the Belgrade embassy bombing, a series of meetings took place where Jiang Zemin instructed China’s military leaders to develop capabilities addressing their leverage deficit — something that would make future conflicts, particularly across the Taiwan Strait, too dangerous and risky for US involvement. They concluded their plan to build a massive conventional missile force wouldn’t be sufficient.

Offensive cyber operations emerged as a promising capability, partly because China’s military studied the Kosovo air war and observed Serbian militias defacing NATO web pages — not particularly sophisticated cyber activities, but they recognized how quickly and inexpensively one could create problems for a more advanced military and society.

China’s awareness of cyber operations dates back to the Gulf War, when they began studying what they called “computer virus weapons” and noticed US interest in these capabilities. However, they didn’t actively pursue this option until after the Belgrade embassy bombing. Subsequently, they allowed “a thousand flowers to bloom” (百花齊放), with everyone of significance in the PLA entering the cyber arena. This resulted in capability development on both offensive and espionage fronts, but it lacked coordination and organization.

This disorganization became problematic for China’s leaders around 2010. They observed other countries using cyber operations for military effects that demonstrated the potential they had recognized. Simultaneously, China was becoming increasingly dependent on information networks for military, social, and economic functions. Seeing these cyber domain developments, they instructed their military to adjust their plans and organization for offensive cyber operations, acknowledging that they would be “throwing stones from a glass house.”

Jordan Schneider: What happens next?

Fiona Cunningham: A process of changing China’s cyber doctrine began around 2012. The PLA held many meetings to evaluate its progress and future direction. These discussions produced revealing statements, such as “We’ve made significant progress in capabilities, but everyone is fighting their own war, so we lack coordination.” Another noteworthy comment was: “If we aren’t careful with how we plan and execute these operations, we’ll harm the national interest — yet we still want to use them for leverage purposes."

Over approximately two years, China’s PLA developed approaches to modify their cyber doctrine. Around 2014, Xi Jinping particularly pressured the PLA to reorganize. This coincided with the US indicting several PLA officers for industrial espionage. The following year, China released a white paper acknowledging their cyber defense capabilities for the first time, though they still refused to publicly admit possessing offensive capabilities.

Subscribe now

At the end of 2015, China established a new organization called the Strategic Support Force, which consolidated all the disparate elements of the PLA’s cyber enterprise into a single entity. The success of this consolidation remains unclear — in fact, China disbanded the Strategic Support Force just as I was completing my manuscript.

Jordan Schneider: Toward the end of this first period, the Obama administration attempted to engage Xi and establish norms about appropriate cyber activities. How does this international dimension interact with the Chinese government’s directives to its ambiguously organized cyber force?

Fiona Cunningham: One fascinating aspect of China’s cyber capabilities is that for most capabilities with direct strategic effects on adversary decision-makers, China’s top leaders never delegate authority to lower-level PLA commanders regarding their use. The only area where some de facto delegation occurred was in the cyber arena before the 2014-15 changes I outlined.

The activities the Obama administration complained about — PLA members hacking US corporations to obtain proprietary information for commercial advantage — appeared to occur without China’s top leaders’ authorization or oversight. The international pressure reinforced China’s trend toward consolidating their cyber forces and subjecting them to tighter oversight and command and control from the top. In some ways, the Obama administration was pushing on an open door.

For industrial espionage that supported China’s national priorities in certain sectors, the activities continued but shifted to non-PLA actors, while the PLA focused on military missions. The administration was pushing on an open door in seeing some activity decline. If you examine the timelines, major decreases in PLA cyber activity tracked by organizations like FireEye (a private cybersecurity company) coincided with Central Military Commission meetings deciding on major military reform packages.

It’s difficult to disentangle whether US threats of sanctions over cyber espionage or broader PLA reorganization trends were more influential. However, this suggests US pressure alone didn’t cause China to change its behavior.

Jordan Schneider: We’ve witnessed many high-profile actions by the Chinese government over the past decade-plus, but we haven’t seen something comparable to Stuxnet, Russian cyber attacks on Ukraine’s power grid in 2024, or the Viasat attack in 2022 — incidents where actors not only penetrated critical infrastructure but actively attempted to disrupt operations. Fiona, is this simply because China hasn’t engaged in a major war, hasn’t faced something as pressing as Iranian nuclear capabilities, or is there a broader hesitance to cause damage rather than just steal information, conduct espionage, and prepare groundwork for potential future conflicts?

Fiona Cunningham: The PLA is certainly willing to lay groundwork for disruptive operations. In late 2023, reports emerged about the Volt Typhoon intrusions, where Chinese actors — though public attribution to specific PLA units is limited — penetrated adversary critical infrastructure. This aligns perfectly with their doctrine of infiltrating an adversary’s critical infrastructure to pressure them in peacetime, crisis, or conflict — shaping their behavior through fear of system disruption.

What China hasn’t done, however, is demonstrate that if they wanted to “hit send” on these capabilities, they could effectively disrupt an adversary’s critical infrastructure. I wish I had an interview where someone explicitly explained their reasoning, but these questions are extremely difficult to raise within China. Even if you could ask, individuals might not know the answers.

This remains a puzzle, even in US analysis — why hasn’t China conducted a major public demonstration proving they can cause disruption in ways that would concern adversaries in future conflicts? The reasons likely align with China’s desire to control these capabilities very carefully to avoid blowback. Nevertheless, countries typically want to demonstrate their capabilities, especially with something as uncertain as offensive cyber operations.

Mike Horowitz: Shout out to the names that people give to things. In the official US Government release on Volt Typhoon, they also called that group Vanguard Panda, Bronze Silhouette, and Insidious Taurus, which is unbelievable.

Jordan Schneider: Are these American code names, or do the groups call themselves like Golden Panda?

Fiona Cunningham: I don’t know.

Mike Horowitz: It’s like “AKA Vanguard Panda, Bronze Silhouette.” There are a couple other names that are less funny, like “Dev-0391.”

It’s less puzzling to me why China hasn’t tried to break things. I view what China is doing in this context as building the capacity to break things. The demonstration that they are in US and other countries’ networks is a signal of capabilities, since the actual use of cyber tools to break things is pretty rare. Stuxnet is the overcited one-off in a way, and relied on a lot of very specific factors.

China demonstrated that it was deep in energy infrastructure and water infrastructure at the state level across the United States, which certainly illustrates the capacity to destroy — especially because of the way they gained access through routers, VPNs, and many common electronic devices that lots of critical infrastructure facilities and Americans have.

This points to one of the issues surrounding cyber capabilities in general and relates to what we discussed last time, Jordan, in the context of offensive cyber strikes in a world of advancing AI. There’s sometimes a tendency to think about offensive cyber as this magic thing you can press “go” on, when in reality, because the accesses are so limited and once you use them they can disappear, the incentive structure even for very competent offensive actors like China is often to hold back on breaking things. The more you use these capabilities, the more it becomes all hands on deck to completely knock you out.

I have no doubt that even in the aftermath of the Volt Typhoon revelation, China has other access points that we don’t know about. From a parochial American perspective, that’s incredibly dangerous. The questions become: under what circumstances would China try to activate that, and what would the impact be? There’s uncertainty surrounding both of those things.

Fiona Cunningham: That uncertainty is one of the reasons why it makes the threat to use them credible. It may be a total fizzle and a flop, but it might also be really bad.

Mike Horowitz: I agree completely.

Fiona Cunningham: What’s interesting to me is that China’s behavior with Volt Typhoon runs contrary to a lot of US academic discussions about the utility of cyber operations over the last five to eight years. It does raise these questions about strategy. US Cyber Command has been saying the real strategic value of cyber operations lies in the “death by a thousand cuts” under the threshold of armed conflict. But the PLA looks like it’s preparing for this “cyber Pearl Harbor” scenario.

Do you actually get the leverage that China thinks it will get from preparing for something the US has said it doesn’t really see as being that big of a problem?

Mike Horowitz: It depends on what they can actually do. Consider some of the things that Russia has allegedly done and how they’ve disrupted Americans. Part of this depends on what the goals are.

There’s “death by a thousand cuts,” where cyber is an enabler to other kinds of operations. This might say something about some of the academic literature on cyber, but we don’t need to discuss that now.

There’s also cyber as a disruption and illustration of possibilities for cost imposition, precisely because there’s uncertainty. It might fizzle, but it might be really important. The theory would be that if you could create a little disruption to the lives of average Americans during, say, a Taiwan crisis, it would bring the costs home in a different way.

The question becomes, how much disruption, and what would be the ultimate impact? If what you’re trying to do is influence American behavior — perhaps more so than affecting US military capacity in the Indo-Pacific — and you’re trying to influence public attitudes and perceptions, no one really knows how that would go, including us and including them.

Jordan Schneider: That’s a key factor with a lot of this. I like your framework of brinkmanship — what are we doing with our force posture? Are we engaging in brinkmanship? Are we doing calibrated escalation? Or are we actually preparing to fight a war?

The problem is that the more forward you are with these capabilities, the more likely you are to get America to take it seriously. When China starts taking cyber seriously, as we see with Volt Typhoon, that leads to more awareness, investments, and local water plant owners updating their systems.

The same applies in the Taiwan context. Do you really think that turning off the lights in Texas is going to turn out better for Beijing? This is almost like Japan’s 1941 logic — “They’ll really want to make peace with us after we bomb Pearl Harbor."

It could go both ways, depending on how focused you are, what timelines you’re working with, and how “feminized” you think America is — to use Putin’s term, not mine. These factors impact the way you’re going to think about what you show and what you don’t do on the world stage, which makes this all very tricky.

Mike Horowitz: Just to be clear, countries have made that mistake about the United States forever. Saddam Hussein made similar assumptions. Bin Laden did as well. Foreign powers often forget that the US can be relentless. When someone attacks the US hard, we respond with full force and persistence.

Arguably, those who understand that we’re somewhat unpredictable might find that beneficial for deterrence in the classical sense. I’m curious about Fiona’s thoughts on how China perceives us in this context. I encountered this topic frequently in defense conversations over recent years. Does China assume that if they attack the US, we’ll respond with everything we have? Or do they believe they could cause enough pain that the US would back down?

Fiona Cunningham: Three factors likely matter — timing, stakes, and nuclear weapons.

Regarding timing, if you strike the US hard during an active conflict, the consequences would be severe because the United States would be fully committed. However, if that threat exists before or during a crisis, it might encourage caution and prudence. China’s strategic deterrence approach encompasses wartime planning but is also designed to influence US behavior during peacetime.

From China’s perspective, there’s an imbalance in what’s at stake. Though not explicitly stated in much of the literature I’ve studied, Taiwan represents an immediate, tangible interest deeply connected to the Communist Party’s sense of mission and legitimacy. While China can tolerate the current situation of de facto separation as it has for years, for the United States, Taiwan is an island without a formal treaty alliance, far from the US homeland, and not essential to America’s territorial integrity. The outcome of any Taiwan conflict would matter more to China. The US can have strong interests too, but they’re more diffuse and indirect, whereas China’s interests are specific and direct — Taiwan represents unfinished business from the Chinese Civil War rather than America’s global position and alliance structure.

Finally, nobody wants to fight a nuclear war. Many of China’s information-age weapons — non-nuclear weapons with strategic effects — are designed to push the US to the threshold of nuclear weapons use and then call its bluff. The open question remains whether America’s seemingly unpredictable nature extends to nuclear weapons use. That’s the million-dollar question.

Jordan Schneider: Another point supporting that perspective is that we’ve now seen two consecutive presidents take Putin’s nuclear threats seriously, even though those threats were less credible and not directly aligned with major US strategic interests. This clearly impacted Biden’s calculations. With Trump, throughout his campaign and in his meeting with Zelensky, he repeatedly warned, “You’re flirting with World War III.” It’s one thing to exhibit Jacksonian intensity when fighting overseas wars, but it’s entirely different when you genuinely fear that Los Angeles could be destroyed.

Mike Horowitz: There’s something interesting about how presidents conceptualize nuclear war and risk. Looking at Trump’s public comments, they suggest he might be even more concerned about nuclear risk than some other presidents. While every president since Truman has worried about nuclear war, Trump seems particularly focused on the dangers of nuclear escalation.

Jordan Schneider: It will be interesting when we eventually see a generational change — a president who didn’t grow up during the 60s, 70s, and 80s when nuclear concerns were front and center. But returning to the PLA — Fiona, let’s discuss counter-space capabilities. We’ve witnessed several interesting and explicit demonstrations of these capabilities from China in recent years.

Fiona Cunningham: I should share a personal anecdote — one reason I became interested in writing this book was China’s anti-satellite weapons test in 2007, when they destroyed an aging weather satellite with a conventional missile. I was in college at the time and remember seeing it on the front page of the Sydney Morning Herald, wondering what it meant. The book actually opens with this ASAT weapons test, perhaps for that unwritten reason.

China pursued counter-space weapons capability for coercive purposes — to exert pressure during a potential conventional conflict with the US over Taiwan, particularly after the Belgrade embassy bombing. This decision was difficult to pinpoint precisely, but based on Jiang Zemin’s speeches, it was likely made between late 2000 and late 2002.

China recognized that anti-satellite weapons could disrupt an adversary’s space capabilities and impede the US military’s long-term objective of achieving space control. They developed a range of counter-space capabilities, including lasers that could dazzle optical sensors on US satellites. In fact, they dazzled a US National Reconnaissance Office satellite around 2005, then tested the missile that destroyed a satellite and created substantial debris, generating international criticism.

Their goal was to develop various weapons that could attack US satellites in orbit and disrupt data transmission between those satellites and Earth. These range from non-kinetic reversible effects, such as lasers and electronic warfare jamming, to irreversible effects like completely destroying satellites. More recently, China has developed co-orbital capabilities — satellites that can maneuver to grab onto other satellites, tow them elsewhere, collide with them, or position themselves close enough to jam or dazzle from proximity.

What’s curious is that China didn’t pursue counter-space capabilities earlier. There’s a common perception that China identified attacking US satellites as a valuable coercive tool following the Gulf War. However, when examining the sources, China didn’t begin contemplating counter-space attacks until the late 1990s, after the 1995-96 Taiwan Strait crisis. Before that, in the context of the Gulf War, China primarily recognized that to fight like the United States, it would need its own military satellites to guide weapons and confirm target destruction.

This desire to build military support capabilities in space constrained China’s approach to counter-space weapons. They consistently emphasized that hostilities in space should be limited, unlike their apparent willingness to consider more extensive use of conventional missiles. That’s the doctrinal capability narrative that has emerged.

Jordan Schneider: The pattern is fascinating — China demonstrates a capability, America becomes alarmed and develops countermeasures. This is the challenge with sub-nuclear capabilities — the more you reveal, the more your adversary adapts. By the end of both the cyber and counter-space chapters, you quote Chinese analysts essentially saying, “This didn’t achieve exactly what we wanted. America now has ten times more capabilities than we anticipated a decade ago.” What are your thoughts on this dynamic, Fiona?

Share

Fiona Cunningham: While this dynamic matters, it may not be as important as we might think. Many of these capabilities were designed to give China credible leverage when it was considerably weaker than it is today. Even if these capabilities become less effective over 20-30 years, they still provided China with methods to deter and coerce the United States when China’s conventional capabilities posed minimal threat.

Currently, I don’t believe China’s conventional capabilities threaten decisive victory against the United States in the Indo-Pacific, but they’re substantially improved compared to 20-30 years ago. The US military would face much greater challenges fighting China today than it would have two decades ago.

China’s pursuit of these capabilities has pushed the US to develop countermeasures, which in some ways validates China’s strategic choices — these capabilities delivered on their promise by forcing the US to reconsider how it approaches conflicts and organizes capabilities across different domains. The cyber changes we discussed earlier exemplify this. Additionally, the US withdrew from the Intermediate Nuclear Forces Treaty partly due to China’s conventional missile force growth, creating new questions for China about how US intermediate-range conventional missiles in the region might affect its security.

Similarly, China’s counter-space capabilities have compelled the United States to rethink its satellite architecture. The US formerly relied on “big juicy targets” — large satellites for sensing and signals intelligence that supported conventional conflicts and precision strikes. Now the United States deploys smaller, less vulnerable satellites to complement these larger assets.

Mike Horowitz: Or haven’t yet.

Fiona Cunningham: Right. This adaptation means China faces many more targets if it wants to use counter-space capabilities effectively. It can no longer neutralize US capabilities by destroying just a handful of sensing satellites. Furthermore, the United States now discusses its own counter-space capabilities much more openly.

China has triggered a big reaction in the space domain. Chinese strategists are now concerned about several US initiatives, including distributed satellite constellations, US counter-space capabilities, satellites that monitor activities in orbit, and US partnerships with commercial space companies that enhance American capabilities. They’re also increasingly worried about the potential for nuclear escalation resulting from space-based conflicts — a new development over the past five years.

Mike Horowitz: Another way to think about this more broadly is the reveal-conceal dilemma when developing capabilities. Everything Fiona said is correct. Generally, you reveal capabilities to deter or coerce, while you conceal capabilities for actual conflict. China’s actions in space represent revelation specifically to impose costs, which they have accomplished by driving many changes to US space architecture — both implemented and planned.

However, some areas are difficult to conceal even when that’s the intention. Space is arguably one such domain — concealment is easier in some ways but harder in others. We don’t know exactly what China’s intentions are when they do things like dazzle our assets. Are they purposely revealing capabilities, or would they prefer to conceal these tests but can’t? Their testing options differ from terrestrial missiles, which can sometimes be tested in relative isolation. This fundamentally changes the dynamic.

Fiona Cunningham: To add one more element, PLA space deterrence doctrine often describes a tiered approach. You begin by signaling your capabilities or repositioning assets to make them visible. Mike’s point is particularly relevant because even when examining PRC doctrine, you can see there’s a strategic place for movement and revelation of capabilities intended to send a message. Determining whether they’re following established doctrine or simply couldn’t hide what they were testing remains a key challenge.

Jordan Schneider: Let’s discuss dogfighting. Mike, would you like to address this?

Mike Horowitz: Recently, a US Space Force official published an article discussing China conducting “dogfighting” in space. This terminology demonstrates how we project our understanding of one domain onto another. However, it clearly illustrates what Fiona’s book demonstrates about China’s capabilities and their testing approaches.

Despite changes over recent years, the US Space Force still feels constrained in its ability to conduct similar testing and faces questions about what would be operationally permissible. One interpretation of this discussion about Chinese “dogfighting” in space is that the US Space Force is signaling that what the PLA is doing is legitimate. The implication is that if we want to deter and defeat these tactics — rather than merely exhausting them through proliferated assets that complicate disruption of American capabilities — we need to grant the Space Force more authority to operate in space, including activities that have been considered dangerous for decades.

Fiona Cunningham: If I’m not mistaken, the report on dogfighting described a scenario where several Chinese satellites — more than three — essentially surrounded another satellite. This demonstrates advanced capabilities in rendezvous and proximity operations — the ability to position several objects close to each other in space. This relates to Mike’s point about constraints regarding how close one might be permitted to approach in these operations.

Regarding the specific terminology of “dogfighting,” I’ll note that I don’t know how you would say “dogfighting” in Chinese. This probably indicates that the term doesn’t appear in their doctrine.

Jordan Schneider: Everyone wishes they were still flying fighter jets over Korea. That’s basically the lesson of all this.

Mike Horowitz: The question for Fiona is, how sophisticated do you believe China’s conventional missile arsenal has become? China has, by my count, at least six relatively modern anti-ship missiles, including two different anti-ship ballistic missiles, something like the DF-17 anti-ship hypersonic missile, along with the YJ-class missiles. How important is this development? Coming from the Pentagon in 2024, the importance of these missiles represents one of perhaps the top five questions facing the US military today, particularly the US Navy.

Fiona Cunningham: This is an Air Force issue as well, since these same missiles can target assets on regional airfields. Examining PLA conventional missile doctrine reveals something interesting: as their forces become more precise, the described target set actually narrows. Rather than infrastructure or economic targets, they now focus on missile defenses, electronic warfare capabilities, radars, airfields, and ships.

Some advances are obvious — China’s missiles have become more accurate. Finding reliable estimates of their accuracy is challenging — how many missiles will land within a short range of their target, and how frequently they’ll hit within a certain radius of that point.

The most important change for China comes in sensing capabilities. Precision missile capability requires not just accurate missiles but also sensing systems to locate targets and confirm successful strikes. This presents a particular challenge for anti-ship ballistic missiles. Extensive analysis has examined what space-based and ground-based capabilities China might employ to locate US carriers — the key question being whether they could find and track them sufficiently for missile targeting.

Recently, China deployed an optical satellite in geosynchronous orbit that provides persistent coverage of that region, significantly enhancing their target acquisition capabilities. The black box for me is what happens between the sensor and the shooter — this can be a very difficult process. I have less visibility on whether this presents as big an issue for anti-ship ballistic missiles targeting large vessels as it would for more precise targeting scenarios.

The other question involves US countermeasures and their effectiveness against China’s attempts to hit both moving targets and fixed installations. The US could potentially use its own counter-space capabilities to disrupt Chinese sensing systems important for locating and tracking vessels. However, fixed targets like Kadena Air Force Base remain at known locations. You may not know exactly what assets are present, but you can still strike these fixed targets even with degraded sensing capability.

This leads to considering what else the US can do to disrupt China’s precision strike capability, which ultimately points toward disrupting missiles before launch. This returns us to the challenging problems of tracking mobile missiles that we discussed earlier. You can see how these issues are interconnected.

I’m not directly answering your question, Mike, because uncertainty exists about the steps in the chain that China must complete to successfully hit the more challenging US targets. Substantial uncertainty also surrounds the effectiveness of US countermeasures, both in terms of disrupting PRC capabilities and in hiding from or diverting those missiles once launched and en route to their targets.

Mike Horowitz: That’s a great answer, super helpful. What about Chinese hypersonics in general? China has tested hypersonic systems that have excited missile enthusiasts and generated concern that the US could be falling behind in this technology. Setting aside the “falling behind” narrative — and I’ll save my personal rant about hypersonics for another day —

Fiona Cunningham: I’d love to hear it — come on, we’re here.

Jordan Schneider: We’re two hours in, Mike. The people are waiting for this.

Mike Horowitz: There’s no capability that disappointed me more when I left the Pentagon than hypersonics, particularly regarding the relative value for investment. They have their place but have been somewhat overhyped considering the overall architecture of missile systems. That’s not surprising coming from me, given my advocacy for precise mass and more attritable autonomous systems.

My question for you, Fiona: How fearsome are China’s hypersonics, and how have they managed to deploy so many hypersonic systems so quickly compared to the United States?

Fiona Cunningham: The short answer to why China has been quicker than the United States: I can’t point to a specific line in a PLA manual or teaching text, but if I were to hazard a guess, it’s because this is a priority for China. Conventional missiles, and missiles in general, represent an area where China has invested a lot of effort ahead of other aspects of its conventional military modernization and certainly its nuclear modernization. When you prioritize something, you naturally progress more quickly.

It also relates to China’s specific problem set. The United States has deployed missile defense systems around China’s periphery, most visibly with the THAAD system in South Korea in 2016, which caused considerable political disruption in Northeast Asian security dynamics. China faces the challenge of defeating missile defenses. If that’s your problem at theater range, hypersonics offer a potential solution because of both their speed and maneuverability in the terminal phase, making them more difficult for missile defenses to intercept.

Regarding their effectiveness — they’ve been tested in controlled environments. Their true capabilities won’t be known until they face actual US missile defense systems, and their success will depend on how US capabilities to track hypersonic missiles progress. This represents an ongoing development with the current space sensing architecture.

This creates a cat-and-mouse dynamic where China’s investment may provide a temporary advantage, but the US can potentially close the gap — either through the precise mass approach you advocate or through countermeasures specifically designed to address the problems that hypersonics solve for China.

Mike Horowitz: To me, that’s an “and” not an “or” for the US, but that’s very helpful.

Jordan Schneider: Thinking about that exchange, I was trying to imagine how to persuade Trump to defend spending on basic research and science. This led me to recall the “super duper missile” and “invisible aircraft” comments. We’re recording this on Friday, March 21, when the NGAD, the sixth-generation fighter, is about to be announced.

My mental model suggests Trump will approve spending on things that are big, fast, and shiny — not slow and attritable. There’s an interesting tension here. Many influential tech companies like Palantir and figures connected to Trump’s circle advocate for one theory of acquisitions and victory, whereas Trump himself seems drawn to technologies you can describe with dramatic action-figure adjectives, rather than, say, cute submersible drones.

Mike Horowitz: One notable aspect of the 17 priorities that the Pentagon announced for its review of the FY26 budget is how both sophisticated systems and “one-way attack” precise mass systems were explicitly identified as investment priorities.

My instinct is that you’re probably correct. Additionally, the now-confirmed Deputy Secretary of Defense Feinberg is a strong hypersonics advocate. What we might see, due to both Feinberg’s influence and presidential preferences, is a push toward a high-low capability mix — investing in both the biggest, shiniest assets like NGAD and hypersonics, as well as more distributed mass capabilities.

Since many US capability investments over the last couple of decades have focused on that mid-tier range, this shift raises questions about those programs. As an advocate for a high-low mix for the force, I don’t necessarily find this troubling, even if the pathway there differs from what I might prefer.

Jordan Schneider: Fiona, were there alternative paths the PLA could have taken? Could they have decided in the late 80s to focus on nuclear modernization, making that threat more credible, rather than building a massive conventional force with advanced capabilities in space, cyber, and missiles? Was that a viable option? And would that approach have made China’s rise less concerning to other nations? Or do you see concerns about China as primarily related to economic growth rather than specific military capabilities?

Fiona Cunningham: That’s an excellent question. It depends on which parameters you change. What if those big incidents with the United States in the mid-1990s hadn’t occurred? In the book, I also discuss the EP-3 crisis, when a US reconnaissance aircraft and a Chinese fighter collided over Hainan Island in 2001.

Mike Horowitz: Remember how big that incident was at the time?

Fiona Cunningham: It was enormous.

Mike Horowitz: We’ve somewhat memory-holed it, especially given the current global uncertainty, but when it happened, it was a major international crisis.

Fiona Cunningham: I was in what Americans would call middle school at the time. It was another event I followed closely, similar to the ASAT test, as it occurred very early in the Bush administration. Looking back, it’s clear why my childhood dream of becoming a human rights lawyer at the UN got diverted by media coverage of these events.

Mike Horowitz: By your one true love for nuclear weapons.

Fiona Cunningham: Well, if I truly loved nuclear weapons, I’d probably study a different country. They’re not China’s favorite.

An alternative narrative might have emerged when China encountered these crises with the US that revealed its leverage deficits. China could have decided, “We have nuclear weapons. We’ve observed NATO’s approach during the Cold War. We’re watching contemporary Russian strategies in the 1990s. We’ll simply adapt our nuclear posture and follow those models,” referencing Iain Johnston’s famous “limited nuclear deterrence” article. That’s one potential pathway.

Another scenario: without these crises, China might still have developed its conventional military, but its modernization and strategy would have responded to different variables — how other countries fight wars, party unity (to reference Taylor Fravel, my advisor’s work). You might have seen conventional modernization proceed without these investments in non-nuclear strategic deterrence.

Economic factors could have reshaped China’s conventional modernization trajectory. Jiang Zemin explicitly stated in the late 1990s that China’s military modernization progress toward joining the ranks of advanced military powers — “world-class militaries,” though he didn’t use that specific term — was contingent on the country’s economic circumstances.

A third consideration: how much of the current situation stems from Xi Jinping being a different type of leader with distinct visions for China’s foreign policy and defense strategy compared to his predecessors? I tend to see more continuity than change. However, with a different leader pursuing objectives unlike those of Hu Jintao or Jiang Zemin — perhaps emphasizing greater international engagement or taking a different approach to domestic politics — China might have followed a different path.

Mike Horowitz: It’s tempting to view everything as inevitable in retrospect. My instinct is that if it hadn’t been the Taiwan Strait Crisis or the EP-3 incident, something else would have triggered similar developments. From a structural perspective, China’s rise combined with frequent US demonstrations of conventional military superiority during the Iraq invasion and subsequent conflicts created conditions where accelerated PLA conventional modernization became highly probable. The question is: what additional steps would they take beyond that?

Fiona Cunningham: That’s where politics enters the equation. Stepping back, one of the major conclusions of my book is that China’s decisions about strategic deterrence and capabilities intended to pressure the United States are all connected to political dynamics. This isn’t primarily about US capability dynamics but rather those political crises that create urgent demand for enhanced capabilities to deter the United States from engaging in future confrontations. The US doesn’t always pay sufficient attention to these political dynamics, while from China’s perspective, they’re paramount.

Jordan Schneider: Let’s close with three things, Fiona. First, recommend one PLA book for listeners who’ve made it to the end of this episode. Second, share your favorite Chinese phrase, perhaps something PLA-adjacent. Finally, which meeting in recent or older PLA history would you have liked to witness, and perhaps what organization you might have worked for during which time period.

Fiona Cunningham: I would decline working for any PLA organizations — I don’t think that uniform would have suited me well.

Regarding a meeting I would have liked to witness recently, I’m curious about why China’s military leaders decided to disband the Strategic Support Force around March or April last year. It previously housed space capabilities, cyber capabilities, and the network maintenance organization for all PLA Defense Information Networks. They announced a separate force for information networks, but almost nothing has emerged about what happened to the cyber and space components. I really want to understand why this occurred and why there’s been silence about the other elements.

For historical meetings, in approximately 1978, there was a meeting where Deng Xiaoping commented that China should continue considering tactical nuclear weapons while making decisions about the future of Chinese nuclear forces. Information about these decisions is scarce, but I’d love to know what choices led China’s leaders to still contemplate whether shorter-range, lower-yield nuclear weapons would help defend their borders against the Soviets, and why that idea eventually faded as China maintained its restrained nuclear strategy.

For a PLA book recommendation, I frequently return to an edited volume from the National Defense University titled Crossing the Strait: China's Military Prepares for War with Taiwan with editors Joel Wuthnow, Phillip Saunders, and others. It contains many valuable chapters summarizing China’s doctrine for a Taiwan conflict. Joel Wuthnow’s chapter on PLA command and control systems is particularly helpful for anyone wanting to understand China’s campaign capabilities as of a year or two ago. It exemplifies the best that PLA studies and analysis can offer regarding the dilemma we discussed at the beginning of this podcast — this remains the primary scenario and most difficult problem facing military leaders in the United States and allied countries.

My favorite Chinese phrase would be “惩戒行动” (chéngjiè xíngdòng), which appears in descriptions of China’s counter-space capabilities and roughly translates to “punitive strikes and disciplinary action” — essentially combining “punish” and “warn.” This phrase resonated with me early in developing my dissertation topic because it emphasized how differently China approaches the space domain and space deterrence compared to nuclear weapons. In many ways, it represented the variation in the dependent variable that made this not just an interesting and policy-relevant topic but one with academic merit as a political science dissertation.

Jordan Schneider: Can you use those words in a teaching or parenting context?

Fiona Cunningham: During my interviews in China, I asked what this phrase meant. People explained that “惩戒行动” referred to conflicts with punitive and disciplinary elements, like the Sino-Vietnamese War. When I asked if nuclear retaliation qualified, they said no. There’s a footnote about this in the book. No one suggested it was terminology they would use when teaching students or parenting children, so I would recommend keeping it within a military strategy context.

Jordan Schneider: Tell us more about the general atmosphere of the Chinese defense analyst community. How would you characterize them as people, using broad generalizations? And what do they gain from speaking with you?

Fiona Cunningham: When I conducted interviews for the book, mostly as a graduate student, I believe they were partly motivated by helping a student learn. The desire to assist students transcends cultural contexts. There’s also genuine interest among many Chinese experts to engage with researchers who make the effort to visit China, learn the language, and understand the strategic studies lexicon. They want to clarify misconceptions and ensure that American discourse about these topics shows a more nuanced understanding.

I was in China conducting this research during the 2015-16 PLA reforms. At times, when I asked questions about certain issues, people would simply respond that they didn’t know — until the reform package clarified, there was a lot of uncertainty.

As a group, the Chinese defense and strategic studies community isn’t fundamentally different from what you’d expect in the United States. There’s diversity in terms of gender, age, and ideological perspective — some more conservative than others. However, it’s a relatively small community, similar to the United States, where many experts know each other well and are familiar with each other’s views.

Jordan Schneider: One more question about the defense community. In a recent paper, you noted that the civilian defense consensus argued against extensive nuclear modernization, but Xi disregarded this view, and modernization proceeded anyway. However, before 2019, expert discussions often aligned with eventual PLA actions across various dimensions. What do you make of this shift?

Fiona Cunningham: One conclusion might be that the expert community has less influence or interaction with decision-makers than previously. Some have drawn this interpretation regarding nuclear policy. I have work in progress — referenced when we discussed China’s nuclear modernization drivers — showing that perspectives within China vary on why nuclear capabilities should be modernized and what specific actions should be taken.

In that paper, I focused on one segment of the PRC community involved in arms control that expressed concerns about threats to China’s retaliatory capability but didn’t advocate for substantial arsenal expansion. However, other voices within China’s strategic and defense community do support a larger arsenal — not based on intricate calculations about warhead targeting or force posturing to make credible threats, but simply because more nuclear weapons provide political leverage. This isn’t connected to military campaigns or outcomes but shows the psychological impact of greater capability. It could even be viewed as status-related, with nuclear prominence offering instrumental advantages.

This mentality helps explain the gap between what the arms control community might recommend and China’s actual behavior. More research on this is forthcoming.

Jordan Schneider: Standing invitation — we could host shows in Chinese if experts want to join us, with Fiona and me co-hosting discussions about specialized aspects of PLA doctrine.

Mike Horowitz: I look forward to the translation of that episode.

Mary Clare McMahon is an incoming Schwarzman Scholar (‘26) and former Winter Fellow at the Centre for the Governance of AI, where she researched compute governance and U.S.-China AI competition. Previously, she worked in the National Security and Cybercrime Section of the U.S. Attorney’s Office for the Eastern District of New York.

Last month, reports emerged that DeepSeek was running a distilled version of its R1 reasoning model on Huawei’s Ascend chips. While DeepSeek trained its model on Nvidia Hopper series chips, Huawei’s deployment of Deepseek R1 underscores a broader strategic question: to what extent can Huawei erode Nvidia’s dominance in the global AI chip market?

Nvidia’s position rests on what has called a “three-headed hydra” of leading hardware, networking capabilities, and, most importantly for this piece, a deeply entrenched software ecosystem. At the center of that ecosystem is CUDA, a proprietary programming framework that allows developers to efficiently map computations onto Nvidia’s GPUs. CUDA’s value lies not only in its performance but in its reach: an expansive set of libraries, optimized workflows, and tight integration with widely-used machine learning frameworks make it the industry standard. And, crucially, CUDA can only be used with Nvidia GPUs. That makes CUDA a core component of Nvidia’s competitive advantage, otherwise known as Nvidia’s moat.

This article explains Huawei’s attempt to replicate and bypass that moat. For now, Huawei appears to be advancing the following three-pronged strategy:

1. Building out its own software stack, including a proprietary parallel programming model and surrounding tools that developers rely on to write, optimize, and deploy code efficiently.

2. Deepening integration with PyTorch, the most widely adopted open-source machine learning framework for model training.

3. Investing engineering resources in developing the Open Neural Network Exchange (ONNX), an open standard for machine learning models that enables portability across hardware platforms, to support the deployment of non-Ascend-trained models on Ascend chips.

Huawei is not the only actor seeking to erode Nvidia’s software lock-in — AMD has made similar efforts with ROCm, and Google has a software stack fitted to run Google TPUs. However, Huawei remains the most significant challenger in the Chinese market. The core question is not whether Nvidia’s dominance is being contested, but whether Huawei’s software strategy can mature enough for a full-stack transition away from U.S. hardware. This article proceeds in two parts: part one provides background on Nvidia’s software moat and how it was constructed; part two analyzes Huawei’s evolving response.

Nvidia’s Software Moat

The roots of Nvidia’s software moat can be traced back to the late 2000s, when CEO Jensen Huang made a long-term bet on CUDA, Nvidia’s proprietary parallel computing platform. In 2007, Nvidia released CUDA as a programming model for scientific computing. At the time, the dominant paradigm for scientific research (and most other computing applications) was CPU-based computation; GPUs were considered niche accelerators, primarily designed for graphics rendering in video games. CUDA’s launch was an explicit attempt to invert that paradigm by positioning the GPU as a general-purpose compute platform.

CUDA allowed developers to write code in familiar C/C++ syntax that executed directly on Nvidia GPUs, thereby accessing the highly optimized functionality of these GPUs. But creating a new computing model meant overcoming a classic chicken-and-egg problem: developers needed hardware to test their software on, and customers needed software to run on their hardware — neither would commit without the other. Nvidia addressed this by seeding the market for CUDA with its consumer gaming cards, which already had a broad base of installation. It made CUDA freely available (without open sourcing the code), created a global developer conference, and worked directly with scientists and researchers to port algorithms to the GPU. As Huang later recalled in a speech at National Taiwan University, “We worked with each developer to write their algorithms and achieved incredible speedups.” This engagement strategy eventually paid off; in 2012, AlexNet was trained on CUDA and Nvidia GPUs.

Subscribe now

As Nvidia’s software and hardware stacks became popular with deep learning researchers, Nvidia continued to invest in — and improve — CUDA. Nvidia created an extensive suite of libraries, such as cuDNN for deep learning, which dramatically lowered the time and expertise required to deploy high-performance models. In short, CUDA became more than just a programming model — it became the foundation of a full-stack software ecosystem.

For the next decade, CUDA continued to improve and attract more developers. And it is still improving to this day — though CUDA is closed source, Nvidia welcomes and often incorporates developers’ feedback. Nvidia also maintains online forums for developers to answer and ask questions about CUDA.

Thus, the CUDA ecosystem embeds substantial switching costs. Developers who migrate away from CUDA usually must rewrite large portions of code — by forgoing access to Nvidia’s finely tuned libraries, developers are forced to substitute with less mature equivalents, if any replacements exist at all. Further, developers also lose support from the large troubleshooting community that has grown up around CUDA.

Today, many machine learning developers do not code directly in CUDA. Instead, they write code in Python, a higher-level and more user-friendly language, using frameworks such as PyTorch and JAX. But even here, CUDA remains central: it acts as the backend bridge between PyTorch and Nvidia’s GPU architecture.

We will discuss PyTorch in greater detail in a later section. For now, it is enough to note that CUDA’s value lies not only in its impressive performance (which has improved continuously for nearly two decades), but also in the ecosystem that has formed around it. That is the essence of Nvidia’s moat — challengers with competitive hardware must also replicate an entire software environment if they want to compete.

Huawei’s Software Strategy

Undermining Nvidia’s software moat requires more than performance parity with Nvidia GPUs — it demands a credible alternative to the tightly integrated CUDA ecosystem. Huawei appears to be pursuing such an alternative. Its strategy consists of three interrelated prongs, each aimed at reducing the friction of switching away from Nvidia.

First, it is expanding its native software stack alongside a growing suite of tools designed to mirror the utility of CUDA’s broader ecosystem. Second, Huawei is deepening integration with PyTorch, the most widely adopted machine learning framework and one that, by default, pairs seamlessly with CUDA. By building backend support through adapters like torch_npu, Huawei is attempting to position Ascend as a drop-in hardware alternative. Third, Huawei is investing in ONNX (Open Neural Network Exchange), an open standard for cross-platform model representation, to allow models trained on non-Huawei hardware to run inference efficiently on Huawei chips. Together, these efforts seek to replicate the full-stack developer experience that has made CUDA so difficult to displace.

1. Huawei’s Software Alternatives

Huawei’s most direct challenge to CUDA comes in the form of CANN (Compute Architecture for Neural Networks), its proprietary programming environment for Ascend NPUs. CANN sits at the same level of the software stack as CUDA, providing the tools needed to execute high-performance machine learning models on Huawei hardware. Paired with CANN is MindSpore, Huawei’s high-level deep learning framework, conceptually analogous to PyTorch. Together, these tools form Huawei’s native alternative to the Nvidia-centric PyTorch + CUDA stack.

​​CANN has been in development since at least 2019, the year Huawei was added to the US entity list. Huawei’s 2024 Annual Report highlighted (on four occasions) the release of CANN 8.0 in September of 2024, promoting this development as a significant step in advancing AI computing capabilities.

However, developers cite serious usability issues with CANN. According to the Financial Times, one Huawei researcher complained that CANN made the Ascend chips “difficult and unstable to use.” One developer described the process of using the Ascend 910B as “a road full of pitfalls” (踩坑之路), sharing the following reflections on Zhihu, a Quora-like Chinese website for academic discussion, in February 2025:

“I have been interning in the company for the past six months. Due to the shortage of computing resources, interns can only use Ascend 910B for training and development… Looking back, every time I encountered various problems and bugs, it was difficult to find the corresponding solutions on the Internet. Some problems were finally solved with the help of Huawei's operation and maintenance engineers. Therefore, I hope that this article, in addition to summarizing my own staged engineering experience, can help more Ascend NPU developers and help the development and progress of the domestic computing ecosystem.”

426 other users upvoted the post. One commenter responded, “It seems that it will take until 2027 for CANN to be truly mature, stable, and easy to use.”

The absence of a robust developer community for CANN further increases the onboarding burden for new developers. Unlike Nvidia’s developer forums, which benefit from community-maintained documentation and rapid peer troubleshooting, Huawei’s Ascend developer portals — both in English and Chinese — exhibit low engagement, with sporadic posts and limited public debugging activity. According to another Zhihu article posted in June of 2024, “When I first started exploring Ascend, I felt quite overwhelmed. Although there is a lot of documentation available, it feels quite disorganized. When encountering problems, the limited user community means you probably won’t find a corresponding solution, which leads to frequent frustration.”

While the Nvidia CUDA Programming and Performance Developer page had multiple live threads posted just days before the screen capture above, the most recent posts on the Huawei CANN developer pages were from January 2025.

Adapting models to run on Huawei’s platform is also onerous. According to that same Zhihu article from June 2024, “Any public model must undergo deep optimization by Huawei before it can run on Huawei's platform. This optimization process is heavily dependent on Huawei and progresses slowly.” By contrast, after testing the Nvidia H100 and H200 for model training applications, Semianalysis reported, “Nvidia’s Out of the Box Performance & Experience is amazing, and we did not run into any Nvidia specific bugs during our benchmarks. Nvidia tasked a single engineer to us for technical support, but we didn’t run into any Nvidia software bugs as such we didn’t need much support.”

To try to increase adoption, Huawei has adopted a strategy reminiscent of Nvidia’s own CUDA rollout in the 2000s: embedding engineers directly into customer sites to assist with code migration. According to reporting from the Financial Times, Huawei has deployed engineering teams to Baidu, iFlytek, and Tencent to help reimplement and optimize existing CUDA-based training code within the CANN environment​. This mirrors the anecdote recounted above, where Jensen Huang described how Nvidia “worked with each developer to write their algorithms and achieved incredible speedups” during CUDA’s early years. Huawei is now attempting to replicate that strategy, pairing onboarding with high-touch technical support in the hope of accelerating ecosystem uptake.

In parallel, Huawei is also trying to improve its native software stack. DeepSeek engineers have reportedly said that the Ascend 910C can achieve up to 60% of the inference performance of the H100, and potentially more with CANN optimizations. As Kevin Xu noted on a prior episode of ChinaTalk, DeepSeek engineers have proven adept at “work[ing] below CUDA to maximize their Nvidia GPU.” If similar techniques were applied within the Huawei ecosystem, they could help close the performance gap between Ascend and NVIDIA hardware.

One particularly intriguing way to close that gap involves using AI to accelerate software optimization. If AI systems themselves can be leveraged to improve kernel optimization, develop the CANN and MindSpore stack, and reduce performance inefficiencies, it could meaningfully shift the competitive landscape. Sakana AI has already demonstrated a version of this approach with its “AI CUDA Engineer,” an agentic framework that translates standard PyTorch code into highly optimized CUDA kernels. According to Sakana, the system achieves 10—100x speedups for AI model training. If comparable AI-driven optimization techniques could be adapted for Huawei software, it would represent a significant step toward enhancing performance within the CANN ecosystem. Developer loyalty might follow.

Despite its investment in a native software stack, though, Huawei appears to recognize that displacing CUDA with CANN is not feasible in the near term. As a result, it has shifted part of its strategy toward interoperability rather than replacement. Nowhere is this more evident than in Huawei’s growing involvement with the PyTorch ecosystem.

2. Huawei and PyTorch

As part of its strategy to reduce friction in migrating away from Nvidia, Huawei has prioritized compatibility with PyTorch, the dominant open-source machine learning framework used across academia and industry. Originally developed by Meta’s AI research lab in 2016, PyTorch was released publicly in 2017, then transitioned to being governed by a wider network of companies under the Linux Foundation in 2022. The resulting PyTorch Foundation is governed by a consortium of premier members, including Meta, Microsoft, Google, Amazon, AMD, Intel, Nvidia, and, as of October 2023, Huawei.

PyTorch enables developers to define, train, and deploy machine learning models using concise and intuitive Python code. The framework's popularity stems from its "eager execution" model, which allows each operation to run immediately, making it easier to debug, prototype, and iterate than other alternative frameworks (like Google’s TensorFlow).

From the outset, PyTorch was optimized for Nvidia GPUs. New operators and features are still tested and tuned against CUDA first, and performance benchmarks are routinely conducted on Nvidia’s hardware. Installing PyTorch via Python’s package manager automatically sets it up to run on Nvidia GPUs. This makes the framework effectively Nvidia-native, and any effort to use it on non-Nvidia hardware requires not just backend substitution, but complete ecosystem engineering.

The challenge for Huawei, then, is not only to make PyTorch run on Ascend hardware, but also to make it run well enough that developers don’t notice they’ve switched ecosystems.

Huawei’s primary technical achievement has been enabling the execution of PyTorch models on its Ascend NPUs through an adapter called torch_npu. Torch_npu bridges PyTorch with Huawei’s low-level NPU drivers and CANN backend. Huawei developers publicized this development at the 2024 PyTorch Shanghai Meetup, pictured below.

Huawei’s torch_npu adapter allows Huawei's AI accelerators to interface with PyTorch, though it exists separately from PyTorch’s main codebase. (The torch_npu adapter uses PyTorch’s PrivateUse1 mechanism, an interface that lets hardware makers test new accelerators without immediately merging their code into PyTorch.) At the 2024 PyTorch meetup in Shanghai, a Huawei engineer noted that devices maintained outside PyTorch’s core, like Huawei’s, often face stability issues because changes in PyTorch's main code aren't automatically tested for compatibility. This challenge is widely recognized by the community.

For this reason, Huawei’s forked version of PyTorch is still less effective than Nvidia’s CUDA-native implementation, and developer feedback points to persistent challenges in runtime reliability and documentation. In a Zhihu thread with more than 700,000 views, senior software engineer “Mingfei” wrote that, “It’s worth emphasizing that plugins [referring to the forked version of PyTorch] are not native” and “several unavoidable issues arise,” including version compatibility; third-party extension support; and test coverage challenges. Another Zhihu contributor noted, “Ascend chips provide poor support for third-party frameworks like PyTorch and TensorFlow, making it extremely challenging to adapt to the latest large-scale models and use them effectively.” Note that the developer seems to be referring to the challenges of deploying models on Ascend chips, not training new models.

While Huawei’s patches have not yet been fully integrated upstream, there are reasons to believe that Huawei might be able to garner political support within the PyTorch Foundation to formalize its contributions. The PyTorch Foundation’s official announcement of Huawei’s status as a premier member noted that Huawei “provides easier access to the PyTorch ecosystem for more hardware vendors… [which] aligns with the PyTorch Foundation’s mission to develop AI as part of a sustainable open source ecosystem and produce inclusive technological feats.” This quote seems to suggest that PyTorch wants to support other hardware options besides Nvidia’s. Further, Huawei’s status as a premier member of the PyTorch Foundation grants it a seat on the Governing Board, as well as a formal role in setting foundation-wide policies and technical priorities. This membership was unanimously approved by existing premier members, signaling at least tacit acceptance of Huawei’s contributions by Meta, Nvidia, AMD, and Google. Finally, Huawei appears to be strongly committed to contributing to open source projects. The company’s 2024 Annual Report highlighted that Huawei is “a firm supporter and major contributor to open source communities” and explicitly mentioned its membership in the PyTorch Foundation.

In sum, Huawei is executing a long-term strategy to allow developers to use PyTorch with its Ascend series of chips. Its success will depend on the company’s continued technical contributions, the size of its developer community, and whether the PyTorch Foundation will incorporate the torch_npu and other Huawei contributions into its main code base.

3. Huawei and OXXN

While Huawei’s PyTorch integration aims to reduce friction in model development, it does little to solve the harder problem of model portability — that is, how to take a model trained on Nvidia hardware and deploy it on Huawei’s Ascend chips. To address this, Huawei has turned to a complementary approach, optimizing the Open Neural Network Exchange (ONNX) format to serve as a bridge between software ecosystems.

ONNX (Open Neural Network Exchange) is an open-source format originally developed by Meta and Microsoft in 2017 to enable model interoperability across deep learning frameworks. It allows developers to export a model trained in one framework, such as PyTorch with CUDA, and run inference in another runtime environment — or on different hardware entirely. It also helps optimize models, allowing them to run faster than they would if they were directly deployed from PyTorch. ONNX operates under the umbrella of the Linux Foundation AI & Data, of which Huawei is a premier member.

Put simply, ONNX is like the PDF of AI models. Just as documents created in Microsoft Word or Google Docs to be exported into a portable, fixed-format PDF file that can be opened and viewed across operating systems, ONNX allows models trained in PyTorch or other machine learning libraries to be exported into a standardized format that can then be run on different hardware platforms.

Huawei has embraced ONNX Runtime, the engine that executes ONNX models. The company maintains a public Ascend ONNX Runtime, available on GitHub, which includes optimized kernels and execution instructions tailored to CANN and Ascend chips. According to the ONNX Runtime documentation, Huawei’s ONNX Runtime page is “community-maintained,” meaning that it is maintained by Huawei rather than by the core ONNX Runtime team, and that it is Huawei’s responsibility to ensure ongoing support for the library.

Subscribe now

Huawei’s goal here is straightforward: to enable developers to train models on non-Huawei hardware, export the files to ONNX, and deploy the models on Ascend chips, all without rewriting core logic. This workflow has clear appeal in the Chinese market. Model developers could still train on Nvidia Hopper chips or train models through the cloud, then shift deployment or inference workloads to Huawei hardware.

It’s important to note that running a model on hardware, even if using an ONNX file, can introduce bugs or compatibility issues. Some PyTorch operations don’t export cleanly to ONNX, while others need rewriting. ONNX models may also need custom operations that the hardware backend has to support. That said, Huawei’s investment in ONNX offers a practical path to inference decoupling. In contrast to the CUDA-first development loop, which binds training and deployment to Nvidia hardware, ONNX gives Huawei a way to insert itself at the deployment stage, even if training remains CUDA-bound.

Huawei’s Future

Nvidia’s enduring dominance in the AI chip market is not due to superior hardware or networking architecture alone — it’s also a function of Nvidia’s deeply integrated software ecosystem. This ecosystem — anchored by CUDA, high-performance libraries, and seamless compatibility with PyTorch — offers a robust developer experience and an active community that reinforce Nvidia’s lead. Huawei’s strategy is to build a competitive stack of its own.

Model deployment may be Huawei’s most immediate opening. Already, it has demonstrated that models trained on Nvidia hardware, like DeepSeek’s R1, can be run in distilled form on Ascend chips. If the US were to ban the export of Nvidia H20s to China, this workaround could become standard. In that scenario, indicators of improvement in the Huawei software stack would manifest not as headlines, but as reduced developer complaints, more seamless deployments, and fewer distinctions between fallback option and first choice.

Huawei isn’t there yet, though. As noted by the exasperated programmers quoted above, working with Ascend 910B chips still requires debugging without community support. But Zhihu threads where developers vent frustrations can eventually become a troubleshooting resource that contributes back to the Huawei ecosystem. With enough developers dedicated to advancing that new ecosystem, the result could be a slow, durable shift away from CUDA. That shift won’t happen overnight — remember, it took Nvidia 18 years to build the CUDA ecosystem of today; building a competitive software ecosystem is a multi-year effort even under pressure. But what started as necessity may, over time, harden into habit — and eventually, into infrastructure that can compete with Nvidia’s software stack.

Special thanks to Jeff Ding and Kevin Xu for thoughtful feedback on prior drafts.

Read more

What’s happening to America’s science and technology ecosystem? How is China interpreting the current state of US research, and how is it working to build its own science and technology base in response? And what can we learn from China's war mobilization exercises?

To explore these questions, we're joined by Divyansh Kaushik and Alex Rubin, who both work at Beacon Global Strategies. Divyansh holds an AI PhD from Carnegie Mellon, and Alex spent the past decade at CIA focusing on China and emerging technologies.

We discuss…

• The historical origins of the US R&D model, and the division of labor between universities, government, and industry,

• How budget cuts will impact the NSF, NIH, NIST, and DoD basic research,

• Why and how China attempts to emulate US research institutions,

• What a leaked wargame exercise from Guangdong province can tell us about China’s grand strategy,

• How institutions like ChinaTalk complement the IC with fresh, independent research.

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

How America Has Won With R&D

Jordan Schneider: Let’s do a 101 on the broader American research ecosystem. What does the interaction between universities, government funding, and corporations look like in the 2000s?

Divyansh Kaushik: To better understand today’s landscape, we need to trace our steps back about 70 years and examine how the American research ecosystem was conceptualized. The original model positioned universities to conduct curiosity-driven research funded by the federal government, while American industry focused on transforming that research into applications.

There were certain industrial monopolies created by the government that also conducted basic research, which Alex can address more comprehensively. However, the overwhelming majority of basic research happened in academia — universities created as land-grant institutions or those existing before the war. This system served us remarkably well, as basic research developments from the 1950s, 60s, and 70s bore fruit 10, 20, 30, or 40 years later. The nature of basic research doesn’t necessarily have an immediate application, but applications may emerge years down the line.

Consider this example: During the 1970s and 80s AI winter, when nobody was funding neural networks research because it was viewed as a dead end without viable applications, the National Science Foundation — created to fund basic research through the federal government — was funding Geoffrey Hinton’s work on neural networks, being the only entity supporting this research at the time. Fast forward 40-50 years, that work has fundamentally shaped how we view AI today. It’s the foundational technology behind all the large language models currently in use.

NSF also funded Andrew Barto’s entire PhD. Barto, together with Richard Sutton, established the field of reinforcement learning at a time when there were few practical applications. Today, reinforcement learning is a critical component behind LLMs, AlphaFold, and similar technologies.

This exemplifies how America has pursued basic research. Currently, there’s considerable criticism about research projects like "shrimp on a treadmill" or "fish on cocaine," questioning why such studies receive funding. While these projects have legitimate scientific purposes, to the general public they appear to be wasteful uses of federal research dollars.

Agencies like the National Institutes of Health fund more applied research on medicines and can point to tangible outcomes — specific drugs developed with NIH funding. The NSF, conversely, funds basic research that may not demonstrate tangible benefits for decades, as happened with neural networks.

We’re now engaged in a deeper conversation about what constitutes waste or abuse of federal research dollars and how to allocate those funds more effectively. Is industry-funded research the optimal approach? Does the federal government have — or should it have — a role in the R&D ecosystem? What about public-private partnerships, which were a cornerstone of the CHIPS and Science Act in creating the Technology Innovations and Partnerships Directorate at the National Science Foundation?

This significant conversation emerged particularly this year, as the National Science Foundation, Department of Energy, Department of Defense, and National Institutes of Health find themselves at the center of questions regarding the appropriate allocation of federal dollars toward research.

Jordan Schneider: Let’s explore more of this history, because I think we can’t ignore the broader defense community’s role in funding R&D over the past 75 years. Alex, would you like to address that?

Alex Rubin: This is best illustrated through an interesting case study. We’re currently focused on the semiconductor industry due to its substantial economic and strategic implications. The foundation of the semiconductor industry can be traced directly back to funding from the Air Force and NASA for both the Apollo program and ICBMs.

One interesting example demonstrating the federal government’s role involves what we call the "Valley of Death" — the challenge of bringing novel research from the laboratory to market. The federal government, particularly the defense procurement establishment, has excelled at intervening at this crucial stage in the R&D cycle by providing customers and markets for these technologies. This allows companies to expand production, build scale, and reduce costs, making it feasible for them to enter commercial markets.

The modern semiconductor industry wouldn’t exist as it does today without those initial purchases from the Air Force, NASA, and other government entities. It’s extraordinarily difficult to transition from the lab to the market, especially when costs are high. Finding consumers willing to purchase these products when they’re expensive is challenging, which is where government plays a particularly important role in advancing that cycle.

Subscribe now

When discussing the federal government’s role in early-stage basic research, there are instances where industry has undertaken this responsibility. The AT&T system — the Bell system — during the 20th century provides a classic example of industry conducting exactly the type of research we’re discussing. However, certain unique characteristics made this possible for the Bell system.

Throughout the 20th century, the Bell system comprised several key units: corporate headquarters in New York, its manufacturing arm (Western Electric) producing equipment for telecom networks, the regional Bell operating companies that eventually became companies like T-Mobile and Verizon providing local phone service, long-distance service, and crucially, Bell Labs.

Extensive literature documents Bell Labs’ history and impact. Innovations including the transistor, discovery of cosmic microwave background radiation, and cell networks all emerged from Bell Labs. The list of remarkable technologies and innovations originating there is extensive.

Bell Labs could invest in both basic and applied research because of their consistent funding stream from what was essentially a government-regulated monopoly. In the 1910s, the government granted AT&T monopoly status over the telecom industry with certain conditions, including a requirement to license any inventions outside the telecom industry under generous terms or without fees.

Bell Labs operated with a consistent revenue stream from their telecom service monopoly, maintained a direct relationship with manufacturers allowing them to troubleshoot as products moved to manufacturing, and had direct connections to customers through their operating companies, enabling them to identify market demands.

Most importantly, they had long-term consistent funding — precisely what makes federal research dollars so crucial for basic research. This consistent funding allows investment in projects that might not yield deliverables for 10, 15, or 20 years. This differs dramatically from typical corporate investments seeking returns within a couple of years.

Companies justifiably need to demonstrate return on investment, which becomes incredibly difficult without consistent market support. Industry can indeed support basic research, but it requires specific enabling characteristics similar to those that enable basic research funding from the public sector.

Jordan Schneider: Divyansh, could you address the university’s role in this ecosystem?

Divyansh Kaushik: Universities today aren’t limited to basic research — they conduct significant applied and industry-funded research as well. However, an often overlooked aspect of universities’ contribution is their role in creating talent pipelines. The researchers going into industry are those who received federal funding at universities.

These individuals enter graduate programs where they develop intellectual curiosity through curiosity-driven research and by solving interesting problems without immediate pressure to generate revenue. They cultivate this intellectual curiosity and bring it with them when they join industry, ultimately driving the industrial innovation we witness.

Universities play a crucial role in regional innovation and economic growth. The spillover effects include startups and jobs created as a result of research funding. Numerous economic studies demonstrate multiple dollars returned for every dollar of federal R&D spending at universities.

Universities also advance national security objectives. Carnegie Mellon University, my alma mater, works with the Department of Defense on several projects directly impacting warfighters. Universities host Federally Funded Research and Development Centers (FFRDCs) and integrate DoD personnel into their research programs.

Universities therefore have a broader ecosystem-driving role, not just a narrow focus on quarterly profits. Both approaches are valid — they’re complementary rather than substitutive. While most companies don’t conduct basic research, some do. Microsoft, for instance, spent nearly ten years developing the Majorana chip for quantum computing, made possible by consistent funding.

If we reduce consistent funding for universities, we’ll see fewer PhD students enrolling, fewer PhDs granted, fewer qualified individuals joining companies like Microsoft, and ultimately fewer innovations like the Majorana chip. This affects the entire ecosystem.

Jordan Schneider: Alex, can you discuss how envious the rest of the world is of the ecosystem America has built?

Alex Rubin: A couple of decades ago, China looked at the U.S. R&D ecosystem and essentially said, “We want that,” and began working to replicate it. They’ve invested considerable resources — money, time, and high-level attention. President Xi Jinping regularly emphasizes the importance of basic research, improving China’s STEM education system, and developing talent as key enablers of China’s technological development and growth as an economic and global power.

China recognizes that the U.S. model is incredibly effective at generating innovations, bringing them to market, and establishing dominance and first-mover advantages in critical new technologies. China’s approach to its R&D ecosystem and education system focuses on three main categories, emphasizing generational investment.

Jordan Schneider: Let me provide some context with numbers. The U.S. spends approximately $50 billion annually on basic R&D, with another $50 billion coming from universities and businesses — totaling around $100 billion yearly.

American firms represent 80% of the world’s global technology market capitalization. Additionally, 80% of science Nobel Prizes over the past 50 years have included winners with U.S. affiliations. These three factors are interconnected.

This ecosystem produces the most advanced companies, which then provide cutting-edge technologies to the national security establishment. It’s a beautiful, self-reinforcing system. The best scientists work at American universities, attracting the best students worldwide. Despite providing only 25% of global basic funding, the U.S. spends it so effectively that the greatest minds globally want to come here and work on these topics.

Alex Rubin: This is indeed a generational investment. It’s no coincidence that U.S. companies initially led in the semiconductor industry, then in personal computers and other computing applications, and now in AI. These advantages build upon each other.

If you establish early leadership in one industry and continue making long-term investments, it naturally positions you advantageously for the next generation of technology. Conversely, if we make decisions now that underinvest in research or otherwise hinder the development of these talent ecosystems, the real impact might not become apparent for 10-15 years.

Unfortunately, once those impacts become visible, it’s often too late. Recovering lost ground requires significant time. The research investments and decisions we make now — whether regarding grants or graduate programs — will show their consequences a decade or more from now.

Jordan Schneider: That concludes our cheerleading session for the American R&D ecosystem. Divyansh, what has been happening over the past 100 days that concerns all three of us — developments that may risk the world-historical R&D golden goose America has built since Vannevar Bush wrote to FDR, envisioning a glorious future made possible by the collaborative efforts of corporations, research universities, and the U.S. government?

Divyansh Kaushik: I would actually broaden the timeframe to consider the last year or so. During this period, we saw the NSF budget cut by approximately 8% from the previous year, the NIST budget cut by about 13%, and the DoD basic research budget reduced by roughly 4%.

Early in the administration, NIH changed its policy on Facilities and Administrative (F&A) benefits, unilaterally reducing them to 15% — a decision that faced legal challenges. This was followed by layoffs at several federal funding agencies.

More recently, DoD and DOE followed NIH’s approach by capping F&A at 15%, which merits separate discussion. Additionally, NSF terminated approximately 400 previously awarded grants. We also witnessed the resignation of the NSF director amid rumors of potential additional layoffs at the agency.

Jordan Schneider: A former Trump appointee, mind you.

Divyansh Kaushik: Correct, and unanimously confirmed by the Senate. Further concerning developments included leaked information about the President’s budget request, suggesting NSF could face approximately 55% budget cuts, with NIH potentially facing similar reductions.

These issues are foremost in the minds of both academic and industry researchers. Brad Smith recently wrote a blog post about quantum computing where he emphasized the importance of basic science funding for workforce development. This concern is widespread.

Several former national security leaders, including former Trump appointees such as his former Homeland Security advisor and others, signed a letter to Congress highlighting the importance of funding basic science research at this critical juncture. As Alex mentioned, China has increased its basic R&D spending by 10% year-over-year for the past seven years.

The CHIPS and Science Act established a vision to double our federal basic R&D spending over the next decade. Instead, we’ve failed to meet this moment. Approximately $50 billion of authorized funding from the CHIPS and Science Act remains unappropriated for the science component.

Regarding talent, certain universities received letters terminating visas for some PhD students — a decision the administration later reversed. This situation weighs on the minds of universities and researchers, raising questions about broader implications on the global stage.

France, Australia, and China have attempted to capitalize on this uncertainty by establishing specific programs to attract U.S. researchers, offering long-term stability, funding, and residency benefits.

Jordan Schneider: You missed one aspect, Divyansh. We’re also seeing targeted actions against specific universities, with significant conflicts involving Columbia and Harvard. Beyond the 400 NSF grants canceled due to DEI considerations, research is being canceled simply because researchers happen to be PhD students or professors at Harvard.

Divyansh Kaushik: That certainly occurred. Interestingly, as we record this, President Trump just announced he’s naming Secretary of State Marco Rubio as interim National Security Advisor and nominating NSA Mike Waltz for UN Ambassador. Developments are unfolding rapidly.

Jordan Schneider: Wow, really? That’s the best possible outcome. I was preparing for the Laura Loomer National Security Advisor era.

Divyansh Kaushik: I mention this because the National Security Council plays a crucial role in this conversation by emphasizing the national security importance of federal R&D. Alex understands this well from his previous position, particularly regarding the critical benefits it provides.

Jordan Schneider: Wait, we need to focus on this for a moment. We haven’t seen this dual role since Kissinger. Is that right, Alex? Has there ever been another person serving in both capacities simultaneously?

Alex Rubin: No, I believe Kissinger was the only one.

Jordan Schneider: This is remarkable. There’s been considerable discussion about NSC reform, as it’s not a fixed organization. I wonder if this presents an opportunity for such reforms, though Rubio’s State Department reforms appear less developed than anticipated or discussed.

I recognize we’re speculating beyond our expertise, but this breaking news deserves attention. There’s a certain Nixon-era quality to these developments. From my perspective, this appointment represents a positive direction — the situation could have been significantly worse than Marco Rubio.

Alex Rubin: Regarding technology policy, Rubio has been at the forefront on issues concerning investments in technology and its centrality to competition with China. During his Senate tenure, his team produced a report examining Made in China 2025. Technology represents the key battleground in this competition. Extending that metaphor, researchers, scientists, and engineers serve as the frontline contributors to American power in this space.

Jordan Schneider: This is astonishing. Earlier today, New York Times articles suggested Waltz had been dismissed — likely someone attempting to shape that narrative. Rubio presents an interesting case, given the contrast between his decade-plus Senate career and his more recent MAGA-aligned positioning. JD Vance seems to embody that perspective more naturally than Rubio. We’ll have to observe how this develops.

Returning to science and technology — before discussing positive aspects, Divyansh, we should address the challenges facing the university ecosystem regarding talent and funding. International students constitute a crucial funding component since most pay full tuition. Government funding represents another vital revenue stream.

Only about 10-15 universities possess multibillion-dollar endowments that would enable them to withstand major external shocks such as losing international students or significant funding cuts without drastic measures like closure or acquisition by private equity firms.

You briefly mentioned immigration concerns, but the situation created genuine alarm among many students who feared leaving the country. Although courts have temporarily reversed certain policies, I worry these uncertainties will linger in the minds of parents worldwide and PhD students considering where to establish their careers.

Divyansh Kaushik: We’ll soon see how this affects enrollment as universities release their yield data. With May 1st approaching and April 15th being the deadline for students to accept or decline offers, that information will become available shortly.

We’ve already observed an 11% decrease in international student enrollment between March 2024 and March 2025, mirroring a similar trend between March 2016 and March 2017. We must monitor this data closely.

Regarding the importance of this population, people often overlook that international students comprise 60% of Computer Science and AI PhDs, and approximately 50% of all STEM PhDs and Masters students. Replacing this talent with domestic students would require considerable time and concerted effort.

Notably, the number of domestic computer science undergraduates pursuing graduate degrees has remained unchanged since approximately 1990-1995. China, with four times the U.S. population, produces twice as many STEM PhDs, twice as many STEM Masters, and four times as many STEM Bachelors graduates.

We can no longer credibly claim that their STEM education or research quality is inferior — they excel in both quality and quantity. Based purely on numbers, our only viable competitive strategy involves recruiting talent globally. China has substantially more human resources to dedicate to complex problems than we do, a critical factor in this discussion.

Jordan Schneider: Let’s return to your "On the plus side" perspective, Divyansh.

Divyansh Kaushik: Consider Michael Kratsios’ remarks at the Endless Frontier retreat approximately 16-17 days ago, on April 14. He described an emerging golden age for America, speaking of "the early light of this new golden age," "American hope," and "the possibility of progress through science and technology."

He emphasized that this golden age will materialize only if we actively choose it, then outlined his approach. He discussed how ours was the atomic age and how we must fight to restore that inheritance. He proposed rethinking federal R&D spending through smarter methodologies.

The Biden administration implemented numerous pilot programs in this direction, but making those approaches the primary R&D strategy would represent a significant achievement for the current administration. New experimentation and prize competitions would be particularly beneficial. We must consider how to optimize every dollar spent on R&D.

Grazio emphasized that beyond a protective agenda to maintain American dominance, we need a promotional agenda. We must create a funding environment that clearly articulates our national priorities, enables scientists to develop new theories, and empowers engineers to implement them. Using advanced market commitments would multiply the impact of government-funded research.

His address contained numerous positive elements that create opportunities for the administration to scale these efforts. Now is the ideal time for those with bold ideas to advance them.

Jordan Schneider: We observed the DOGE approach during the first hundred days — not implementing reforms to unlock a better version of government, but simply making cuts. As the DOGE era concludes, we recognize you can’t forcibly impose creative meta-science reforms, though these organizations do need restructuring.

The current energy, insight, and understanding acknowledge that conditions aren’t ideal. Breakthroughs have become less frequent and more expensive relative to expenditures compared to the 1950s-70s. This presents an opportune moment to experiment with new approaches. However, these efforts become significantly more challenging with half the funding and without international talent — risks created by the budgetary constraints, visa restrictions, and confrontational stance toward universities we’ve witnessed in recent months.

Alex Rubin: Yes.

Divyansh Kaushik: My friend Caleb Watney offers a valuable perspective: viewing federal R&D through a venture capital lens, given the substantial VC presence in government. We should measure performance by return on investment rather than by minimizing expenditure. The critical question is how to maximize outcomes from our investments.

Regarding reforming and restructuring agencies, these institutions are generally receptive to change. The National Science Foundation created the Technology, Innovation, and Partnerships (TIP) Directorate before Congress even passed the CHIPS and Science Act, despite some quiet resistance from other directorates. The agencies welcome innovation.

Consider how Department of Energy national laboratories are experimenting with OpenAI’s models as scientific peers for brainstorming. These represent fascinating initiatives by research agencies to reinvent their approaches to research and funding. If the administration pursues this direction, they’ll likely find substantial support from within the agencies themselves, as well as from universities and industry.

Jordan Schneider: We should campaign for Irwin as NSF Director!

To conclude our discussion on America’s research ecosystem, my assessment is that the Vannevar Bush “Endless Frontier” model has, over the past 75 years, delivered some of humanity’s greatest benefits. Setting aside national power considerations — which should be self-evident given that this system helped overcome the Soviet Union and created history’s wealthiest nation — this ecosystem could benefit from reforms. However, it represents the quintessential golden goose that we’ve managed to develop through work, consistency, and some fortunate circumstances.

This represents a national treasure, and what disturbs me most is the risk of crossing thresholds we cannot reverse. Ecosystems like this, when supported, demonstrate remarkable resilience. However, they contain inherent vulnerabilities related to institutions, funding streams, and talent that require continuous replenishment to maintain previous levels of success.

We’ve covered this extensively over the past eight years, following the excitement surrounding and ultimate passage of the CHIPS and Science Act — a period when bipartisan consensus seemed to favor increased investment in basic research. Watching immigration restrictions, culture war issues, and DOGE priorities converge to create perhaps the greatest threat this ecosystem has faced in decades is deeply concerning and something we’ll monitor closely in the coming months and years.

Divyansh Kaushik: American R&D is globally envied, and we should intensify our commitment to it. Universities serve as powerhouses in this system. Simultaneously, they aren’t blameless in many respects and need to engage in introspection regarding why our commitment to academia and universities faces questioning today. I hope many institutions will undertake this self-reflection and emerge stronger.

Alex Rubin: I’d like to add that while we’ve focused extensively on laboratories and academia and higher education, the R&D ecosystem extends beyond these components. It encompasses community colleges, vocational schools, and technical training programs that produce technicians who operate equipment in these laboratories — an absolutely crucial function.

Many major technology companies, particularly those with significant manufacturing operations, primarily employ community college graduates or individuals with technical training rather than PhDs. The semiconductor industry, for instance, has a substantial veteran population, recruiting former mechanics with relevant skills to maintain equipment. These aren’t PhDs, but they possess essential skills for equipment maintenance.

The final component, which speaks to generational investment, is K-12 education. Truly enhancing the quantity and quality of graduates from PhD and master’s programs begins at these early stages. We’ll discuss China’s approach in this area later, but the foundation lies in K-12 education, gradually building technical literacy so that by the undergraduate or graduate level, students’ mathematics and science skills match global standards.

Divyansh Kaushik: The administration recognizes this priority, evidenced by the recent AI in K-12 executive order, which aims to integrate AI education throughout K-12 curriculum to develop an AI-ready workforce in the coming years. Alex’s observation is entirely accurate, which further supports my optimism regarding future opportunities.

Monitoring Chinese Innovation

Jordan Schneider: Let’s discuss China. Alex, when fundraising for the ChinaTalk Institute, which has enabled me to hire exceptional talent tracking China’s developments in AI and biotech, several funders questioned the necessity of such an organization. They assumed the U.S. government adequately monitors China’s commercial technology through open-source intelligence. As someone who has spent the past decade primarily following Chinese science and technology in the commercial sector, how would you respond to that assumption?

Alex Rubin: My response is that it’s fundamentally a team sport. Different organizations — the intelligence community, other government agencies — have comparative advantages in what they monitor. However, when discussing commercial technology and areas where the primary actors aren’t governments but companies, universities, and laboratories, many strategically significant developments emerge from industry rumors and corporate insights.

Effectively monitoring these developments can’t be limited to individuals like myself in my previous role, working in secured environments to examine these issues. It requires a comprehensive approach that incorporates companies and universities.

The space for organizations like ChinaTalk involves engaging the general public. During the Cold War, nobody questioned why developments within the Soviet Union mattered — there was an inherent understanding of their connection to the American economy, jobs, and security. We need to establish similar connections today, explaining why developments like Huawei creating an advanced GPU matter to average Americans.

This is precisely where podcasts like ChinaTalk and similar outlets contribute value — bringing perspectives well-understood in Washington and disseminating them throughout the country.

Jordan Schneider: It’s interesting how you frame this through tactical, operational, and strategic perspectives when analyzing these questions. The flexibility available in think tanks, academia, or whatever category ChinaTalk occupies allows for different approaches.

Alex, what are the Chinese government’s long-term strategic intentions regarding science and technology?

Alex Rubin: I’ve settled on what I believe is the most accurate characterization, paraphrasing Matt Damon in “The Martian” — Xi Jinping plans to “science the hell out of China.” That’s his fundamental approach — an all-in bet on science and technology.

Whether examining the economy, military, or internal stability, technology permeates everything. Looking at the economy, Xi’s new catchphrase is "new quality productive forces" — a reinterpretation of classic Marxist-Leninist productive forces theory that essentially asks how technology can improve economic performance.

Key components include upgrading traditional manufacturing through robotics and AI automation. Another focus involves eliminating technological choke points by making China more self-sufficient through innovation and R&D investments, enabling Chinese companies to develop domestic alternatives to technologies they currently source from foreign providers. A classic example is photolithography, where significant investment is directed toward Chinese companies like SMEE to reduce dependence on lithography systems from the Netherlands.

The third component focuses on future industries. Last year, China identified six broad categories and numerous specific technologies for targeted support in their Future Industries Development Action Plan. Some might seem far-fetched, including humanoid robots, quantum technologies, artificial general intelligence, and brain-computer interfaces.

They’re absolutely serious about leveraging these technologies for economic benefits. China recognizes that the United States, through its R&D ecosystem, positioned itself to dominate high-revenue sectors of the modern economy. China aims to dominate these sectors moving forward and is investing accordingly.

Regarding social concerns, Chinese leadership prioritizes issues like social stability that could potentially undermine the Party’s control. Their solution involves technology — AI-based tools to enhance surveillance through facial recognition, gait recognition, voice recognition, and predictive analysis. These technology-based solutions monitor and control the population.

Throughout Chinese history, food security has represented the leading cause of revolutions and rebellions. For 22 consecutive years, the first document issued annually by the State Council and CCP Central Committee has addressed rural policy and agriculture — reflecting their significant concern about food security partly due to limited arable land and pollution. Again, their solution involves technology-based approaches to improve agricultural output.

Examining China’s strategy from the reform and opening period to the present reveals consistent prioritization of scientific and technological investment, seeking to leverage these advancements across multiple objectives. So yes, the aim is to “science the shit out of China.”

Divyansh Kaushik: To add to what Alex was saying, China has openly stated in many documents how they want to copy the US system. I was testifying last year to Senate Energy and Natural Resources on this topic. The Chinese 13th Five Year Plan explicitly identifies Argonne, Los Alamos, and Lawrence Berkeley national labs as crown jewels of US innovation. China aims to mimic the US national laboratory system to focus on national goals, strategic needs, and target international technological frontiers — all the points that Alex highlighted.

Jordan Schneider: Alex, could you tie that to the basic research ecosystem?

Alex Rubin: Everything I’m discussing and everything China is attempting to do is fundamentally based in the basic research ecosystem and the talent flowing into it. Xi Jinping himself has talked about how basic research is the foundation of China’s technological progress and how talent is the key enabling factor for their development.

There’s a recognition within senior leadership circles in China that to succeed in dominating future industries and technologies, they must start with investments in basic research. They’re facing challenges in shifting their investments away from applied research toward basic research, given their long-standing investments in applied areas. However, there is broad recognition that to be competitive as a global technology leader, you must invest in early-stage research, basic research, and crucially, train people to staff those facilities.

When we discuss basic research, we often focus on building infrastructure — whether purchasing GPUs or constructing data centers to train models. However, you can build the best infrastructure in the world, fill it with the best equipment, and provide unlimited funding, but if you lack people who know how to use that equipment and what to do with it, it accomplishes nothing. It inevitably comes down to having the right people working together with the appropriate training, experience, and connections to advance science and technology.

Jordan Schneider: Alex, what’s your take on the argument that China is looking for "good enough" technology as opposed to Nobel Prizes and truly frontier research?

Alex Rubin: China is essentially pursuing both approaches. China has a different interpretation of what it means to be a technology leader than the US does. The US defines technology leadership as having the most advanced technology and leading cutting-edge research. China defines it as that plus dominating markets and owning most of the world’s markets for key technology products.

For that second part of their definition, you don’t necessarily need the most advanced technology. What you need is technology that achieves perhaps 80% of the capability at 80% of the cost. When you’re looking at dominating markets, you’re considering Sub-Saharan Africa, Southeast Asia, Latin America — places that aren’t necessarily capable of affording the most advanced technology but still want the benefits advanced technology can provide.

A classic example is Huawei circa 2019. For a very long time, from its early stages, Huawei wasn’t seeking to be the world leader in telecommunication equipment. Instead, it developed technology that was "just good enough" and offered it to countries that couldn’t afford the best American options. Eventually, once the US exited the telecom equipment market, Huawei competed with European offerings by providing discounted prices and generous financial incentives — possible because of government funding, subsidies, and state support.

If you track the rollout of telecom networks through the 2000s and 2010s, from 2G networks to 3G and 4G networks, Huawei’s secret sauce was entering emerging markets in the Global South, offering technology that was perhaps not as good but cheaper, thereby gaining a foothold in these network buildouts. This strategy gave them significant revenue and market share, which they reinvested in research and development.

By 2019, they not only owned most of the world’s 4G network infrastructure but had leveraged their profits from markets like Sub-Saharan Africa to invest in 5G technology, which at the time was both better and cheaper than competitors’ offerings. This put the US in a difficult position because, arguably for the first time in modern history, it faced a major critical infrastructure buildout without a US company in the running, confronted by a Chinese company offering equally good technology at a lower price.

When we talk about "good enough" technology, it’s about broadening our definition of what it means to be a global technology leader. It emphasizes that leadership isn’t just about cutting-edge innovation but also about scale and presence in markets worldwide.

Jordan Schneider: Eva Dou, who wrote the excellent The House of Huawei, is on maternity leave, but we’ll get her on ChinaTalk at some point. One of the fascinating lessons from the Huawei story is that even though the government was pushing firms to do more R&D, the decision to spend an absurdly high percentage of revenue on R&D was Ren Zhengfei’s decision, not a government mandate. This made Huawei an outlier compared to rivals in China like ZTE, which invested only 5-7%. It demonstrates the interaction between government support, domestic scale, and visionary founders who see the long game. These founders understand that to build the most advanced technology company on the planet, you need to do the work yourself — you can’t just steal it.

Alex Rubin: You can steal your way to parity, but you can’t steal your way to leadership.

Jordan Schneider: Totally.

Alex Rubin: Another key point we mentioned earlier is the interconnection between research, customers, and manufacturing. That’s exactly what Huawei built for itself after studying success stories in the US and elsewhere. Huawei functions as both manufacturer and designer with a secure domestic market, where government support was particularly crucial in its early days. This creates an interplay that makes for a very successful, efficient model for advancing R&D.

The concerning part is that we’re seeing these same dynamics play out across multiple sectors today. If we’re not careful, we could find ourselves in the same position later this year or within two years, where critical infrastructure sectors are either reliant on Chinese technology or forced to choose between a Chinese supplier or paying more to be less competitive by going elsewhere.

Jordan Schneider: The broader American media and political ecosystem is only starting to process that China will be ahead in major commercial technologies over the next five years. We’ve already seen it in drones, telecom, and electric vehicles. Regardless of where the Chinese macroeconomic environment or American science funding goes in the coming years, we’re entering a new dynamic.

The trade-offs of keeping these technologies out of the US — which is broadly what we’ve decided to do for drones, telecom, and cars — creates a strange situation. Another important part of the China story was export discipline and the fact that many of these firms, at least in their early days, really had to compete to achieve scale domestically, both with other firms from different provinces and against companies like Apple and Tesla.

From a policy perspective, we need to remember that just because we don’t see these cars here doesn’t mean they don’t exist. They’re getting better, winning in third markets, and forcing GM, Ford, and Tesla to improve.

Alex Rubin: Even if what a Chinese company offers right now isn’t as good or is more expensive than what a US company offers, they’ve consistently shown they can leverage legacy technology to eventually move up the stack and position themselves to achieve world-leading technology. Focusing solely on where they are right now and the current quality of their offerings misses the future risks of their ability to leverage "good enough" technology to eventually generate world-leading technology, whether through theft or innovation.

Jordan Schneider: Divyansh, do you want to say anything about this?

Divyansh Kaushik: Look at where Huawei is now compared to where it was in 2019, as Alex pointed out. It’s expanding everywhere — building data centers, producing cell phones and laptops, operating undersea cables, and investing in EVs. It’s no longer just a telecom company. The same pattern is true for many other Chinese companies.

Alex’s point about projecting forward rather than just looking at a static moment in time is extremely critical, especially as we try to implement more "protect and promote" actions. We should consider where these companies want to be. They’ve laid it all out openly. Made in China 2025 was not a hypothetical document — they met every objective. The AI 2030 plan was not hypothetical either — they’re on track for their 2025 goals.

We’re sometimes overconfident about how significant our lead is. We have an uncanny ability to underestimate China’s capacity to out-hustle everybody. This is something people should be careful about.

Jordan Schneider: When I was fundraising for the ChinaTalk Institute, which now exists and is still taking donations — we’re doing great work around Chinese AI, biotech, and strategic competition — a number of funders asked if the Intelligence Community already has all this China technology information covered. What would you say about what the IC can and can’t do, and the utility of people writing about these topics independently from the government and publicly?

Alex Rubin: Fundamentally, we’re talking about commercial technologies and commercial entities — companies and academia. Many of these industries are relatively small where key players know each other. There’s an inherent need for platforms like ChinaTalk and a key role for industry participants.

Many key insights that might be technological in nature but have significant strategic implications are rumors circulating within industry or insights that companies gain from talking to customers or partners, whether in China or elsewhere. There’s significant value in that information.

Unlike analyzing the Chinese military or leadership, which is a very different challenge, the targets and developments you’re looking at here are fundamentally different. There must be a role for entities that monitor the open-source ecosystem. Otherwise, you risk missing significant trends and developments.

Last point — while we spend a lot of time discussing how China is different and how Chinese companies operate differently, they still have profit motivation, even if somewhat reduced or circumventable when needed. Many of these companies actively publicize their developments in industry press and within their ecosystem because they want visibility. They want to broadcast their technological advancement and development.

There’s so much available in open sources — small technological developments with massive strategic implications. Something like China making progress toward more advanced semiconductor manufacturing is highly technical but has enormous strategic implications for U.S. export controls and AI policy.

Jordan Schneider: I think my answer is that if you’re comparing military to commercial intelligence, maybe it’s 80/20 or 90/10 on the military side — the interesting information requires hunting and digging in an intelligence community way. On the commercial side, it’s the inverse, where 90-95% of what you need — maybe not in a specific tactical way, but at a larger strategic level of what it means for America, industrial policy, or science and technology policy — you can get by just reading publicly available information.

The relative openness of the Chinese media ecosystem when discussing commercial technologies versus operational military plans is completely different because companies need to win domestic market share, hire people, get workers excited about their companies, and raise money from investors. All of that happens under a journalism ecosystem which is, for the most part, fairly free. It’s valuable to surface this information if you have the language skills and context to process it and share it with an English-speaking audience.

Subscribe now

Alex Rubin: China’s war plan for tech is essentially their industrial policies, which they release constantly. Made in China 2025 was a very specific tech dominance plan that detailed their goals down to controlling specific percentages of industries or producing certain percentages of components. You can’t get more detailed than that.

Jordan Schneider: Before this episode, Alex, I asked if there was one document you wanted to discuss, and you pointed me to this mobilization war plan. We’ll link it in the show notes, but could you give listeners some context on why you think it’s interesting and important?

Alex Rubin: The document is from May 2022 — it’s a leaked transcript from a tabletop exercise, a war mobilization simulation in Guangdong province. What makes it interesting is that it includes representatives from the party, military, and government, all brought together in one room.

The scenario presented is essentially: "We’ve decided to invade Taiwan. What does the province do?" The focus isn’t necessarily on military movements like positioning naval vessels, but rather on how to mobilize the population and economy. In very detailed fashion, it discusses converting civilian manufacturing industries to wartime production, specifically calling out the shipbuilding sector, drone manufacturing, and other high-tech industries.

This provides a fascinating example of China preparing for potential major conflict with the US, not at the national level, but at the provincial level. They’re thinking through how to leverage their economy in wartime. If this sounds familiar, it’s basically similar to the US approach to war mobilization during World War II — that’s the scale and framework they’re considering.

They’re planning to leverage the benefits of China’s decades-long investment in expanding manufacturing capacity to essentially outproduce the US in the event of a conflict. The transcript is surprisingly detailed and covers all their considerations, from mobilizing reserves and recruiting people to converting maritime industries, aerospace repair yards, and organizing militias — everything is covered.

Jordan Schneider: When I read this, I thought it might be somewhat performative — Americans do nuclear war games for entertainment, after all. There’s something about the history of the Chinese Communist Party where national mobilization is portrayed as the most exciting time to be alive. But your sense is that I shouldn’t dismiss this entirely. Convince me otherwise, Alex.

Alex Rubin: You can find evidence of these activities in local Chinese press — I found examples just by using ChatGPT to search for relevant articles. There are numerous instances at county and prefectural levels, as well as provincial levels, of similar exercises being conducted. These are part of a comprehensive national system called the National Defense Mobilization system, which establishes cross-party government-military committees at national, provincial, county, and prefectural levels.

For example, in November 2020 in Chongqing Municipality, they conducted a mobilization exercise where civilian manufacturing companies temporarily switched their production lines to make ATVs. While not particularly advanced technology, it demonstrates them testing their capabilities.

To put this in a US context, these county and prefectural level exercises would be equivalent to officials in Fairfax County or New York City planning how they would mobilize to support a national-level conflict in the Pacific. It shows the scale and depth of their preparation and system-building.

Jordan Schneider: For our next episode on ChinaTalk, we’ll Twitch stream America’s national mobilization war plan.

Mood Music—Divyansh’s request as the song capturing the essence of China’s S&T strategy

Created by the United States Congress in 2022, the National Security Commission on Emerging Biotechnology (NSCEB) is an independent, bipartisan body tasked with assessing biotech’s national security implications. Modeled after the earlier AI Commission, the NSCEB recently released its report of recommendations to ensure the US's future leadership in biotechnology.

Their report signals something important: biotech is moving from policy background noise to strategic priority. But key parts of the discussion remain murky. The term “biotech” is used broadly, without disaggregating research from regulation and pharmaceuticals from agriculture. “China” is everywhere in the report — proving the need for additional analysis of the real ins and outs of Beijing’s biotech system. This article attempts to provide that supplemental understanding.

Let’s start by unpacking the first problem: what is biotech?

What is biotech?

The NSCEB defines biotech as “the design and engineering of biological systems, devices, and parts,” which intersects with a range of sectors: defense, industrials, consumer goods, healthcare, agriculture, and energy.

A whole-of-nation, holistic approach to biotechnology makes sense — to a point. Gene editing tools, bioengineering platforms, and computational models of biology do have broad cross-sector potential. But to craft policy, allocate funding, and set regulatory guardrails, the catch-all term “biotech” is too blunt to be useful. For each biotech sector, the stages of development, degrees of maturity, funding dynamics, regulatory environments, and end-user stakes vary so widely that generalizing becomes a liability.

Consider a few examples:

• Health biotech: a mature, well-capitalized sector about one-half of the total biotech market. It operates within a highly structured regulatory system, is driven by both public research and large-scale private investment, and faces challenges with cost, access, and long timelines.

  • Examples: mRNA vaccines, biologic drugs for cancer treatments

• Agricultural biotech: an established sector with capital markets less than one-tenth the size of pharma. It operates in its own silo of regulation and is highly influenced by public perceptions.

  • Examples: genetically modified crops, biofertilizers

• Industrial and biomanufacturing biotech: an emerging field that faces major scale-up challenges. Success depends less on regulatory approval and more on economics: whether these products can compete with traditional alternatives.

  • Examples: biofuels, bioplastics

• Defense biotech: mostly in the R&D and prototyping phase. Funds come almost entirely from government sources, goals are strategic rather than commercial, and products are often dual-use or classified.

  • Examples: biosurveillance,1 combat medicine

There’s no easy way to capture everything biotech touches in a single report, and the NSCEB’s efforts to visualize the big picture are important. Still, the complexity and rapid evolution of the field is exactly why specificity matters. Effective biotech governance requires a modular, sector-specific approach that aligns policies with actual economics, risk profiles, and social impacts.

Tailoring policy to the needs of each biotech sector will also sharpen how we evaluate progress. One of the benchmarks the US is already using is China — so it’s all the more important to take stock: what’s happening across China’s biotech landscape?

What’s going on in China?

China is the organizing principle of the NSCEB report. Chinese government and private sector activity shaped policy recommendations, defined success metrics, and fuelled a sense of urgency.

But if the Commission wants to meet the challenge it outlines, US policy must be grounded in a more detailed understanding of China’s biotech system: its strengths, its weaknesses, and its trajectory. Misreading the landscape risks building American strategy on flawed assumptions, allowing public resources to flow into the wrong problems or reinforce the wrong incentives.

So, let’s turn headlines into hypotheses. What follows is a toolkit for transforming common claims about China’s biotech sector into sharp, curiosity-driven questions – the kind that demand deeper analysis but can inform good policy.

1. The Money Question

Claim: China is spending billions on biotech.

Question: How much is China investing in biotech, and how well is it being spent?

China has identified biotechnology as a strategic priority. Budget allocations reflect this shift: China spent roughly US$34.5 billion (RMB 249.7 billion) on science and technology in 2024, and earmarked US$55 billion (RMB 398.12 billion) for 2025. However, disaggregating biotech from broader S&T initiatives — and parsing where spending falls across research, development, and commercialization — remains difficult. For instance, China’s 2024 investment of US$4.17 billion in bio-industrials and biomanufacturing is hard to categorize by stage. What’s clear is that the bulk of China’s biotech policy continues to prioritize biopharma, the most dominant part of the sector.

The efficiency of China’s public biotech funding ecosystem also requires investigation. One Chinese professor observed: “Money is no longer a major issue… the problem is with the funding system.” Government ministries, national foundations, and provincial and municipal governments all deploy biotech-linked funds, and a lack of coordination and transparency between actors can lead to redundant or wasteful spending. Fund allocation is also uneven, often favoring large, top-down projects aligned with national priorities — an approach that can be ill-suited to biotech, where breakthrough directions are inherently unpredictable. High competition and the short-term structure of grants further limit incentives for more high-risk innovation.

Private capital is equally important to understand. In key application areas like pharmaceuticals and agriculture, private investment plays an outsized role in shaping what gets developed and what reaches the market. China was the second-largest destination globally for biopharma venture funding in 2018–2019, raising around US$60 billion. Such a metric is less impressive in comparative terms: the US raised US$212 billion in the same period, and China’s innovative capabilities remain limited by its less robust private sector. But since the COVID-19 pandemic, China’s capital markets have cooled, and many biotech firms have scaled back R&D. The survival rate and long-term viability of these Chinese start-ups remain undecided.

What should the rest of the world do?

Following China’s actions, other nations could also identify biotech as a national priority deserving serious investment. The NSCEB’s primary recommendation for the US government to spend “minimum of US$15 billion over the next five years” on biotech is a necessary first step – especially at a time when US R&D spending as a share of GDP has declined significantly2 and federal science funding faces mounting pressure from grant freezes and budget cuts.

At the same time, China’s “whole-of-nation” approach to biotech comes with clear challenges: bureaucratic fragmentation, inefficient capital deployment, and misaligned incentives. For the rest of the world, effective biotech policy need not require the same holistic approach. For instance, the US already has a strategic advantage in its robust, innovation-driven private sector – the question is how to unlock that capacity more systematically. The NSCEB’s proposals for an ‘Independence Investment Fund’ to support high-potential startups, advance market commitments (AMCs) and offtake agreements to smooth demand signals, and the restoration of full R&D expensing under the tax code are smart practical levers to promote biotech.

Strong policy deploys public capital to make innovation easier: incentivizing early-stage R&D, derisking scale-up, and building stronger bridges between public goals and private execution.

2. The People Question

Claim: China is building a massive biotech workforce of domestic and global talent.
Core Question: What is the size of China’s talent base, and how sticky is it?

Over the past two decades, China has modernized its education system with the goal of cultivating world-class science and technology talent. Government education expenditure as a share of GDP rose from 2.5% in 1998 to over 4% since 2012 and has stayed above this benchmark through 2024. Such investment spans all levels of education: primary and secondary schools are integrating more STEM into their curriculums, and the number of STEM undergraduate and PhD degrees awarded in China grows each year. 45% of China’s STEM PhDs now come from top “double-first class” universities, indicating their quality.

However, academic programs may still fall short when it comes to preparing talent for biotechnology. In one survey, a third of Chinese biopharma companies reported persistent R&D hiring gaps, citing academic curricula that lag behind industry needs. Industry experience, including the ability to translate research into commercial products and to build and lead globally competitive biotech firms, remains underdeveloped. This gap in managerial and translational expertise has become a core constraint on ecosystem growth. While China’s high-skilled STEM workforce is expanding, it still represents a relatively small share of the overall population. The imbalance between supply and demand has led to intense competition for talent — reflected in an 18.0% active turnover rate in pharmaceutical R&D in 2020.3

Beijing is also seeking to bring back Chinese professionals educated or employed overseas. The National Science Fund for Distinguished Young Scholars supports scientists conducting basic research, provided they spend at least nine months a year in China. The Thousand Talents Plan offers returnees signing bonuses, high salaries and funding, housing assistance, and family support. Such programs, while not unique to China, intend to strengthen the country’s domestic research and innovation capacity.

These programs have mixed effects. Government statistics report a growing proportion of returnees each year. At the same time, both the flow of Chinese students going abroad and the rate at which they stay overseas after graduating have held steady. As of 2019, for instance, over 90% of Chinese AI talent educated in the US have chosen to remain in the US. While China has become more attractive for returnees — thanks to rising living standards, a growing private sector, and increased R&D investment — many of the original push factors remain. Concerns about academic and political freedom, limited job prospects, and digital censorship continue to shape decisions to stay abroad. Some of those who do return are frustrated with lower-than-expected salaries, shortages of postdoc positions and jobs, and reverse culture shock.

Demographic pressures are pushing China to diversify its talent base. China’s college-aged population is declining and has been on a marked downward trend for more than a decade. Ongoing labor market challenges are likely to further drive down Chinese postgraduate enrollment. Beijing has made efforts to attract global talent, including a series of immigration reforms in 2017, but those reforms have yielded limited results. Since 2017, China has risen only modestly in the Global Talent Competitiveness Index4 and still ranks below the top 35 countries. Fewer than 7% of the country’s PhD enrollments are foreign. Political, professional, linguistic, and cultural barriers continue to limit China’s appeal to global researchers.

What should the rest of the world do?

As the NSCEB observes, demand for biotech talent is growing faster than supply. China’s comprehensive workforce strategy — from early education to postgraduate opportunities — should push countries like the US to invest just as broadly in domestic and global talent. And with clear cracks in China’s own approach, now is the moment for others to act.

Fewer than 30% of American public school biology classes include molecular biology content,5 the foundation of most biotech today. Undergraduate biology curricula can be overly rigid and siloed, failing to match biotech’s interdisciplinary nature. Federal and state investments in modern lab infrastructure, teacher training, and interdisciplinary STEM curriculum development could begin to close the preparation gap.

Undergraduate and graduate research opportunities also remain unevenly funded, and too many STEM students face degree-to-career dead ends. In a field as technical as biotech, talent needs to accumulate experience and credentials in a way that stacks. That means investing in well-paid postdocs, lab-intensive training, and academic–industry bridges that allow talented people to do innovative work. One analyst notes this is where the US holds a comparative advantage over China: postdoctoral programs that offer real research experience and career growth.

At the same time, the US’s ability to attract and retain global talent is one of its greatest strategic assets. Over 75% of international students who earn STEM PhDs in the US stay for at least a decade, contributing to vital sectors like AI. Chinese officials regularly cite such US retention of Chinese talent as a key obstacle to their national goals. The NSCEB report cites Jeremy Neufeld, who emphasizes the importance of STEM immigration for the defense-industrial base, writing:

Existing restrictions on STEM immigration — and the resulting backlogs and waiting times for STEM talent — hamper the defense industrial base’s growth. Without reducing the barriers to high-skilled immigration, efforts to onshore and strengthen critical industries in the United States will face significant hurdles, and may fail altogether.

To maintain an edge in biotech, the US needs to double down on reinforcing the pull factors that make it the world’s premier destination for scientific and technical work while addressing the growing push factors that drive talent away. Staying attractive for global talent requires investing in research and building secure career pathways. At the same time, streamlining visa transitions and expanding green card availability for advanced STEM grads can help send a clear cultural and political signal that global talent is welcome.

3. The Supply Chain Question

Claim: The US is dangerously dependent on China for pharmaceutical supply chains.
Core Question: How concentrated is China’s role, and how relevant is this to biotech competition?

Pharmaceutical supply chains are commonly cited as a source of concern when it comes to China and biotechnology. However, available data does not support the narrative of overwhelming US dependence on China for pharmaceuticals. While China plays a significant role in upstream chemical manufacturing, pharmaceutical supply chains are globally distributed, multi-step, highly fragmented, and frequently outsourced — making US exposure indirect and difficult to isolate.

For example: in 2024, China was responsible for less than 6% of US imports of finished pharmaceutical products.6 As for APIs (the component of a drug responsible for its therapeutic effect), current data suggests that between 2014 and 2022, 17% of APIs used in medicines in the US were imported from China. In 2024, that share fell to 8% for prescription medications. The oft-cited “80% of APIs come from China” is an unsubstantiated misinterpretation of a government statement that has since been clarified. India and Ireland are both larger exporters of APIs and finished medicines to the US than China.

China’s footprint in pharmaceutical supply chains is primarily upstream, producing key chemical inputs and APIs that pass through multiple global intermediaries before reaching consumers. This prominence is largely in small-molecule generic drugs, which are manufactured through chemical synthesis. This matters because chemical manufacturing and biomanufacturing are fundamentally different in terms of core technologies and processes. Much of what’s described as “biotech supply chain risk” is actually about commodity chemical inputs for generics — a sector adjacent to, but distinct from, modern biotechnology.

China’s footprint in biomanufacturing requires investigation into biologics (medicines derived from living cells, rather than through chemical synthesis; typically large and complex molecules such as proteins and antibodies). In biopharmaceutical supply chains, China’s presence is limited but its capabilities are growing.

Chinese biopharmaceutical manufacturing capacity represents about 10% of the worldwide total, third globally behind the United States and European Union. Much of this capacity was built after 2016, following a major regulatory change7 that opened China’s market for contract manufacturing organizations (CMOs) that provide small-scale and large-scale drug production services to pharmaceutical companies. Multinational CMOs like Boehringer Ingelheim, Lonza, and Samsung Biologics now operate facilities in China. Among domestic CMOs, WuXi Biologics is widely considered a biomanufacturing leader by the industry, though it has recently faced setbacks due to political scrutiny. ​Following Beijing's recent pledge to accelerate biomanufacturing, China's capacity in this sector is poised for significant growth.

What should the rest of the world do?

De-risking pharmaceutical supply chains from China starts with recognizing their complexity. Exposure to China is real but often indirect. When it comes to small-molecule generics in particular, the economics of onshoring rarely add up on their own. These drugs operate on razor-thin margins, giving manufacturers little incentives to invest in quality or resilience – even with tariffs in place. Building more resilient supply chains at home will require accepting trade-offs such as higher costs for payers and consumers. Meaningful incentives for domestic production will require government intervention through policy tools such as targeted subsidies, publicly-funded manufacturing infrastructure, and expansion projects such as BARDA’s work to build out domestic biomanufacturing for vaccine production and distribution. And as supply chains are restructured, regulatory oversight will need to scale as well. In the US, a stronger, better-resourced FDA will be essential to maintaining quality, consistency, and compliance.

On the biotech side of pharmaceutical supply chains, the focus should be on reinforcing homefield advantages and lowering the barriers that keep domestic firms from scaling. One key advantage is talent: biomanufacturing depends on engineers, technicians, and regulatory specialists with deep expertise. Strengthening training pipelines — through programs like the National Institute for Innovation in Manufacturing Biopharmaceuticals, a public–private partnership highlighted by the NSCEB — can help expand this critical workforce. Regulatory credibility is another asset the US shouldn’t take for granted: the FDA still carries far more global trust than its Chinese counterparts. Meanwhile, high startup costs for facilities, equipment, and compliance systems remain a major barrier for domestic firms, requiring targeted investment to lower the threshold for entry.

In Sum!

“China” is often invoked as if it were a unified actor with perfect coordination and seamless execution. Reality is more complicated. Behind some impressive numbers and case studies are real growing pains: capital inefficiencies, inexperienced management, limited regulatory expertise, gaps in early-stage funding, underreported failures, uneven enforcement, and dependence on global partners.

In this respect, China is not an exception. Innovation is rarely a simple narrative of linear progress. But as more eyes turn to China, and as real changes unfold within its biotech ecosystem, careful observation and accurate analysis become all the more essential. Meeting this moment requires asking the right questions to understand how capital is deployed, how talent systems function, and which parts of the supply chain actually matter.

The point isn’t to downplay China’s rise in biotech. It’s to understand it on its own terms — and in doing so, craft smarter, more grounded policy.

2

6

How is Trump changing America’s extended deterrence regime? I got Polymarket to make a market on whether a US ally will acquire nuclear weapons in 2025. It’s currently trading at 6%. Are we buyers or sellers?

To discuss, I interviewed Vipin Narang, professor at MIT, who served as Acting Assistant Secretary of Defense responsible for nuclear deterrence policy during the Biden administration; Pranay Vaddi, a senior fellow at the Center for Nuclear Security Policy at MIT who worked on arms control and non-proliferation on Biden’s National Security Council; and Junichi Fukuda, senior research fellow at Tokyo’s Sasakawa Peace Foundation.

We get into…

• The historical development of the American nuclear umbrella, including the “software” and “hardware” components of deterrence,

• The probability that an American ally will proliferate by 2030, and which countries are the most likely candidates,

• Why France proliferated despite US objections,

• How the world might respond to nuclear ambitions from Poland, Japan, or Saudi Arabia,

• China’s nuclear modernization and deterrence strategies for a multi-polar world.

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

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

Foundations of Non-Proliferation

Jordan Schneider: Let’s start with a history of American extended deterrence. Where did this concept originate, and what has it achieved over the past 75-plus years?

Vipin Narang: Extended deterrence is the notion that the United States will use its range of capabilities — including nuclear weapons — to defend its formal allies and partners against nuclear attack from a common adversary. It also covers high-consequence non-nuclear strategic attacks, which is a mouthful but really means an existential threat to our allies.

The concept developed in the mid-1950s after World War II. European states had been devastated by the war and could not defend themselves against the emerging Soviet threat. The United States maintained its forward conventional deployment, but our huge advantage over the Soviet Union was nuclear weapons. The Eisenhower administration quickly realized that against a conventionally superior Soviet Union, we might need to increasingly rely on nuclear weapons to defend our forces and our allies, which were still rebuilding.

We never implemented extended deterrence out of altruism or charity. It benefited the US by preserving the openness of Western Europe and our East Asian allies, particularly South Korea and Japan, after the Korean War and World War II. This allowed us to keep the economies and political systems of our allies free, democratic, and capitalist, which created open markets for the US.

Over time, extended deterrence became important for another reason that is often overlooked — it became our best non-proliferation tool. We could prevent our allies from pursuing their own nuclear weapons if they believed our nuclear umbrella would credibly deter common adversaries. Kennedy predicted that we’d have 25 nuclear weapon states by the end of the 1960s — that was largely prevented by a credible extended deterrence architecture that the United States developed and maintained.

We worked with allies not just to deter common adversaries — China, Russia, North Korea — but to assure allies that we would come to their defense. Making this credible requires convincing allies that we can fight on their behalf without losing American cities or allied cities. The core of making extended deterrence work is being able to substantially limit the damage that Russia or China can impose on both the ally and the US homeland.

Regarding the non-proliferation point, extended deterrence has a remarkable record of preventing ally proliferation. Of our 34-35 formal allies, France is the only state that consciously decided to acquire nuclear weapons because it believed the US umbrella wasn’t credible for France’s defense. We can discuss some corner cases, but France remains the only example where perceived lack of credibility in extended deterrence caused a formal ally to develop nuclear weapons. Until now, it’s only been one case, and we hope it remains that way.

Jordan Schneider: Let’s explore this “Berlin-for-Boston” dynamic. The assumption throughout the Cold War and after was that no president would really trade Boston for Berlin. What was needed was a defense deterrence package that could plausibly threaten to keep Berlin free while avoiding an escalatory spiral that would leave Cambridge, Massachusetts in thermonuclear flames. What strategies has the US used over the past 75 years to make deterrence credible both to Moscow and to our friends in Asia and Europe?

Pranay Vaddi: You can’t separate the military dynamics and strategy from this discussion. What matters is the relationship between the head of state in the United States — the nuclear guarantor — and the head of state in the allied country receiving the nuclear umbrella. They need to be aligned. There must be a close political relationship between the United States and the 35 allies that Vipin referenced as nuclear umbrella allies.

When you put that relationship at risk or question whether the US will honor its defense arrangements — for example, if the United States declares at the presidential level it won’t defend an ally unless that ally spends a certain percentage of GDP on defense — it creates problems. Regardless of whether the United States maintains its nuclear modernization program, keeps forward-deployed dual-capable aircraft in Europe, or shows off B-2 bombers every time North Korea tests a missile, these mechanics can continue. However, every ally knows that it’s the US President who decides whether to use nuclear weapons in defense of an ally.

When the US President isn’t committed to an extended deterrence strategy as we have been for the past 70-80 years, it creates doubt. This is what we’re seeing today — countries wondering if the US will actually honor its guarantees. Because of these doubts, Poland and South Korea in particular have started discussing how to hedge against potential US abandonment of nuclear guarantees.

Just this week, we’ve seen continued statements from French officials about what a European deterrent could look like — something that Vipin and I have written about extensively. Poland has indicated it would welcome both US and French weapons. But ultimately, you need to combine military strategy with political relationships.

One example of this approach is when Vipin and I, in our previous roles, worked closely with South Korean counterparts to establish the Washington Declaration at the presidential level when President Yoon visited Washington in April 2023. This required extensive work below the surface — Vipin leading discussions at the Defense Department with his counterparts in South Korea’s Ministry of Defense, alongside White House interactions with the presidential office in Seoul. The goal was to create public language that both countries could agree to, demonstrating a renewed and enhanced commitment to extended deterrence with commitments from both sides.

South Korea needed to recommit to the Non-Proliferation Treaty, while we needed to recommit to South Korea’s defense. We agreed to do so more transparently, giving South Korea greater insight into US nuclear planning and more input into how nuclear operations could be conducted. We needed to address both military and political aspects to provide reassurance in that bilateral relationship.

Jordan Schneider: Let’s stay on history a little more. We’ve got the credible military component — “I can actually bomb places” — and we’ve got the credible political aspect — “I actually care about your country enough to do something if it gets attacked.”

Vipin Narang: We called it “hardware” and “software.” The hardware piece is actually easier because it’s about having flexible capabilities to limit damage to the ally and the homeland. The theory is relatively simple, but implementation is not. This is something the Department of Defense spends considerable time contemplating.

Pranay’s point is really important — we underemphasize the importance of the software piece both historically and today. The only change from the Biden administration to the Trump administration is the software piece. That alone explains why we’re having this podcast. It’s almost like a blue screen of death for some allies. They don’t question our capability — they question whether, when the time comes, we will actually attempt to limit damage to the ally and US homeland and fight on their behalf.

Jordan Schneider: Let’s explore the history a bit more. What have been the sketchiest software moments since 1950?

Vipin Narang: The Eisenhower administration raised concerns right out of the gate. The plan was to reduce our conventional footprint, but that conventional presence in Western Europe reassured allies because we had skin in the game. Without skin in the game, the fear was we wouldn’t pull the trigger if the Red Army came crashing through.

The debate during the transition from the Eisenhower administration to the Kennedy administration centered on whether we had flexible enough options for a theory of escalation management. There have been various ups and downs, but what we’re seeing today is new: questioning whether we even care about our allies. Article 5 is a hallowed pillar of NATO and essentially our commitments to South Korea, Japan, and Australia.

For the first time, we’ve had an administration that basically says we might not honor Article 5, and allies might be on their own if they don’t pay enough. One of the bigger implications of the Signal chat leak was revealing that in private, the Vice President essentially said, “I hate Europe.” If I were Europe, I might question the willingness — not the ability — of the United States to step in when necessary.

Junichi Fukuda: Regarding the crisis of extended deterrence, we should consider the case from the early 1970s and late 1960s. At that time, the Guam Doctrine, or Nixon Doctrine, resulted in conventional forces withdrawing from the Indo-Pacific area alongside the withdrawal from the Vietnam War. This created a genuine crisis.

There was no clear nuclear extended debate then. Most countries were deciding whether to sign the Non-Proliferation Treaty (NPT) around 1969-1970. At that time, the Japan-US alliance didn’t explicitly mention the nuclear umbrella. The term “nuclear umbrella” was first used in 1975, yet Japan decided to sign the NPT during détente. The tension between the US and Soviet Union had decreased, allowing us to join the NPT.

Officially speaking, there was no extended nuclear debate at that time. This provides an analogy for our current situation — the United States withdrawing forces while the credibility of extended deterrence appears to be in crisis. We must learn from the experiences of the early 1970s.

Jordan Schneider: Perhaps the distinction between then and today was détente. You didn’t really see nuclear war on the direct horizon. China was busy with its Cultural Revolution, and the Soviet Union was being friendly. It wasn’t the most frightening period in thermonuclear history. Does that change the dynamic compared to today?

Junichi Fukuda: Because it was the détente era, tensions were reduced, so we didn’t have to deeply consider the implications of the extended deterrent crisis — we simply accepted it. In our current situation, US-China relations have deteriorated, and US-Russia relations are poor. Tensions are increasing while the United States appears to be withdrawing its engagement with the outside world. This creates a crisis.

Pranay Vaddi: Another historical element worth mentioning — something I always try to write about when Vipin and I collaborate, though he always deletes this paragraph — concerns extended deterrence from the early 1960s. There’s an interesting story about when a German Chancellor visited JFK after his White House victory.

Konrad Adenauer represented the CDU — the center-right in West Germany — and was essentially JFK’s opposite. Adenauer was about 85 years old while Kennedy was this young, new president. Kennedy was deeply concerned about allied proliferation, as Vipin outlined regarding developments in the 1950s and 60s.

According to declassified documents recording their meeting, Adenauer essentially said, “I hear you have weapons in West Germany, but I’ve never seen them. I’ve been repeatedly told those warheads are there, but I haven’t been shown them.” He wanted his Defense Minister to understand what the US was actually doing in their territory — what they had based there.

Kennedy assured Adenauer it made sense and promised that Defense Secretary Bob McNamara would brief him and his defense minister on the US stockpile operational plans — when we would actually use them if the Warsaw Pact attacked.

To me, this marked the beginning of the types of requests we in the US Government would regularly hear from allies — “What are you actually doing in a nuclear scenario?” As mentioned regarding the Washington Declaration and US work with South Korea, Japan, and NATO, we conduct tabletop exercises, scenario-based discussions, and simulations to demonstrate that the US has thoroughly considered extended deterrence scenarios. This marries hardware and software to make extended deterrence effective. Its origins date back to that conversation.

Retrenchment vs Burden Sharing

Jordan Schneider: If we’re using a relationship analogy, we’re saying, “We’re going to give you a tour of our place. We’ll let you bring some toothbrushes into the bathroom.” The US has been doing this 75-year dance with its allies, broadly trending toward reassuring everyone so they feel safe and secure enough not to develop their own nuclear weapons.

The question now is, what does the decay function of that look like? The Trump administration isn’t going to destroy our B2 bombers anytime soon, so the capability isn’t disappearing. However, the trust has developed some real hiccups, particularly on the software side.

Where do we start with this? What’s the right conceptual framework to understand the signals that the Trump administration is sending to the world?

Vipin Narang: Let me quickly address the hardware aspect. We shouldn’t overlook the hardware adjustments that were ongoing or being laid as a foundation at the end of the Biden administration. This includes ensuring the modernization program for each leg of the nuclear triad — we’re replacing each leg simultaneously for the first time in history, and we’re facing challenges.

The Sentinel ICBM is facing delays. Attention and focus are needed to prevent the modernization program from falling further behind schedule. The current administration hasn’t issued an executive order on nuclear forces, though it has addressed missile defense architecture, which is related but represents a very long-term strategy.

The immediate concern is our progress on this modernization program and how we’re dealing with China’s emergence — something we didn’t anticipate when we sized the modernization program 15 years ago. We don’t need forces matching the combined strength of our adversaries, but we might need more than what we planned when discussing New START numbers in 2010. At that time, Russia was considered a partner, and China and North Korea hadn’t begun their expansions.

For an ally examining the hardware aspect, there are mixed signals. The president has stated he doesn’t want to allocate more funds to nuclear weapons, which might be concerning. However, the more pressing issue, as Pranay and Fukuda-san mentioned, is that the software component isn’t experiencing a decay function — it’s a complete discontinuity. The posture toward our European allies has transformed dramatically since early January 2025.

If I were in the Indo-Pacific region — and I’d welcome Fukuda-san’s thoughts on this — there’s considerable uncertainty. The administration has declared China the pacing threat and pledged to focus on China, yet simultaneously continues to pressure Japan and South Korea economically in ways that may undermine the credibility of our commitment to defend them against existential threats from China or North Korea. This credibility is arguably much weaker than it was in January 2025.

Pranay Vaddi: The fundamental question is whether this is about right-sizing burden-sharing between the United States and its allies or pursuing a retrenchment ideology. If you’re an ally like Japan, you can approach the United States saying, “We understand you want everyone to pay their fair share. You want Japan to prioritize China as your primary concern, and we’re contributing to the extended deterrence strategy that Indo-PACOM needs for a Taiwan Strait crisis."

Japan is developing air and missile defense capabilities, long-range conventional strike capabilities, and a blue-water navy. They’re willing to open their shipyards for replenishment and repair of US naval vessels. In return, the US needs to maintain a strong nuclear umbrella and guarantee. That’s the basic exchange, but we should develop this strategy together.

If it’s primarily a burden-sharing question from the US, that’s a strategy to present to the White House: “Let’s determine the appropriate payment amount and capabilities within a bilateral strategy — or trilateral if South Korea is included, or quadrilateral with Australia as another extended deterrence ally in the Indo-Pacific."

However, if this is about pursuing a retrenchment foreign policy — if that’s what “America First” means — which I believe is an unseen battle within the administration, then allies have limited options. As an extended deterrence ally, you can only be so persuasive because ultimately, you need a United States willing to engage with allies, be transparent, and fulfill the software element of extended deterrence. Being credible on the software side seems contradictory to a retrenchment strategy.

Allies need to determine whether this is about recalibrating burden-sharing in the new political climate or if the United States is reducing its international commitments. The second scenario is much more challenging and drives allies to consider hedging strategies.

Jordan Schneider: It’s complicated because the answer is both. Pranay pointed out that some people believe one thing and others believe something else. We have a president who frequently changes his mind. So what’s the appropriate response?

Junichi Fukuda: Regarding the hardware issue, I’m not overly concerned about the modernization of US strategic nuclear forces. The United States is modernizing its strategic nuclear weapons, and there’s an option for uploading nuclear warheads on existing platforms.

My concern lies more with the future trajectory of developing theater nuclear forces, such as sea-launched cruise missiles from submarines. These types of theater nuclear weapons would be essential for Japan’s defense in the future. Given the current administration’s position on cutting the defense budget, I question whether such new nuclear capabilities will ultimately be realized.

On the software issue — the political relationship between the United States and Japan — there are several possibilities. First, there’s conflict in the trade area. The United States has increased tariffs on Japan, and we’re currently negotiating this matter. Japanese people might find it difficult to rely on the United States for security while we’re engaged in economic disputes.

Simultaneously, the Trump administration is attempting to address the challenge posed by China. Most security experts in the administration appear to be “prioritizers” — they would withdraw from the European theater but concentrate on strengthening capabilities in the Pacific to counter China’s threat. This approach provides some reassurance to us.

The situation remains difficult to analyze fully at present. Overall, we remain confident in continuing the alliance relationship for now, though the future is somewhat unpredictable.

Jordan Schneider: It’s interesting, Junichi, because you’ve seen this incredible earthquake in European politics triggered by just a handful of speeches. It seems to me that only people like you are paying close attention, not necessarily the entire Japanese or South Korean political establishments. Am I wrong? Do you think it would take something as explicit as a J.D. Vance or Pete Hegseth speech saying, “We don’t care about you guys,” to trigger the same kind of awakening that seems to have happened in Poland and Germany?

Junichi Fukuda: That’s a nightmare scenario we hope to avoid. Currently, there appears to be conflict within the administration. One faction focuses on confronting China, recognizing the necessity of maintaining US hegemony in the Indo-Pacific region. The other faction consists of isolationists who advocate withdrawing US forces from every theater worldwide, preferring to concentrate resources on domestic politics. The future direction of the current administration remains uncertain, with potential internal struggles ahead.

Jordan Schneider: The very existence of this debate is unprecedented from a post-1945 perspective. Perhaps now is a good time to discuss the politics of nuclear armament in both South Korea and Japan. If we remain in this awkward, subterranean debate — or move toward the more alarming scenario of J.D. Vance expressing indifference toward Japan — how might this affect Japan and South Korea’s nuclear postures?

Junichi Fukuda: Japan’s nuclearization has traditionally been an unthinkable issue, not widely discussed. However, as someone studying nuclear deterrence, I’ve had many opportunities to address this topic. Last month, I attended a conference where American experts frequently discussed the possibility of Japan nuclearizing in the future.

The primary condition under which Japan might consider nuclearization would be a lack of confidence in the credibility of US extended deterrence. A secondary scenario would involve South Korea pursuing nuclearization first, which might prompt Japan to follow suit. The specific circumstances that would trigger such scenarios remain unclear to me.

Pranay Vaddi: From a US perspective, it’s difficult to imagine any scenario where the United States benefits from ally proliferation. We’ve seen comments by the Vice President questioning whether stationing US nuclear weapons in Poland makes sense — indicating that proliferation of US weapons to other allies isn’t necessarily on the table. This struck me as an off-the-cuff reaction rather than the product of an interagency process.

The challenge for the Trump administration is reconciling several objectives — preserving the status quo regarding the number of nuclear-armed countries, upholding the Nuclear Non-Proliferation Treaty, and renegotiating burden-sharing agreements or pursuing retrenchment. These goals conflict with each other.

Japan is a technologically advanced nuclear country with nuclear technology and reprocessing capability. They’ve committed to neither possessing, producing, nor introducing nuclear weapons into their territory as part of their peace constitution. They remain an importantUS partner in nuclear energy, extended nuclear deterrence, and nuclear disarmament. Japan uniquely embodies all the contradictions of nuclear policy in one bilateral relationship.

Countries like Japan, South Korea, or Poland cannot simply decide to acquire nuclear weapons overnight. The process requires specialized facilities, legal and regulatory changes, withdrawal from the NPT, and consideration of potential US export controls and sanctions. This timeline gives adversaries opportunity to react — not just economically. If North Korea observed South Korea pursuing nuclear weapons while US extended deterrence guarantees appeared shaky, they might preemptively strike those facilities or initiate conflict. The risk of conflict increases when an ally with weakened US security guarantees pursues nuclear capabilities — an important consideration for countries weighing this decision.

Vipin Narang: Those who identify as “America first” — which includes all of us, as I reject the notion that only the Trump administration puts America first while the Biden administration didn’t — should oppose allies acquiring nuclear weapons for at least three reasons.

First, we have historically opposed allied proliferation in Washington because we wanted to minimize the number of independent decision centers for initial employment or escalation decisions. This is a sanitized way of saying we didn’t want France starting a nuclear war we’d have to finish. The French — whom I genuinely respect — had a strategy where their independent nuclear capability would be employed when the French President determined vital interests were at stake. However, they couldn’t finish such a conflict, leaving the US to manage the aftermath against the Soviet arsenal. This thinking persists today when discussing the credibility of European deterrence. France can initiate a conflict, but cannot conclude it.

Similarly, other allied nuclear powers could never develop damage limitation capabilities comparable to the United States independently. An allied proliferator would acquire nuclear capabilities that could initiate a nuclear war and chain-gang the US into it — precisely what the US has always sought to avoid.

Second, while Kenneth Waltz famously argued “more may be better” from a retrenchment perspective, the risk of accidents increases substantially. Smaller states with limited arsenals facing more capable nuclear adversaries with head starts might develop itchy trigger fingers or lack the institutional infrastructure to effectively manage these weapons. The accumulated risk of accidents grows with each additional nuclear-armed state.

Third, as Pranay and Fukuda-san mentioned, the threat to the NPT is real, as is the potential cascade effect if one ally — particularly South Korea, given the growing political consensus there for at least developing a hedging capability — acquires nuclear weapons. I doubt the NPT could survive a democratic state in good standing withdrawing to pursue nuclear weapons. Japan might follow, along with Poland, Ukraine, Germany, Saudi Arabia, Turkey, and a number of other countries if the floodgates opened.

Pranay’s point is important: these countries would pursue nuclear weapons in an environment where their adversaries have both a vote and a head start. Historically, no state has acquired a credible nuclear weapons capability in less than five years from the decision point, and a survivable arsenal typically requires a decade or more.

Regarding probabilities — there’s a zero percent chance an ally proliferates by the end of 2025? The timeline doesn’t work. By 2030? I’d wager the odds are more than 50 percent, though as an academic I don’t have much money to bet — that at least one ally will have a credible latent nuclear capability by 2030-2035 if current trends persist.

Jordan Schneider: Just wait until you get all those Polish and Saudi contracts to help write their nuclear doctrine.

Vipin Narang: You can reach Pranay, my agent, at prvaddi@mit.edu.

Jordan Schneider: Let’s talk about France. They were exceptionally difficult throughout the Cold War, no?

Vipin Narang: I love them, but one France is enough.

Jordan Schneider: This is precisely the point about “America First.” Nuclear weapons are incredibly powerful — they provide tremendous leverage, allowing you to dictate terms to other countries, constrain their actions, and influence whether they initiate conflicts.

Consider how the US effectively prohibited Chiang Kai-shek from invading mainland China in 1962 because he lacked nuclear weapons. Our ability to maintain global stability and, as Vipin mentioned, prevent conflicts that we would ultimately have to resolve actually enhances American security. The alternative makes the world a lot more dangerous.

The notion that Putin would remain passive for five years while Poland develops nuclear weapons is unrealistic. We’ve already witnessed this scenario play out with Israel, Syria, and Iran. Russia would undoubtedly take action. Imagine the consequences of a Polish nuclear facility being bombed by Russian aircraft — the global ramifications would be catastrophic.

Vipin Narang: Poland remains a NATO member, so that scenario involves Article 5 and risks war with Russia over a counter-proliferation strike. These are low-probability events.

Pranay Vaddi: Attempts at retrenchment could actually accelerate the path to great power confrontation that you’re trying to avoid. Regarding our discussions with French colleagues — you don’t necessarily need to implement extended deterrence exactly as the United States has done. We’ve pursued an expensive strategy heavily reliant on nuclear capabilities that requires an enormous defense budget.

Perhaps in their Euro-deterrent approach, the French don’t need 3,800 nuclear warheads or a triad similar to ours. They might achieve more with conventional forces and missile defense systems. However, they ultimately need an approach that accomplishes the same objectives: reassuring eastern flank allies on Russia’s borders that you could meaningfully engage in conflict and either deter or defeat Russian aggression against those allies.

This is extraordinarily challenging. The United States has invested over seven decades in continuously refining this approach. The work on extended deterrence is never complete. As Fukuda-san knows well, the meetings between the United States and Japan throughout the year on extended deterrence are extensive, regularly scheduled, and ongoing. It requires dedicated staffing and expertise on both sides. This isn’t something another country can simply adopt if the United States withdraws those commitments.

Jordan Schneider: What frustrates me about much of the MAGA-influenced foreign policy is its simplistic thinking — reminiscent of middle school debate logic.

There are reasons why the world exists in its current state. I thought conservatives valued the principle of Chesterton’s Fence — not dismantling established systems without understanding their purpose. The transition to a new world order could be dangerous and catastrophic.

Spending an extra $500 billion annually to prevent 20 countries from acquiring nuclear weapons actually represents an excellent deal.

Vipin Narang: It is indeed an excellent deal.

Jordan Schneider: A historically unprecedented deal. You would have accepted this arrangement 100 times out of 100 if you were present at Alamogordo in 1945.

Vipin Narang: This represents a bipartisan consensus. Republican or Democrat — Pranay and I often find ourselves aligned with traditional Republican views on strategic deterrence and extended deterrence, sometimes surprising ourselves with how much common ground we share. Those traditional voices are being marginalized, if not completely excluded, from the Trump administration.

This represents a different approach than anything we’ve seen in the post-World War II era. There was bipartisan consensus that we were committed to extended deterrence, with debates focusing on implementation methods rather than questioning the fundamental commitment itself. That’s why this moment seems distinctive.

To be fair, I don’t know where the Trump administration will ultimately land on this issue, as Fukuda-san and Pranay both noted — there are various competing voices within it.

Vipin Narang: At the core of extended deterrence credibility lies predictability and consistency. The very existence of this debate creates problems for the credibility of extended deterrence.

Jordan Schneider: Junichi, how are you feeling about all this?

Junichi Fukuda: At the administration level, I cannot predict the future because President Trump is unpredictable. However, at the congressional level, there exists concrete bipartisan agreement. The United States must continue extended deterrence toward its allies. Two years ago, a congressional report from the Bipartisan Commission — the Strategic Posture Report — emphasized continuing and even increasing the credibility of extended deterrence. At the congressional level, bipartisan agreement will persist in the future. Administrations last just four years, but Congress continues.

Jordan Schneider: I appreciate your optimism. I’m curious, Junichi — when you first entered this field and began studying nuclear extended deterrence and nuclear policy, what were the most extreme scenarios being discussed? How much more extreme is our conversation over the past 45 minutes compared to anything you’ve encountered before?

Junichi Fukuda: I began studying international relations in the late 1990s. At that time, the most severe concerns involved bioterrorism around 1997 or large regional civil wars. Nuclear war between great powers was definitely not considered a possibility. Now we’re discussing nuclear war. The situation has changed dramatically, forcing us to contemplate the unthinkable.

Pranay Vaddi: For the past 30 years, nuclear policy as a subject area was limited to specialists and experts. Now this debate about extended deterrence reminds us how deeply connected US nuclear strategy is to other elements of US grand strategy and foreign policy goals.

After the Trinity test, as nuclear strategy developed, it became integral to American grand strategy for confronting and containing communism, rebuilding decimated countries including close allies and former adversaries like Germany and Japan. This created a remarkable market for American goods and innovation through the Marshall Plan and other foreign aid efforts.

If you disconnect extended nuclear deterrence from foreign policy and our economic strategy, you can have simplistic conversations about whether we should continue extended deterrence or maintain forward-based troops. The reality is these elements have been interconnected for over 70 years.

Vipin and I have written about how much effort the United States put into rebuilding the world in the 1950s and 60s, ensuring that the United Nations, economic integration, and globalism — which nationalist trends in every democracy now resist — would help prevent another world war. Now the administration is attempting to remove the economic pillar while simultaneously questioning sacred principles of extended nuclear deterrence.

Our assumption when we wrote about this in January was that the economic benefits to the United States would eventually lead back to a coherent extended nuclear deterrence strategy. However, when economic interdependence and the accompanying political relationships diminish, I might question the American commitment to extended nuclear deterrence if we become less reliant on international trade due to efforts to retrench and reshore everything from industry to military capabilities.

These elements are all connected. Retrenchment, if implemented comprehensively, can severely damage US national security.

Deterrence with Chinese Characteristics 玉石俱焚

Jordan Schneider: We’ve progressed quite far into this discussion without mentioning China’s efforts to update its nuclear arsenal. This factor seems irrelevant in the Trump calculation, but it’s worth discussing how countries in Asia perceive this development. Junichi, could you provide a brief overview of what has happened in China over the past decade, and what this modernization means for countries in the region as they evaluate their strategic environment?

Junichi Fukuda: Historically, China’s nuclear assets were limited to approximately 200 nuclear weapons. At that time, analysts described China’s nuclear strategy as “minimum deterrence” — though this is a Western concept that might not align with China’s actual thinking.

China has gradually changed its approach to nuclear weapons. Recently, China has rapidly increased its nuclear arsenal, adding approximately 100 nuclear warheads annually. In the long term, China’s strategic nuclear arsenal may reach 1,500 warheads, potentially achieving parity with the United States and Russia.

The purpose behind this expansion remains unclear, but several possibilities exist. One apparent reason is to gain national prestige and achieve the same international status as Russia and the United States by creating parity with them.

Another potential reason involves creating a strategic stability-instability paradox in the Taiwan Strait contingency. By increasing its strategic nuclear arsenal, China might deter United States intervention against Chinese aggression toward Taiwan.

Their final objective, however, remains unknown. China has stated that by 2029, they intend to build a “world-class military,” which likely means achieving the same scale as the United States. This suggests their nuclear forces will eventually match the current US strategic nuclear arsenal.

Jordan Schneider: A question from Japan’s perspective — does the difference between 200 versus 1500 nuclear weapons actually matter?

Junichi Fukuda: Currently, it doesn’t matter as much, but we’re concerned about what would happen in the next decade. Our nuclear concern is focused on the medium to long term, particularly around 2030-2035.

Jordan Schneider: Why? How do more nuclear weapons with better delivery capabilities matter strategically?

Junichi Fukuda: If China possesses 1000 or 1500 nuclear weapons, they might be able to deter the United States from intervening in a Taiwan Strait contingency, creating a decoupling and instability paradox. This would be a major concern for us. In the short term, we’re more concerned about China’s increasing conventional capabilities, which are changing local military balances and might enable China’s invasion of Taiwan by 2027. But regarding nuclear issues, our concerns focus on the 2030-2035 timeframe and beyond.

Vipin Narang: China’s expansion is real and occurring faster than we anticipated. It’s designed to challenge the United States’ force sizing principles. While we don’t fully understand their motivations, we must prepare for the possibility that China is developing a nuclear posture and strategy to provide cover for regional aggression and coercion against US interests and allies.

We took this very seriously in the Biden administration. For context, we have these formal documents like the Nuclear Posture Review, which is unclassified and available to everyone. Following this parent document is the Presidential Nuclear Weapons Employment Guidance, which I worked on with Pranay. There’s an unclassified report to Congress known as the “491 Report” that I would recommend reading closely.

When the previous employment guidance was issued during the Trump administration around 2019, China was just emerging as a nuclear concern. There was recognition that China would challenge the US force posture, but the specific nature of that challenge wasn’t yet clear. The current document represents an evolution, with continuity between the Trump and Biden administrations’ guidance. However, it takes seriously, for the first time, the prospect of multiple peer strategic challenges, where China can no longer be treated as a “lesser included case.”

An important point that’s often overlooked is that it’s not just about the numbers. China will reach approximately 1,000 nuclear weapons by 2030 and potentially 1,500 by 2035, which represents quantitative parity with New START treaty levels. What’s more concerning is the composition of their force development.

If China had remained consistent with its long-standing assured retaliation strategy and focused solely on a sea-based force similar to the UK, we wouldn’t need to adjust our nuclear force posture. Instead, they’re developing hundreds of hardened silos in remote areas, perfectly spaced, which serve as nothing other than counterforce targets for the United States. Given our strategy to limit damage to the US homeland and our allies, China knows we must target and hold those forces at risk.

Under New START, we only have 400 ICBMs with single warheads and a limited number of SLBMs (Submarine-Launched Ballistic Missiles) at sea at any given time. Hundreds of Chinese ICBMs in remote regions would consume a large portion of our strategic forces and could make us vulnerable against Russia. We’re observing collusion and coordination between Russia and China. If we’re engaged in conflict with Russia and have expended some nuclear forces to maintain strategic deterrence because of their actions in Ukraine, China might see an opportunity to act because we no longer have sufficient forces to maintain strategic deterrence against them.

Their force structure appears carefully and intelligently designed to undermine our capabilities and sizing principles. We recognized the need to respond in ways that enable us to maintain both strategic and regional deterrence in a multiple-peer world across multiple theaters, especially given the coordination between Russia and China — and we shouldn’t underestimate North Korea either.

We conducted a year-long process to assess adjustments to the US force posture for strategic and regional deterrence in this new reality. As Fukuda-san mentioned, we’re implementing changes because Congress mandated it, but these changes also offer deterrence advantages. The security environment continues to deteriorate, and we can no longer deny the value of having a purpose-built maritime regional deterrence capability that frees up strategic platforms for strategic deterrence.

This challenge demands our attention. I’m concerned about the administration potentially rushing to reach a deal with Russia on arms control to extend New START without considering the implications for China. During the first Trump administration, Marshall Billingsley actually attempted to coordinate arms control strategies. I’ll defer to Pranay on this since he knows more about it than I do. But that approach is crucial — these issues must be coordinated within our own strategy. We need to consider what we’re willing to accept with Russia while keeping China in mind, and vice versa.

Pranay Vaddi: I’ll just highlight one additional point. Beyond everything that Vipin and Fukuda-san have raised regarding China’s changing posture — the warhead buildup, the decision to place nuclear weapons in silo ICBMs (which are vulnerable to nuclear strikes and therefore need to launch quickly in a nuclear war scenario), and their approach to survivability that the US would have to contend with — the Chinese are also modernizing cyber capabilities, counterspace technologies, and conventional precision strike systems. These capabilities could be categorized as counter-intervention measures since the US is also arming itself for a potential conflict in the Indo-Pacific. Alternatively, they could be used aggressively as part of a strategy to coerce under the nuclear umbrella.

The reality is that much of this will depend on one person’s decision-making and whether China decides to maintain or change its historic nuclear doctrine established since Mao’s time — characterized by restraint, minimum deterrence, and a no-first-use policy. If any aspects of this doctrine are changing, we probably wouldn’t be the first to know, as we’ve been unable to sustain diplomatic dialogue with China on strategic stability and nuclear policy issues for the past decade or two.

The Biden administration held just one consultation with Chinese colleagues on this particular topic. As with many other national security issues in the US-China relationship, China’s willingness to discuss nuclear matters depends on the political climate between our countries. If conditions seem favorable for US-China discussions across various topics, they might engage on nuclear issues as well. However, if relations are unfavorable — as they have been for several years due to incidents involving surveillance balloons, Nancy Pelosi’s visit to Taiwan, or tensions during the Trump administration — they won’t discuss these matters with us.

This creates a fundamental problem — how are US policymakers supposed to understand the nuances of China’s policy regarding its modernizing nuclear posture if we can’t meet for discussions? Setting aside any attempts at arms control agreements or diplomatic efforts to limit China’s buildup, progress is impossible unless both sides can engage in dialogue. Despite having one meeting during the Biden administration, all our attempts to encourage, incentivize, or facilitate a second meeting were unsuccessful.

There have been some positive developments. Before leaving office, Presidents Biden and Xi established a joint understanding regarding the use of artificial intelligence in nuclear employment decisions, essentially agreeing that humans should remain in the loop for any nuclear deployment or termination decisions made at the presidential level. Additionally, the two countries now exchange notifications before intercontinental ballistic missile tests. While these are small confidence-building measures, they don’t address the central issue — China is the only nuclear weapons state under the Non-Proliferation Treaty that is quintupling its arsenal.

The United States and China must discuss this issue, or else, in worst-case scenario planning conducted by the Pentagon, we’ll have to assume that China intends to build an arsenal qualitatively equivalent to that of the United States — one designed to undermine US nuclear strategy, as Vipin mentioned. This would force the US to respond with its own buildup, creating a world that benefits no one.

Jordan Schneider: Junichi, you’ve been the most positive and sanguine out of all four of us throughout this conversation. Would you like to conclude on a high note? What are you looking forward to over the next few years?

Junichi Fukuda: Actually, I’m not particularly optimistic. Having nuclear weapons in Japan is politically and economically challenging, with very high barriers. We simply don’t want to consider this option, which is why we’re always concerned about strengthening the credibility of extended deterrence and maintaining our alliance relationship.

The future is unpredictable, especially under the Trump administration, so we must consider various scenarios. However, nuclear weapons are not the most desirable path for Japan. The least desirable option would be to bandwagon with China, which we must avoid at all costs. If we face a truly existential situation where we lose confidence inUS extended deterrence, we might decide to pursue nuclear weapons despite the damage this would cause to the Non-Proliferation Treaty. But this is certainly not our preferred outcome.

Jordan Schneider: Why are you so confident that nuclear weapons would be the preferred path versus an accommodation with China?

Junichi Fukuda: Bandwagoning with China would require compromising Japanese sovereignty. Looking at Japan’s long history with China, there is no precedent for such a decision. Japan has at least 1,500 years of recorded history, and during that time, our country was occupied by a foreign power only once — from 1945 to 1952 by the United States. We have never experienced Chinese occupation of our land.

This represents a fundamental difference between South Korea and Japan. South Korea has a long history of being part of China’s tributary system, but Japan has no such experience. We cannot align with China under any circumstances.

To avoid that scenario, we might need to consider independent nuclearization if we lose confidence in the alliance.

Jordan Schneider: Vipin, Pranay — we can’t end on that note.

Vipin Narang: Let me try to be more optimistic. There’s considerable uncertainty right now. We transitioned from an administration that was very careful with its words to avoid sending mixed signals or giving the impression of internal divisions. For the most part, we spoke with one voice and maintained discipline on these issues.

The Trump administration has a different style. The president speaks frequently, and his statements aren’t always consistent. He doesn’t like nuclear weapons, but if that’s true, he should appreciate extended deterrence — though that connection isn’t always obvious. There’s certainly a battle within the Trump administration over the direction of foreign policy and grand strategy, as Pranay mentioned.

We should take a moment and wait to see which direction this goes. The sky isn’t yet falling on extended deterrence or NATO. The NATO ambassador was very clear about the US commitment to NATO. It would be prudent for allies — and this was true during the Biden administration as well — to contribute more, which they have done. The burden-sharing issue is real, bipartisan, and consistent. This focus isn’t necessarily bad for American security, regardless of which administration is in office.

However, there’s a possibility that things could move in a different direction. It’s important for allies to consider what a world without American leadership in the alliances might look like, while also recognizing that the bipartisan continuity we’ve seen on these issues might prevail in practice. It’s still early — only 90 to 100 days have passed — so I remain in wait-and-see mode. I continue to hope for the best, but it isn’t imprudent to prepare for a very different world.

Subscribe now

Pranay Vaddi: This issue is central to US national security. The more people outside the nuclear space learn about extended nuclear deterrence — its history, our current strategy, what we’ve done and why — from people like Vipin who’ve held important positions at the Pentagon or from allies who have experienced being recipients of extended nuclear deterrence, the better.

Every time I see foreign policy generalists learn about extended nuclear deterrence, a light bulb seems to go on. They understand that this is fundamental to what the United States has built in the world, benefiting the average American. I’m somewhat optimistic that as more people become aware of these issues and critically examine foreign policy decisions that might undermine extended nuclear deterrence, they’ll speak out in support of not just maintaining the status quo but improving our approach.

I want to thank you, Jordan, for having us on. Podcasts, articles, news coverage — all of these help people understand why we approach extended nuclear deterrence the way we do and recognize that having more states with nuclear weapons is inherently detrimental to the United States. I believe people in the current administration instinctively understand this as well. The key will be connecting these dots to help them conclude why specific actions in extended nuclear deterrence are necessary. That’s the source of my optimism.

Jordan Schneider: Let’s close with some reading recommendations. How about two from each of you? Jun, would you like to start with two books?

Vipin Narang: I’d like to recommend Under the Nuclear Shadow by one of my former students, Fiona Cunningham, which just came out.

Jordan Schneider: You can’t recommend that one! We already did a show with her.

Vipin Narang: I highly recommend it. My second recommendation is the updated version of Managing Nuclear Operations. This is quite technical. While Fiona’s book is high-level and accessible for anyone interested in China, Managing Nuclear Operations is on the detailed end of the spectrum. Charlie Glaser, Austin Long, and Brian Radzinsky re-edited this classic with updated chapters from experts like Franklin Miller and James Miller. It provides an excellent introduction to how US nuclear strategy and policy are developed. If you’re interested in that topic, I highly recommend it — it’s a very good update to the classic 1987 volume edited by Ash Carter, Steinbrenner, and a third co-editor whose name escapes me.

Jordan Schneider: Are we telling people not to read Annie Jacobson’s Nuclear War?

Pranay Vaddi: I haven’t read it, so I can’t speak to whether it should be recommended or not.

Jordan Schneider: You guys are so diplomatic. I’ll say it directly — don’t read it. It reads like a journalist who doesn’t really know what they’re talking about. I caught one or two historical errors just from my limited knowledge. It gave me the impression of someone with too much of an agenda to be careful with their sources.

Vipin Narang: I appreciate the original sources. I recently taught my class on extended deterrence and discovered an obscure article by James Schlesinger from 1962 explaining why we extend deterrence to Western Europe and NATO. Remarkably, the arguments haven’t changed between 1962 and 2025. It’s a RAND chapter that’s quite difficult to find. I was stunned when I read it carefully, realizing how much we’re simply reinventing the wheel. [Link here!]

I’m very humble about the fact that scholars like Schelling, Brody, Schlesinger, Earl Ravenhall — all published in International Security — along with Ash Carter, Frank Miller, and others who preceded us, had already thought through all these issues. We’ve forgotten many of their insights because we had a 35-year interregnum when we didn’t have to consider the importance of extended deterrence and nuclear deterrence more broadly. For the audience, I recommend going back to the original texts — you’ll be surprised how little has changed.

Jordan Schneider: Pranay, Junichi — any recommendations come to mind?

Pranay Vaddi: I have two recommendations. First is a report by the American Academy of Arts and Sciences authored by Tong Zhao and Dmitry Stefanovich. They write from Chinese and Russian perspectives about why those countries are concerned about US missile defense and its implications for strategic relationships. For those of us who focus on arms control policy and try to understand how the administration might approach hypothetical denuclearization talks with Russia and China, this is essential reading. Both authors are colleagues who have studied this area extensively. It’s not a book, so you can gain valuable insights from two genuine experts in just 50-60 pages.

The second recommendation is an older book from the late 1980s by Richard Betts called Nuclear Blackmail and Nuclear Balance. I found it particularly interesting as the Ukraine conflict progressed while I was still at the NSC. It reminds nuclear policy experts that nuclear weapons aren’t magical solutions. Attempts to use them in world crises — either to force adversaries to back down or to coerce for other military objectives, as in the Vietnam War — have produced mixed results at best. Understanding this history is important for our current era, where we’ve seen Russia issue almost daily nuclear threats in the context of a conventional war it initiated. Those are my two recommendations.

Junichi Fukuda: I’ve chosen two books. The first is another classic, Glenn Snyder’s Deterrence and Defense: Toward a Theory of National Security. This book is quite important because it addresses the problem of the stability-instability paradox, which directly applies to the current situation in the Pacific. If you want to study the stability-instability paradox, I recommend reading Graham Allison’s work on deterrence and defense.

The second recommendation is Brad Roberts’ The Case for US Nuclear Weapons in the 21st Century from 2015. This book is widely read in Japan as it directly discusses options for strengthening extended deterrence for allies in the Indo-Pacific. Many Japanese experts interested in this subject have read this book, so if you read it, you’ll have common ground for discussions with those experts in Japan. I highly recommend it.

Thanks again to the US-Japan Foundation for sponsoring this episode. It’s the first in a series about Japan-adjacent topics. I promise they won’t all be this dark.

Mood music:

为全球华人游荡者提供解决方案的平台：游荡者（www.youdangzhe.com）
这世界的辽阔和美好，游荡者知道。使用过程中遇到问题，欢迎联系客服邮箱wanderservice2024@outlook.com.

【和放学以后永不失联】订阅放学以后Newsletter，每周三收到我们发出的信号：afterschool2021.substack.com 点击链接输入自己的邮箱即可（订阅后如果收不到注意查看垃圾邮箱）。如需查看往期内容，打开任一期你收到的邮件，选择右上角open online，就可以回溯放学以后之前发的所有邮件，或谷歌搜索afterschool2021substack查看。

截至目前，放学以后Newsletter专题系列如下：“在世界游荡的女性”系列、“女性解放指南”系列、“女性浪漫，往复信笺”系列、莫不谷游荡口袋书《做一个蓄意的游荡者》系列、“莫胡说”系列”《创作者手册：从播客开始说起》,播客系列和日常更新等。

大家好，本期Newsletter由刚刚结束西欧十日游荡之旅，回到西班牙的霸王花木兰轮值。原本今天是由金钟罩轮值，但他因为想要放松休息下，所以这段时间先暂停Newsletter的写作，由我和莫不谷继续轮值，等他休息好了以后，再来继续轮值。

也和大家提前预告一下，这两天金钟罩刚刚剪辑完放学以后第54期播客《我不要只做世界的承受者，我要对这个世界一顿发起！》初稿，预计很快就可以在5月上线，敬请期待！

今天这篇Newsletter我想和大家分享刚刚结束的西欧游荡之旅的思考和感受，同时还在游荡者平台（www.youdangzhe.com）分享了这次游荡的具体攻略《生命活力复活之旅:四月，到西欧去！荷兰、比利时和德国攻略》，方便需要的朋友们参考！

不知道是不是因为这次游荡之旅刚刚结束，新鲜感还没过去，这次的西欧游荡想要思考和表达的内容有很多，一是每天暴走三万步大晚上回去睡前还在选图发小红书“游荡者的日常”，莫不谷还笑说我这是有什么KPI要完成吗，二是每天咔咔记录游荡荷兰、比利时和德国的各种攻略，所以回来西班牙休整结束就立刻在游荡者平台更新了一篇攻略，后续会陆续再更新出来，其实前两者对我来说都不太需要调动很多情绪和感知力，而这次游荡之旅我发现，自己的感知力在慢慢恢复，或者换句话来说，我有更多勇气和活力面对自我了，我基于自我感受想要记录、表达的时刻多了很多。

这次游荡的主题是生命活力复苏之旅，莫不谷还给我们一路拍摄的视频素材想了名称“四月生命活力复活之旅”，副标题是“到西欧去”。为什么是到西欧？刚回来西班牙的我立刻感受到了西欧和南欧风景的差别。南欧大多是地中海气候，春季温暖但干燥，生长的也是耐旱的植物，而西欧如荷兰、比利时、德国西部，则是温带海洋性气候，春天温暖湿润，所以目之所及草地茂密、绿树繁盛，特别是春天刚长出的嫩叶，深浅不一的绿，让人无法抗拒地沉迷。

还记得十天前的早上8点，飞机从温暖的西班牙落地有些清冷的荷兰埃因霍温机场，和煦明亮的阳光洒在机舱里，让人不由得心情大好。从埃因霍温机场坐上公交，透过宽敞干净的玻璃就能看到一路的风景：看到野鸭子和像乌鸦一样的小鸟在嫩绿的草地里低头吃草，它们像是充分知道自己不必担心、害怕那样安心平静；看到路边干净的房屋和低矮的窗户，没有高楼大厦的压抑，房主不仅不必担心窗户低矮的风险，而且还要透过窗户展示审美，像是充分在展现我们没有什么可害怕和隐藏的；看到稀疏的行人跑步、骑自行车、遛狗，像是在看一部安静的纪录片；再仔细探头看一看青绿色的草地里，还有郁金香一样淡黄色的小花，错落有致。人有机会看着这一幕幕场景，饶是再冰冷坚硬的心也会忍不住动容。这是我第三次来荷兰，但却像是第一次来荷兰一样，止不住和莫不谷感叹，荷兰真是发达的资本主义国家（特别是对比西班牙），荷兰的春天真是太美好了！大概是因为我已经居住在西班牙，终于有了欧洲在地者的视角，就更能充分欣赏荷兰的美丽，而荷兰的美不仅在于春天的风景，还有越深入了解，你就越会爱上的这个把人当人，把生命当生命，崇尚自由、法治和理性的社会文化和理念。

记得在荷兰我们每日暴走三万步，回到家常常筋疲力尽，但到了大半夜我不仅不困，甚至兴奋地睡不着，得靠阅读《空洞的心》帮助入睡。莫不谷惊讶地问我说：“你人生有如此充沛地活过吗？”这是我西欧游荡之旅很明显的一个变化：“我不困了！”这次和莫不谷还有荷兰的台湾朋友一起去森林野餐，台湾朋友还记得我2023年野餐睡着的事情，但是这次即便我躺在野餐垫上，我想睡我也睡不着，好像身体在和我说，天气这么好，有啥好睡的？去年五月我和莫不谷在纽约游荡的时候，莫不谷每天high到不行，我直接在博物馆里坐着睡着，这世界对莫不谷来说是个知识的游乐场，对我却是随时随地睡觉的一张床。那时候莫不谷常常被我随时入睡震惊到，我则是回答说，我从小就这样，这没什么奇怪的。

但我现在回看想改口说，随时随地睡觉真的不正常。就像我随时掉电的苹果手机，电池绝对有问题。那时候我还没意识到，我的大脑一直处于超级省电模式，已经无力支撑一个人正常一天的活动。2023年5月以来到现在，快两年的时间过去，我辞职休息刚满一整年，去世界游荡无数次，我的身体终于开始充上了电。这让我想到看潜水视频时知道的一个知识，深度下潜的潜水员在浮出水面前必须先减压，否则会得减压病。如果潜水员下潜至 100 米，每多呆 1 分钟，需要减压的时间就要增加1小时。当我们持续压抑剥削自我以适应学习、工作和威权社会时，每多一分钟，我们恢复正常的时间都相应需要增加，而这个恢复的时长甚至是以指数级在增长。不是离职了人就会健康，不是离开了人就会自由，身体、大脑、情绪的恢复取决于此前受伤的时间和程度。好在我放弃继续伤害自己，好在我决定拯救自己，好在我的身体在慢慢恢复中了。

（鉴于以下内容涉及个人信息，安全起见设置付费门槛）

Read more

Lennart Heim is the Compute Team Lead at RAND’s Technology and Security Policy Center and a Professor of Policy Analysis at RAND School of Public Policy.

China will likely match U.S. AI model capabilities this year, triggering inevitable concerns about America’s technological edge. However, this snapshot comparison misses the bigger picture. While Chinese models close the gap on benchmarks, the U.S. maintains an advantage in total compute capacity — owning far more, and more advanced, AI chips. This compute advantage, if leveraged strategically, will play an extraordinary role in driving economic transformation, securing technological leadership, and shaping the global AI ecosystem. U.S. policymakers risk squandering this edge by focusing on the wrong metrics and overreacting to predictable Chinese advancements.

TSMC's Secret Pipeline to China

Central to the U.S.-China AI competition are U.S. export controls that restrict China from importing advanced AI chips, acquiring semiconductor manufacturing equipment to build indigenous advanced AI chips, and using leading chip manufacturers such as Taiwan Semiconductor Manufacturing Company (TSMC). Despite these measures, a massive failure occurred in September 2024: TSMC, lacking basic due diligence, breached export restrictions by producing advanced AI chips for Huawei through a Chinese proxy company. This violation allowed Huawei to secure approximately 3 million chip dies using TSMC’s 7nm process, enabling the production of China’s best AI chips, the Huawei Ascend 910B and the upcoming 910C. This dwarfs even the highest estimates of smuggled chips, which typically involved tens of thousands of units, not millions. Although these chips trail the U.S. state of the art by about four years, they collectively provide China with computing power equivalent to approximately 1 million export-controlled Nvidia H100s (Nvidia's previous-generation chip from 2023) — substantial AI compute capacity that compensates for China's lack of indigenous production capabilities.

A Pattern of Failures, Not an Exception

The U.S. government responded to TSMC’s illicit production issue with a harsher rule in January 2025 and a likely investigation, but this incident is just one of many failures.

The 2022 chip controls contained specification errors that allowed Nvidia to slightly modify existing chips, creating the A800 and H800, which enabled DeepSeek's rise. Entity-listed Chinese companies built literal bridges to access advanced chipmaking equipment in unlisted facilities. Chinese firms stockpiled enough high-bandwidth memory (HBM) to meet their needs for the next few years, after the industry leaked the December 2024 restrictions in July. Until recently, AI chips with deployment-optimized performance, such as Nvidia's H20, continued to flow to China despite evidence that they enable new reasoning capabilities, with U.S. government officials taking months to act despite warnings. In every case: too little, too late.

What to Expect from Export Controls—And What Not to Expect

Despite these implementation failures, export controls remain a valuable strategic tool. While export controls do not create absolute barriers, they impose costs that have substantially slowed China's progress in AI and semiconductors, thereby maintaining America's lead.

But why is China still producing competitive models? It's harder for export controls to affect individual training runs than an entire ecosystem. Think of AI compute like factory equipment: Having fewer production lines doesn't prevent you from manufacturing a single product—you might still create that flagship product—but it severely limits your production capacity and market reach. Without the latest machinery, your production costs are higher and efficiency lower while competitors optimize their operations. What's more, the economics of scale work against you: while your competitor with ten factories can spread fixed costs across massive production volumes, driving down unit costs and enabling experimentation with new product lines, you're stuck with higher per-unit costs and limited ability to diversify. Similarly, when China has less total compute, they can still develop competitive models. After all, even in the U.S., companies only spent a fraction of their compute on training frontier models. But China misses out on the economies of scale that allow the U.S. to deploy AI broadly, experiment with more approaches, support a diverse ecosystem of leading AI companies, and continuously reinvest efficiency gains across their entire AI economy. Critically, this compute disadvantage also limits China's ability to project soft power globally—whether by providing AI services to international markets or by having the capacity to export advanced AI chips to other nations.

However, if current trends in AI training continue—shifting from two chips in 2012 to clusters of hundreds of thousands in 2025—export controls may eventually bite harder. China would need to build significantly larger clusters to compensate for less powerful chips: a cluster of 100,000 Nvidia B200s (a leading U.S. AI chip) might require a Chinese equivalent of 300,000 Ascend 910Cs (China's leading chip), resulting in higher energy consumption and greater engineering complexity for distributing AI workloads. Compounding this hardware gap is Nvidia's sophisticated software and networking ecosystem—crucial for orchestrating massive AI chip clusters and currently unmatched by Chinese alternatives.

10x More AI Workers: The Power of U.S. Compute Dominance

This focus on preventing model parity misses the fundamental question: What are the true metrics of AI leadership? We reacted to DeepSeek because it narrowed the gap between U.S. and Chinese model capabilities as measured by benchmark leaderboards and model rankings. However, the fixation misses the broader technological competition.

For example, think of AI systems as “virtual employees” that can perform cognitive labor, replacing remote colleagues you’ve never met in person. The number of AI employees an economy can deploy hinges directly on its compute resources. With roughly 10 times more compute capacity than China, the U.S. can field proportionally more AI employees across its economy, in sectors as diverse as drug discovery, logistics optimization, industrial robotics, and AI research itself. This compute advantage therefore compounds into a broader economic advantage over time. As AI systems drive productivity growth and innovation, they strengthen the underlying factors of national power. While China may catch up in individual model quality for now, the true metric of AI leadership might lie in deploying and integrating these systems at scale. In that sense, America’s true moat isn’t just better models—it’s the capacity to deploy and integrate AI in the economy at scale.

Subscribe now

Maintaining and Leveraging America's Compute Advantage

Both the Trump and Biden administrations deserve credit for recognizing early on that compute would be central to AI competition, taking key steps like restricting extreme ultraviolet (EUV) equipment exports to China in 2018 and banning AI chips in 2022—before ChatGPT captured public attention. But to stay ahead, the U.S. government must now build capacity to act quickly when reinforcing export controls while developing a clearer understanding of their actual impact. Rather than hoping to prevent temporary model parity, the U.S. strategy should leverage its substantial compute advantage to transform the broader economy and establish leadership in AI deployment worldwide. China may achieve competitive individual AI models this year, but this narrow benchmark gap is neither permanent nor strategically decisive. China will continue to tout breakthroughs to rattle U.S. confidence, but getting distracted and abandoning America's fundamental compute advantage would be a profound mistake.

Mood Music

How should America maximize its odds when competing with China? Can Trump’s approach to alliances succeed in strengthening deterrence? And what will it take for China to diplomatically capitalize?

To find out, ChinaTalk interviewed Rush Doshi, author of The Long Game: China’s Grand Strategy to Displace American Order. Rush served as deputy senior director for China and Taiwan on Biden’s NSC and is now at CFR. His new article with Kurt Campbell, entitled “Underestimating China: Why America Needs a New Strategy of Allied Scale to Offset Beijing’s Enduring Advantages,” I sincerely hope becomes a seminal document for American grand strategy in the 21st century. This is the most important show I’ve recorded all year.

Our conversation covers…

• Strategies for countering China’s strengths while accurately assessing weaknesses like demographics, debt, and slow growth,

• Historical lessons for US-China competition, from the newly-industrialized UK to Gorbachev’s USSR,

• Capacity-centric statecraft as an underrated type of international partnership,

• Persuasive versus coercive approaches to alliances,

• The evolution of China’s grand strategy and how Beijing’s diplomatic overconfidence in the face of tariffs could backfire.

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

America’s Odds

Jordan Schneider: Let’s start with net assessment, that is, understanding where the US is relative to China. Why is this important for effective strategy?

Rush Doshi: You have to compare yourself relative to your adversary or competitor in order to understand exactly what your strategy should be. Without good net assessment, you can’t really form a strategy. There used to be something called the DoD Office of Net Assessment. It’s since been eliminated.

Our assessment of China has swung widely over the last few years. Back in 2020, there was a perception that the US was in decline. We’d had January 6th, we weren’t handling the pandemic particularly well, and we had alienated our allies and partners. At the same time, China’s economy was booming. It seemed to be handling Covid well. In that moment, there was a perception that America was on its way out.

Just two years later, however, the zeitgeist changed completely. Russia invaded Ukraine in 2022 but achieved only limited success. The removal of zero Covid in China didn’t bring huge prosperity or reignite the economy. Later that year, America debuted ChatGPT to the world. It seemed like we had a rabbit we could pull out of our hat again — we could always rely on American innovation to sustain our lead.

These factors created a new zeitgeist — America ascendant. “America’s got this, and China can’t possibly keep up.”

But it’s difficult to strategize around such vastly differing conclusions about where we are relative to our competitor.

In our article, Kurt and I try to present a nuanced evaluation of our position relative to China.

Jordan Schneider: This is a fascinating case study of the past four years because it almost felt like these swings occurred faster than during the Cold War. Back then, they took longer and the triggering events were bigger than COVID policy. Events like the Vietnam War — America losing a conflict — would cause you to reassess relative trajectory, or decades-long realizations that the Soviet system was not delivering the technological and economic growth that people thought it might have been in the 30s, 40s, and 50s.

The big lesson is that it’s not necessarily a flows question — it is a stocks question. The US and China each have roughly a quarter of the world’s GDP and are likely going to continue having roughly a quarter of the world’s GDP each in the decades to come.

Rush Doshi: It’s important to note that two things can be true simultaneously. China could be slowing economically and have serious economic problems, but it’s also possible that it has strategic advantages and technological advantages in the timeframe that matters most for geopolitical competition.

Poor assessment simply concludes, “They are facing tough economic times, therefore, strategically they’re going to be on their way out.” It never ends up being that neat. The fundamental point of our piece is that you have to think about how particular advantages and disadvantages in one sector translate strategically to great power competition. This isn’t intuitive. It requires some degree of guesswork, but also careful analysis to understand correctly.

Jordan Schneider: There are cones of expectation for how well the US will do and how well China will do from a national power perspective. If you base your assessment on the third standard deviation performance of one country over another, you’re probably not picking the right strategy for the vast majority of futures. The way the US and China are discussed often defaults to either “this country has it figured out” while “that country is doomed,” or vice versa.

Rush Doshi: Thinking about it probabilistically is exactly the right approach.

Jordan Schneider: What do you think about the psychology of the American public and our leaders in Washington? Do you predict a preference for one direction or another, or is there appetite for a nuanced middle stance?

Rush Doshi: There’s a fundamental sense that things are changing — that China is formidable but also has a number of fragilities. The stakes of getting this question right are very high. This leads people to promote narratives they want to believe.

Sometimes there’s a market for saying, “Everything’s fine, don’t worry. American advantages are enduring, stable, real, and decisive.” On the other side, people might suggest things are in tough shape because they have a particular domestic policy agenda to advance. We saw that with the trade war. These assessments have political valence and salience.

Currently, China isn’t even a top priority for most American strategic thinkers. The focus continues to be on Russia-Ukraine or the Middle East. China lurks in the background rather than occupying the foreground where it deserves to be. That’s partly why there’s not always much attention on getting these assessments right.

It was important to have an office that did this kind of work in the US Government, and it would be monumental if that function was restored in the Trump administration. Hopefully, they make that decision soon.

From a government perspective, you almost never have anybody doing true net assessment, because the intelligence community isn’t authorized to consider American strengths and weaknesses relative to an adversary. They can talk about “red” (the adversary), but they won’t necessarily talk about “blue” (us). This causes complications.

Jordan Schneider: I agree that net assessment is important. It’s striking that so much of the commentary about the US relative to China comes from people who aren’t necessarily China experts. It seems that there’s more appetite to use China as a rhetorical talking point, on either the left or right, rather than to make accurate assessments about China.

Rush, you point to scale as the key metric for evaluating the US and China on a decadal or multi-decadal horizon. Can you explain your US versus UK analogy?

Rush Doshi: We only had 5,000 words for our piece, but a full net assessment should be much longer. Not all large countries achieve great power status. There’s a distinction between size and scale. Scale is the ability to translate size into meaningful outcomes. Scaling up is that process.

The Asian tigers had an incredible ability to scale on a small foundation. When you take those techniques and apply them to a foundation as large as China’s, the consequences are world-shaking.

The UK found this out the hard way. The UK had a first-mover advantage in the Industrial Revolution, and a small island in the northwest corner of Europe dominated the entire world for a long time — which geographically is quite remarkable. But that first-mover advantage wasn’t permanent. British industrial methods eventually transferred to Germany, the US, and Russia.

The British knew this was happening. Lord John Seeley wrote a book in 1883 about the rise of Great Britain that ended with a warning. He worried that just as Florence was surpassed by the great nation states of Europe, Great Britain would be surpassed by larger countries — the US, Germany (which had a much larger population than the UK), and Russia. He believed the British methods for industrializing and creating wealth could be transferred to others. When that occurred, they would have what he called “scale” — unimaginable scale relative to Great Britain — and Great Britain would decline.

During that period, these countries leveraged scale to push Britain out of key markets. They used their larger domestic economies to drive down marginal costs and outcompete in third-country markets. Today, that dynamic resembles the US and China.

As the US grew in scale and became more efficient and competitive, our industrial base exploded. By 1910, before World War I even happened, we were manufacturing four or five times as much steel as Great Britain. By World War II, our size and scale were even greater.

Hitler warned that the US was an unimaginable productive power. Yamamoto thought Japan could succeed in the first six months of war, but eventually American scale would defeat them. Even the Italians feared a contest of stamina would favor the United States.

That fundamental scale advantage we had shaped the thinking of our adversaries in the run-up to World War II and during the conflict itself. It wasn’t a hidden fact of politics — it was well understood by our rivals. Now that sense of daunting scale probably belongs to China.

Jordan Schneider: You make the point that even when Nazi Germany was at its peak from 1938 through 1941-42, the relative industrial weight of the Axis powers doesn’t hold a candle to what China can bring relative to the US today.

Rush Doshi: Hitler called the US a “giant state with unimaginable productive capacities.” In many ways, he was hoping to create that kind of productive scale for Germany, which obviously led to disastrous consequences. The fact of American manufacturing prowess was well established and understood around the world.

Jordan Schneider: For a couple of decades in the middle of the 20th century, America was building everything for everyone. Then globalization arrived, which was beneficial as more people gained wealth. We prevailed over Great Britain, Nazi Germany, and Japan, and later over the Soviet Union. But we’re in a very different industrial race today. Rush, let’s look at China’s statistics.

Rush Doshi: Every great power competitor the US has faced previously lacked the size and scale that China possesses. The Soviet economy was much smaller than the US economy, less productive, and its absolute manufacturing capability was inferior to America’s. The US was stronger than Germany and Japan combined in World War II. However, China represents the first competitor with true size and scale advantages against the United States.

Consider manufacturing shares. About 25 years ago, the US share might have been 30% while China’s was 6%. Within two decades, China’s share has quintupled to nearly 32%, while the US share has fallen by half to 15%. According to the UN, by 2030, China’s manufacturing share will be four times that of the United States — about 40% to our 11%. This represents a surprising turnaround in just 30 years.

The last comparable shift occurred between the US and UK. From 1870 to 1910, the British share of global manufacturing fell by 50% — the same proportion we’ve seen fall for the US, except it took us only 20 years rather than 40.

China’s current output is remarkable: twice American power generation, three times American car production, 13 times American steel production, 20 times American cement production, and approximately 200 times US shipbuilding capacity overall (though only three times our capacity for warships).

China produces about half the world’s chemicals and ships, 67-70% of the world’s electric vehicles, more than three-quarters of the world’s batteries, 80% of the world’s consumer drones, 90% of the world’s solar panels, and 90% of the world’s refined rare earths.

They’re also betting on the next industrial revolution. The US is installing many industrial robots, but China is installing seven times more. Half of all robot installations in 2023 happened in China. Regarding nuclear power for fueling the AI revolution — China leads in commercializing fourth-generation nuclear technology that we invented, and it’s planning 100 nuclear reactors in the next 20 years.

In science and technology, people claim China can’t innovate, but they exceed us in active patents and top-cited publications. You can question these numbers and argue they manipulate statistics. There’s some truth to statistical manipulation, but the trend line is incredible. They’re probably about even with us in those categories, and in 10 years, they’ll be well ahead of us, even accounting for statistical manipulation.

This manufacturing power translates into two forms of advantage. One is military advantage, as we discussed regarding World War II. The second is technological advantage — innovation from the factory floor, tacit knowledge, process knowledge, and special production capabilities that improve over time. This creates enduring advantages for China.

We should recognize this because it mirrors American innovation. When we were becoming a manufacturing power, we lagged behind Europe in science and technology. We didn’t have the Nobel Prize numbers that we have today. Manufacturing was the leading edge, and scientific recognition came later. We’re seeing the same story with China.

Jordan Schneider: Let’s not underestimate China — I’m with you this far, Rush. What’s the answer? One approach from the administration is that we need to reindustrialize. The problem is that even with perfect policies, you might gain only 0.5-2% per year in global manufacturing production. Setting reindustrialization and manufacturing growth as long-term goals aside, the trends you’ve outlined aren’t going to reverse anytime soon.

Rush Doshi: These built advantages are very sticky. Poured concrete is poured concrete — it exists. China’s supply chains won’t immediately relocate. They’re also more resilient than people think.

Some argue that China has macroeconomic and demographic problems, questioning why manufacturing statistics matter. But these problems are probably overstated. Consider their economy — it’s smaller than ours in nominal dollar terms, and many take comfort that it’s shrinking relative to the US economy. However, much of that is due to a strong dollar.

If you remove that financial advantage and look at purchasing power parity, China’s economy probably surpassed the US economy 10 years ago and is 30% larger today. Adjusting for purchasing power has limitations, but it’s a good attempt at capturing the local price of key strategic inputs — infrastructure, weapons, government personnel — which are priced in local terms. While nominal GDP might be important for quality of life, purchasing power adjusted GDP better captures factors that generate strategic advantage.

Regarding demographics, China is aging, but when will that matter? By 2100, their population will have fallen by half based on current trends. However, the strategic harm of aging won’t materialize for another 20 years or so. Between 2010 and 2020, the under-14 share of China’s population increased in absolute terms and as a percentage of the total population. The population is still aging, but the younger share grew due to a Mao-era baby boom — these are the grandchildren. Demographics are lumpy, which can buy time.

China’s dependency ratio won’t be as problematic as Japan’s current ratio until 2050, giving them time. Their investments in industrial robotics and embodied AI could help address labor problems from a declining population. Chinese factories like BYD plants already operate with few assembly line workers.

Regarding debt, China’s level is serious — it’s 300% of GDP when combining government, household, and corporate sectors. But that figure matches the United States. The composition of debt matters, as does who holds it, but this aggregate statistic provides a sense of overall indebtedness. If comparable, China might weather this challenge by properly recapitalizing local governments.

Finally, people note that American companies have high market capitalizations — the most valuable companies on earth are American with the highest profit shares in technology. But that market cap partially reflects investors wanting dollar assets and a strong dollar. Profits matter, but American companies maximize profits while Chinese companies maximize market share. They’re willing to operate at a loss to achieve market dominance. They’re optimizing for something different that might matter strategically — if they can remain solvent longer than our companies, they can put them out of business and deindustrialize us.

China has weaknesses — aging population, economic slowdown, potential deflation, high youth unemployment. All are important, but they may not matter as much in the timeframes that determine great power competition. That’s what we’re addressing with the question of scale.

Jordan Schneider: You’ve convinced me about the unlikelihood of second or third standard deviation downside outcomes for China. You’ve addressed many bearish arguments about China’s 20-year outlook, but not political instability. Can you discuss that? Xi will eventually die.

Rush Doshi: This is a critical point. My assessments focused on macro indicators, demographics, and financial factors because those are commonly cited when people claim China might never catch up to the US. Political leadership is challenging to evaluate.

If there’s a path to Chinese failure — an inability to succeed in the 21st century — it runs through a failed political transition from President Xi to his successor. Interestingly, Deng Xiaoping set up 20-plus years of succession by selecting Jiang Zemin and Hu Jintao, but he didn’t pick Xi. We lack that same long-term planning now.

China is living in the house that Deng Xiaoping built. Xi Jinping is changing the architecture, and we don’t know if these changes will cause the house to collapse or strengthen it against external pressures — to use terminology Xi often applies about strengthening the house against the elements.

China’s ability to maintain political stability is extremely difficult to assess. The Chinese Communist Party has endured the Great Leap Forward, Cultural Revolution, Tiananmen Square massacre, and globalization, remaining powerful throughout. However, elite instability at the top can threaten to bring down the entire party, as seen at various points in China’s history — such as when Deng Xiaoping was repeatedly imprisoned during the troubled transition from Mao to his successor.

We don’t know if Xi Jinping will select a successor or be too nervous to do so, given that he’s alienated vested interests, powerful families, and parochial groups that historically exercised power more freely than they do under his leadership.

This represents one potential path to Chinese problems. However, you can’t count on systemic failure or institutional collapse, especially when China has studied the Soviet collapse intently and tried to avoid those same mistakes. We must give them credit for considering paths to decay and attempting to close off as many as possible. They won’t get everything right, but that effort can’t be ignored.

Jordan Schneider: Even if you give Kyle Bass as much credit as possible — say a 50% chance of being right — that still leaves a 50% chance that you have an incredibly legitimate superpower breathing down your neck. It makes sense to buy the insurance plan to implement all the policies needed to live in a world where China actually succeeds.

Rush Doshi: That’s a great way of putting it. In our piece, we discuss “allied scale” as the alternative grand strategy for the United States. You’re right, it is an insurance plan. You can hope that China somehow gets in its own way — which wouldn’t be great for 1.4 billion Chinese people, but geopolitically, you could hope for that. However, betting on China’s collapse, governmental system change, or a botched leadership transition isn’t sensible. You need to buy insurance by investing in an American path to scale. Few great powers have achieved scale when confronted with a bigger rival, but the US can do it through its allies and partners.

Beyond Traditional Alliances

Jordan Schneider: Let’s talk about capacity-centric statecraft. What must Washington do?

Rush Doshi: Working more closely with allies sounds like a cliché. Kurt and I argue that we need to revolutionize how we work with allies by putting capacity building at the center — flowing both from the US to its allies and from allies to the US. If done right, the benefits are substantial.

When quantifying allied scale, the US with its allies represents approximately three times China’s nominal GDP, twice China’s purchasing power adjusted GDP, and more than twice China’s defense spending, even using high Chinese estimates that aren’t publicly available. We would have 1.5 to 2 times China’s share of manufacturing, and we would dominate in patents and top-cited publications.

China may be the top trading partner for 120-140 countries worldwide, but an allied consortium would be the top trading partner for virtually everyone except perhaps North Korea. This represents an incredible scale advantage that’s readily available.

The central task of American foreign policy in this era is transforming that theoretical advantage into reality — realizing that scale. Too often, we think of allies in hierarchical terms based on post-World War II and Cold War habits. We focus excessively on military issues and categorize alliances as tripwires, protectorates, vassals, or status markers. Our contention is that alliances must function as platforms for building capacity.

We could envision Japanese and Korean investment in American shipbuilding, where they have three to five times more productivity per worker. We could provide more technology to our allies; AUKUS exemplifies this approach. Kurt and I worked closely on launching AUKUS, which involves the US and UK helping Australia acquire nuclear submarine capability — our most sensitive military technology — to build meaningful Australian capacity in the Indo-Pacific.

We should consider novel joint military formations. The British and French created a joint brigade; the US and Japan or the US and South Korea could create anti-ship cruise missile battalions to train each other on the best shoot-and-scoot tactics, learning best practices from the Marine Corps under General Berger.

Allied scale also means allies helping each other, with the US facilitating collective action and connections. South Korea could help Europe rearm with its incredible defense industrial base. Scandinavian countries with excellent anti-ship cruise missiles could sell them to Southeast Asia. France, with expertise in LEU nuclear-powered submarines, could assist India, which has the same nuclear submarine program.

This vision of allied scale doesn’t always place the US at the center, but the US provides gravitational pull to facilitate collective action. And that’s just the security dimension — there’s much more to discuss on the economic and technological fronts.

Jordan Schneider: Let’s stay on the security side for a while. You were in government for the past four years trying to implement these ideas, which didn’t emerge from nowhere. The fact that the US plus its friends will be 2-3 times larger than China and its friends over the next few decades — assuming the balance of alliances remains stable — isn’t a new insight. As you and Kurt pushed for this approach with a president who was extraordinarily ally-focused, what roadblocks did you encounter from both US and partner perspectives?

Rush Doshi: Let me first address whether allied scale is intuitive. It is intuitive — we collectively know that if we don’t hang together, we’ll hang separately. What’s been missing from most analyses is the sense of urgency. People say working with allies is good — like apple pie and motherhood — but don’t explain why it’s essential.

It’s essential because we’re underestimating China’s scale. Great Britain failed as a great power because it couldn’t achieve scale. Lord Seely looked to unite with British colonies to create scale that could rival other great powers of that age, but those efforts were too little, too late, and everyone drifted apart. The question for the US is whether we’ll face the same fate.

We have an advantage: instead of an empire, we have allies who are independently capable. That’s a huge advantage compared to the UK. What Kurt and I are arguing in our piece is that this is urgent — we must get this right or cede the century to China. This isn’t the usual framing around allies. People typically see allies as beneficial, not as a four-alarm fire necessity. Part of the challenge in implementation is that people don’t perceive it as urgent, but rather as conventional common sense.

Jordan Schneider: To be clear, the US and China independently will likely run neck and neck — or too close for comfort — in the vast majority of future scenarios over the next few decades. Therefore, America’s most important strategy is to leverage the rest of the world, which is much more likely to partner with us than with a CCP-controlled PRC. That’s how you change the balance of power — the most straightforward way to dramatically shift the balance over the coming decades.

Rush Doshi: I’ll add that we may not be neck and neck in all metrics that matter for strategic value. We’re not neck and neck in manufacturing — we’ll be one-quarter their size. Comparing the US versus China on many metrics, we’re behind. We don’t have scale. We have several advantages: capital markets, immigration, innovation, talent, and general political stability. But we lack scale.

My concern is that without allies, the US won’t run neck and neck with China. The only way we can compete on metrics for great power competition is with allies. That’s why this is urgent. Without them, we’ll be like the UK was to the US and Russia, or like Florence was to the UK — one-quarter their size.

As productivity equalizes around the world — partly due to institutions and technology — China will remain productive. They already outmanufacture us and lead in many industries. The US urgently needs its own path to scale, which comes through allies.

The British couldn’t achieve scale because they failed to unite with their colonies. What differentiates the US is that we have capable, independent, sovereign allies who share our values — not colonies or subjects. If we join together with them, it’s not even a contest with China. We completely outscale them.

Jordan Schneider: This is a really important point. Modern-day Japan and Germany are not comparable to India circa 1890. We’re talking about very different allies in terms of their latent ability to contribute to the metropole’s national power.

Rush Doshi: Even in Lord Seely’s case, he excluded India, considering it a resource pit — obviously wrong about India’s potential. He focused on Canada, South Africa, Australia, and New Zealand — the white settler colonies. But compare Japan or Korea to any of them. The American alliance system contains incredible performers — powerful, technologically advanced, economically impressive nations who also make excellent military equipment and broadly share our values. That’s an extraordinary advantage in world politics. Almost no one else has this.

Don’t take my word for it — take China’s. China believes our single greatest asymmetric advantage is our allies. In 2017-2018, their delight at rising populism stemmed from one thing: the sense that populism was breaking American alliances apart. They’ve always seen alliances as our critical advantage, and they understand what we often don’t — that alliances bring economic and technological power to bear in competition.

The Biden administration had a theory for achieving scale with good and bad elements. The Trump administration also has a theory with strengths and weaknesses. Our piece isn’t partisan — we’re not advocating for one approach. What we’re saying is that the destination must be scale. There can be a path to allied scale with “Trumpian characteristics” — great, let’s pursue that. My concern is unilateralism that resolves bilateral issues without considering the larger strategic context vis-à-vis China.

The Trump administration has an opportunity to secure several bilateral deals worldwide. My hope is they’ll do so in a way that focuses on pooling our shared capacity and building scale against the PRC.

Jordan Schneider: The J.D. Vance perspective of promoting the AfD in Germany and suggesting that if the German right doesn’t win an election, they’re no longer our allies, represents a very insidious mindset. As you said, Rush, we forget how fortunate we are to have so many countries essentially aligned with our worldview.

Rush Doshi: We fought two world wars against Germany and one against Japan. These were the preeminent industrial powers of their regions in the 20th century and part of the 19th century. Today, they’re stalwart American allies. We have disagreements, but they’re on our side and share our values. That represents an incredible accomplishment and success story.

The idea that culture war issues should dictate our approach to these allies on the greatest geopolitical challenge America has ever faced is like cutting off your nose to spite your face. It doesn’t make sense.

That said, the Trump administration could still achieve scale. One critique of our piece is that it’s an idealistic dream that made sense five years ago but not now. I strongly disagree. There is a Trump path to scale, likely less focused on persuasion — which President Biden emphasized — and more on coercion. You could criticize the Biden administration for insufficient coercion.

Subscribe now

My point is this — you probably need to mix and match different tools depending on whom you’re dealing with. But we all know where we need to go. It cannot be the US versus China one-on-one. A unilateral America retreating to its own sphere would be poorer, less powerful, with deteriorating quality of life and diminished global influence, including on factors that make us prosperous. That’s obviously not a good path.

Jordan Schneider: Rush, you’ve outlined the two failure scenarios. The Trump administration would attribute one to the Biden administration — being too soft, allowing allies to avoid defense spending and ride our coattails into oblivion. I’ll let you elaborate on the contrasting errors.

Rush Doshi: One critique the Trump administration makes of the Biden administration is that it focused too much on persuasion and not enough on pushing allies — what you might call coercion. This resulted in free-riding. There’s a good counter-argument: progress occurred with AUKUS, the Quad, European involvement in Asia, and Asian involvement in Europe, which began moving toward allied scale.

However, the Trump administration’s critique that more needed to be done has merit. Kurt and I also believe more scale is immediately necessary. AUKUS was essentially a down payment on the kind of allied approach we need to institutionalize across the board. As someone deeply involved in AUKUS in 2021, working on negotiations for the White House under Kurt Campbell, I recognize it was novel then. We need to reach a point where such arrangements become routine.

Conversely, the Trump administration’s potential error would be excessive coercion — thinking you can bully everyone into compliance. Another error would be believing allies don’t matter, that America can go it alone, and that relations with Japan or Korea should focus solely on bilateral trade deficits and Treasury purchases.

Consider the economic picture: If we want to ensure scale for our industries when China is more competitive on a marginal cost basis, protecting only the American market isn’t enough. Without access to allied markets and allies’ access to our market, our companies cannot achieve scale relative to China. We need pooled market share.

The Trump administration has discussed erecting a tariff wall. In my view, a regulatory wall might be even better against certain Chinese exports. Beyond that, reducing trade barriers among allies would facilitate greater scaling up within the protected zone. Achieving this requires a mix of coercion and persuasion — combining Biden and Trump approaches.

Currently, President Trump has taken the coercive route with Liberation Day, sparking negotiations and deal-making opportunities. Hopefully, these deals will incorporate persuasive elements. Ideal deals would involve allies investing more while we reduce trade barriers — perhaps to zero-zero tariffs as some on Trump’s team have suggested. Allies would increase defense spending or work with us to improve their military capabilities. Additionally, coordinated steps would prevent China from dominating allied markets, including shared tariffs and regulatory policies to build a protective moat for all participants.

This kind of deal is achievable. As the Trump administration pushes these negotiations, I hope that’s their model. Some say they need to negotiate 75 deals, which seems excessive. I believe they need 8-10 key agreements. If they get those right, they can build allied scale differently from President Biden’s approach. They can claim victory and say they succeeded where others failed — that’s fine. The point is to achieve scale, and if they accomplish that, I’ll applaud them.

Jordan Schneider: Let’s stay on the Germany example for a second. The Biden administration was not able to get Germany to break its debt covenants and spend $1 trillion on defense. On the other hand, it basically took administration officials saying things so intense that Germany and Poland decided they can’t really trust extended deterrence anymore.

I don’t think the wake-up moment has necessarily come for Taiwan, Japan, and South Korea, but we seem to be in a very different timeline in the European context.

Rush Doshi: The risk of the current approach is that it’s so intense it pushes Germany away from being a core US ally on issues we care about. There’s a lot of flirtation in Europe right now with different forms of strategic autonomy, which could be interesting on their own. But if driven by dissatisfaction with the US, it might mean they’re not on our team when we need them.

Consider a concrete example: in 2020, the Europeans and Chinese negotiated the Comprehensive Agreement on Investment (CAI). This was fundamentally negotiated because Europe was concerned they couldn’t rely on the US. The Biden administration was upset — we had just taken office. That agreement eventually fell apart, but there’s a world where it doesn’t.

What worries me now is seeing Europe signal they want to hedge with China. That’s not sensible policy, but great powers can “commit suicide for fear of death” — a political science phrase I’m applying to this context. When you push an ally too aggressively, you can change their domestic politics to make it more anti-American. You can activate questions of face and respect that lead them to make poor decisions. Countries don’t always act in their rational self-interest, especially in times like these.

You can reasonably critique the Biden administration for not pushing allies hard enough, and the Trump administration for pushing too hard in ways that undermine transatlantic unity on China — which is extremely important for preserving America’s position. The truth is you need to do both. You need a nuanced approach, which either administration could achieve if they recognize the urgent need for scale and exercise more humility in American statecraft.

The capacity-centric statecraft we’re advocating is humble. We acknowledge there are capacities America has that its allies need, which we can provide. But we also recognize there are capacities America has lost that only our allies possess, and we need to regain them. How easy will it be to get Germany or Japan to manufacture in America if their domestic politics turns anti-American and their governments feel that accommodating the United States contradicts their political self-interest? That’s the risk — we might lose the transatlantic relationship we need if we’re not careful.

Jordan Schneider: In dark timelines, how does Trump 2.0 turn into America’s Gorbachev moment?

Rush Doshi: That was a provocative line in our piece. Personally, I’m hoping that doesn’t happen. Kurt and I intended to provide a strategic theory of how America should conduct itself with alliances — different from past approaches and valuable for the Trump administration. It’s not meant to be partisan.

The Gorbachev reference relates to two things.

1. There’s a risk that we neglect our allies’ agency, respect, face, and autonomy, causing them to move away from us if pushed too hard. Mikhail Gorbachev didn’t assume every Soviet republic would abandon his experiment, but that’s what happened. The same could happen to the American system if we renegotiate foundational elements, become unpredictable, interfere in allies’ domestic politics by supporting specific parties, or dismantle tacit bargains that underpin the American alliance system. This fragmentation would be catastrophic.

2. Successfully competing with China requires state capacity. Gorbachev’s reforms, and especially Yeltsin’s shock therapy, destroyed Soviet and then Russian state capacity, making it harder to accomplish geopolitical and domestic aims. I worry we’re dismantling our institutions. I’m not saying they shouldn’t be reformed or renewed — there were different approaches possible for both DOGE and the trade war. What concerns me is that these critically important initiatives weren’t implemented in the most strategic way.

Consider the trade war as a thought experiment. Imagine if, before Liberation Day, the Trump administration had presented a coherent theory of the case — policy papers explaining they would raise tariffs on China, with everyone else on notice. Imagine they had outlined what would happen to other countries in 90 days if they didn’t strike a grand bargain covering economics, trade, finance, security, and relations with China. What if they had already worked out preliminary deals with a few countries before the announcement, surprising China with this collective approach? And what if markets saw a logical grand design behind this strategy?

Would yields be spiking? Part of what’s happening is that markets don’t think we have a plan. The rationalization the Trump administration created after the fact about the trade war — imagine if it had come first. We’d be looking at a very different strategic situation.

You don’t go to war without preparation, as Russia demonstrated in Ukraine. That applies to global trade war also. What worries me is that there appears to be no document that articulates exactly what we hope to achieve globally, bilaterally, and multilaterally in the trading system. Without that framework, we’re improvising — and that’s frightening.

Jordan Schneider: As much as I want the Rush Doshi version of the Trump administration, I’m not optimistic that we’ll get it.

Rush Doshi: The purpose of my thought experiment is to imagine if we had done this differently. This could have been a moment of strategic advantage where the United States finally changed the global trading system to better reflect what’s needed regarding China’s participation. Instead, we have a unilateral, diminished American position with the entire world angry at us. I don’t believe that was the optimal approach.

You’re right — there were people with different proposals. My argument is that they should have made a different case, and I’m not certain anyone made this case. This is what they’re saying now — their post-hoc rationalization. Rewind the clock two months to the transition period. Had they presented this case then and informed the press — even if they ultimately imposed tariffs on allies — at least there would have been a coherent narrative before implementation. That story never emerged.

Jordan Schneider: Assuming the bear case from an American policymaking perspective for at least the next four years, we face several potential futures. One scenario involves democratic allies realizing they can’t count on the US, but needing to band together, put their houses in order, increase defense spending, and establish beneficial agreements between countries like South Korea and Germany. Let’s explore this path first. What might the global order look like with China checked out, while America and the rest of the developed world — uncomfortable with Chinese ascendancy — attempt to establish a new equilibrium?

Rush Doshi: China currently appears quite excited. President Xi is traversing different countries and striking deals. They clearly see an opportunity. I recently met with Chinese academics who indicated that China’s propaganda department is ecstatic, believing they haveways to position China as a better partner than the United States. Nobody should readily accept this narrative — skepticism is warranted. Nevertheless, China is actively attempting to drive wedges between the US and its allies, using trade as their primary mechanism.

My concern revolves around the bear case scenario. If we fail to reconstitute our alliances, experience greater transatlantic ruptures, and cannot establish meaningful agreements with our East Asian partners, America might retreat to the Western Hemisphere, attempting to make its stand there. This approach is doomed to failure. We cannot cede the entire world to China’s industrial power and expect that an America retreating westward can somehow escape the Mackinder trap. Mackinder warned against allowing any power to control the Eurasian “world island” — America’s pullback to the Western Hemisphere would enable precisely that outcome. This strategy was dismissed during World War II and makes even less sense now.

It’s particularly illogical in our globalized world where, even if the trading system fragments, it will inevitably reconstitute itself with another power at its core — potentially China. In that scenario, America becomes frozen out globally, unable to access the innovation, technological progress, and efficiencies developed elsewhere. America as the Galapagos Islands represents a bleak future. We must maintain our global presence and advocate for a world order reflecting American interests.

This isn’t merely “America First” rhetoric. For those embracing an America First perspective, this approach actually aligns with those principles by safeguarding American prosperity, not just geopolitical leadership. We’re discussing the quality of life Americans enjoy, which stems from constituting just 5% of the world’s population yet achieving extraordinary wealth — a reality made possible by a system designed to sustain this quality of life. We cannot simply demolish this system hoping everything will work out. That resembles Yeltsin’s shock therapy approach, which proved disastrous for the Soviet Union and Russia. My fear centers on an America that withdraws, creating a world that becomes China’s to lose — an outcome we must avoid.

Jordan Schneider: Do you believe the Trump team has learned lessons about alliance building from missed diplomatic opportunities during the first administration? What potential playbooks might China be implementing at this moment?

Rush Doshi: We should never discount China’s capacity to undermine itself diplomatically — essentially tripping over its own feet. Your observation is accurate, Jordan. There’s considerable confidence in Beijing. In my previous work on Chinese grand strategy, I’ve noted that the most important variable shaping their approach is their perception of American power. Beginning in 2016, this perception shifted dramatically. They concluded the world was experiencing unprecedented changes, leading to more aggressive and assertive Chinese policies, even toward countries they should have been cultivating.

One of history’s greatest strategic puzzles is China’s persistent alienation of India. China consistently pushes India toward the United States, when India might naturally prefer a more balanced position. China simply makes such neutrality impossible. They could certainly repeat these mistakes.

However, Jordan, we must acknowledge that our current approach differs fundamentally from anything attempted during the first Trump term. That administration conducted a trade war with China — arguably overdue — and attempted to renegotiate certain security agreements with allies. The current approach, however, essentially applies “Control-Alt-Delete” to the entire global trading system. We have a 90-day pause, but also a 10% blanket tariff — itself a revolutionary act. If we considered this tariff in isolation, without the chart the President unveiled, we would be shocked by it. Now, because our comparative reference is that chart, we’re virtually ignoring the importance of the 10% tariff.

To answer your question directly: American policy has become much more revolutionary than during the first Trump term, creating greater opportunities for China. The question becomes: even if they execute their strategy imperfectly, but marginally better than in the past, they could still accumulate substantial advantages. That prospect deeply concerns me.

Jordan Schneider: This connects to our earlier point about political instability. They might mishandle the opportunity, or they might not. Even granting them a 50% chance of failure still leaves several undesirable futures where smaller countries worldwide must navigate coordination problems without the United States providing necessary cohesion and facilitation. That’s frightening.

Rush Doshi: Precisely. A divided world where the US abandons its role as facilitator of collective action presents serious challenges. In political science, collective action represents one of the fundamental puzzles — what enables people to cooperate despite incentives to pursue individual interests? America functions as an anchor for collective action, making cooperation possible. If we withdraw, everything potentially fragments. Allies become divided and conquered. We might currently be in the “divide” phase of China’s “divide and conquer” strategy, but we’re inadvertently assisting them with the division — the conquest follows later.

Jordan Schneider: Every once in a while people ask me, “Jordan, why don’t you become the Matt Levine for China policy?” My answer is fundamentally that Matt Levine thinks securities fraud is funny, whereas I consider this stuff deadly serious. You too are someone who cannot look at the stakes of the conversation we’ve had for the past hour, shrug your shoulders, laugh about it, and simply move on with your life. Do you have any advice, Rush? It feels very helpless when the potential geopolitical errors seem larger than they’ve ever been in either of our lifetimes. Give me the pep talk. How do I keep putting one foot in front of another with the podcast?

Rush Doshi: There are a few things to keep in mind.

First, the Trump administration can negotiate agreements with allies and partners in ways that actually achieve scale. This remains conceivable. There are signs that some individuals running the negotiations are thinking in these terms. We know many people staffing the Trump administration at the mid-level, the key principals — not specifically the Cabinet, but many people below that level — believe in allied scale. They may not have called it that, but they understand its logic. That group exists, and there remains the possibility of landing the ship in that spot. I cannot predict the odds, and I do worry, but we shouldn’t be despondent.

Part of what must happen is continued discussion of these issues. One reason Kurt and I wrote our piece was to argue why allied scale matters. I hope as that argument circulates and others reinforce it, it may help shape some of the Trump administration’s thinking. The piece wasn’t meant as a partisan attack but as a strategic framework any administration could adopt.

The last thing I’d note is that this administration won’t last forever. Congress might flip in two years. America constantly changes. For allies and partners, they will find ways to work with the United States — whether through the legislative branch, pockets in the executive branch, or a suddenly enlightened executive that negotiates better deals. It’s not over.

Great power competitions can be lost in short periods, but China faces many challenges, too. They often struggle to consolidate and press their advantages. America possesses real advantages we shouldn’t neglect: two oceans, abundant resources, 70% of the world’s capital markets, the ability to attract the best talent from around the planet, and general political stability. These represent huge advantages that won’t disappear overnight. Some might be damaged or tarnished, but others might even strengthen.

We must remember this competition won’t resolve itself in two or four years — we must make investments for the long haul. Congress can do much. Everything the Biden administration did to build allied scale received bipartisan support, funding, endorsement, and praise from Congress. We must remember Congress wields power in this domain as well.

Jordan Schneider: One last question, Rush. I remember DMing you during the administration saying, “They’re not going to let you talk about this book, right?” And you replied, “Yeah, sorry,” but here we are now. We’ll have you back to do a deeper dive and an updated discussion about it. I’m curious — assuming nearly everyone in our audience has read the book — how did your experience being on the inside, reading intelligence, and interacting with the Chinese government affect some of the conclusions you reached toward the end of your research for that project?

Rush Doshi: Thank you, Jordan. I’m really proud of “The Long Game.” It was unfortunate I couldn’t discuss it while in government, and I’m glad people found it useful. The book was based on an assessment of 5 million words of Chinese Communist Party material, as well as a rigorous social scientific approach to their behavior. The conclusion was that China has maintained, since the end of the Cold War, a grand strategy to displace American order — first regionally, then globally. My time in government reinforced my view that these conclusions were correct.

In the book, I argue that going back to the 1980s, the US and China were essentially quasi-allies against the Soviet Union. Everything changed after the Gulf War, the Soviet collapse, and the Tiananmen Square massacre. Suddenly China viewed America as the biggest threat and inaugurated a new policy — “hiding capabilities, biding time” (韬光养晦 tāoguāng-yǎnghuì) — what I call blunting American power quietly, not assertively, while benefiting from America’s system.

In military terms, they pursued anti-access/area denial approaches to keep us out. Economically, they sought Most Favored Nation (MFN) and Permanent Normal Trade Relations (PNTR) status to tie our hands. Politically, they wanted to stall American-led institutions in Asia lest they become platforms for challenging China. This approach worked quite well until the global financial crisis.

In Central Committee materials, we see China change its perception of American power, adopting a new strategy formulated by the Central Committee. China should “actively accomplish something.” This approach had military, economic, and political components focused on building order within Asia, not just blunting American order. Militarily, China invested in power projection capabilities to influence its neighbors. Economically, we saw the Belt and Road Initiative and efforts to use economic statecraft against others. Politically, China built international institutions meant to serve as the foundation for order-building within Asia.

I wrote that this worked until 2016, when China’s assessment changed again. They adopted a new phrase to guide Chinese policy: “great changes unseen in a century.” The idea was that China’s current opportunities and risks were unlike anything they’d faced in a hundred years. We witnessed the inauguration of a global Chinese grand strategy focused on global military bases and winning in Taiwan; dominating supply chains economically and making the world more dependent on China (what President Xi calls “dual circulation"); and technologically, leading the fourth industrial revolution to ensure China wins those technologies not just for prosperity but for power. Politically, it aimed to change global institutions to make them more conducive to autocracy.

All of this is motivated by a desire to “rejuvenate” China, which represents a 100-year goal. The Chinese Communist Party is fundamentally a nationalist party seeking rejuvenation. They’ve always been that way. They’re also a Leninist party wanting to centralize power in pursuit of that goal. That’s why I say China has a grand strategy — it has the concepts, capability, and conduct to pursue that vision. They don’t always execute perfectly and make mistakes, but there exists a strategic intention to create a partial hegemony over part of the world, reflected in those military, economic, and technological indices I mentioned.

The intelligence community had previously outlined these conclusions in threat assessments but hadn’t substantiated them using open sources. I’m more confident of these conclusions now than when I wrote the book. Looking at President Xi’s behavior and the Chinese Communist Party’s actions over the past five years sustains the argument that they are pursuing a global grand strategy — not always well or brilliantly executed, but nonetheless deadly serious.

Jordan Schneider: Let’s look toward the future. Rush, what’s the broader open source project you’re launching at CFR?

Rush Doshi: I’m continuing my research at Georgetown as a professor and running the China program at the Council on Foreign Relations. One of our most exciting initiatives builds on my approach in “The Long Game,” where I relied on Chinese texts and materials to make my arguments. Now we’re going to mass acquire, digitize, and translate those kinds of texts at scale, making them more available to the public so people can essentially read China in its own voice.

If you do that, China essentially tells you where it wants to go. The Chinese Communist Party must communicate with itself and its cadres, and we want to make those communications available to others. This is something the US government used to do from FDR in 1941 through 2013 under Barack Obama, when we had the Foreign Broadcast Information Service that translated foreign material. That resource largely stopped being available to scholars in 2013. Our hope is to rebuild it at scale with a far greater source base and with the help of artificial intelligence. We’ve made major investments in this effort, have access to a large cache of material, and will share more very soon.

Jordan Schneider: It has been pretty frustrating over the past few months to see China be so important rhetorically to every policy decision while genuine curiosity about understanding the country seems at a relative low point — both regarding net assessment and strategic intention. It’s on us to make this material accessible and interesting. We can’t just complain about it; we must do the work and present it effectively, even if we need to be persistent. We need to make a strong case that this is both important and engaging.

Rush Doshi: Well that’s what you do on this podcast!

Jordan: Thank you! And look, The Long Game sold well, so there’s an appetite for this content somewhere! I fully believe in the mission and have more faith in you than virtually anyone else to make Chinese strategic discussion something the world cares about and takes seriously. Godspeed, Rush!

Mood Music

如果人生是长夜，照亮漫漫人生路的又是什么呢？我想，那一定是爱。

——崔恩荣《明亮的夜晚》

祝生于春日的王嗣珠女士健康快乐。你将人生的智慧写进具体的生活，将平凡的日子过成诗。我无时无刻不被你启发着，鼓舞着。当我意识到自己是你生命的延续，我感到无比骄傲。

——你的朋友、孙女，陌生女人4号

亲爱的媎妹：

见字如面！

最近我在练习人生教练对话，希望能通过专业服务帮助更多女性重建主体性，获得丰盛的人生。我发现很多姐妹在探索人生愿景时会遇见各种限制性的信念，这些信念常常源于母亲甚至祖母。但同时，我们前进的动力、决心与勇气也常常来自于这些了不起的女性。总之，母女关系、祖孙关系似乎是女性自我探索中绕不开的话题。

在一次跨代际疗愈分享会上，一位来自旧金山的心理咨询师说：“在了解我与母亲和祖母共同的人生体验中，我找到了爱和自己。”然而，每当我想要穿过代际的鸿沟、走进母亲和祖母的世界时，却总是体会到一种复杂的情绪：害怕自己的主体性被削弱、担心被“亲情”束缚，为无法回报她们的牺牲而愧疚。但尽管如此，最后占上风的往往依然是埋藏于心底的思念和深切的爱——母亲和祖母对生活的热爱无时无刻不启发着我，是她们用自己的双手构建了我最初的精神家园。

所以，虽然心中五味杂陈，我还是鼓起勇气开启了这场连接三代女性的疗愈之旅。一个家庭不同姓氏的三个女人生长在同一片文化土壤中，在不同的年代成为了女儿、母亲、祖母。其中有不甘与心酸，也有幸福和释然。为什么如此相爱的我们总是互相埋怨？我们为什么总是那样充满矛盾、无法自洽？父权社会如何塑造了我们之间“拧巴”的关系？母女和祖孙又该如何跨越代际的鸿沟，寻回最本真、纯粹的爱？

1、作为女儿和孙女：爱何以沦为“枷锁”？

童年时期，母亲与祖母是我们最亲密的伙伴。她们的关爱给我们安全感，也塑造了我们最初的安全感和世界观。于是，我们理所当然地将她们看作无所不能的存在。这份深厚的情感纽带使女儿对母亲与祖母产生了高度的依赖，让我们形成了近似「共生」的关系。

人生旅途中，母亲和祖母就像范本一样潜移默化地教导我们“如何做一个好「女人」”。她们教会女儿坚强善良、建立边界、不畏困难；也提醒她们要乖巧懂事、内敛自持、无私奉献。出于全然的信任与依赖，女儿就像一块海绵，毫无保留地接受女性长辈的影响，并无条件地爱着她们、努力满足她们的期待。

然而，随着年龄增长，女儿逐渐意识到，母亲和祖母并非完美无缺。当我们走出人生最初的“精神家园”，进入学校与社会、接触到更多元的女性形象和价值观，我们就会开始渴望独立、想要追求自我价值，因此对长辈的人生选择产生了诸多质疑和批判。

“女儿总是母亲最严苛的审判者”——许多女性的第一次顿悟就源于对母亲的反叛。当她们看到母亲和祖母日复一日地承担无偿家务劳动、为了支持男性亲属打拼事业而放弃自己的社交生活和梦想，甚至默默忍受家庭暴力，那个心目中如战士般无所不能的英雌形象可能会在瞬间崩塌。在觉醒的女儿看来，母亲和祖母是如此平庸甚至软弱。我们哀其不幸、怒其不争，为母亲“心甘情愿”接受家庭的束缚感到痛心，也怨怪母亲和祖母为什么不能勇敢反抗、为小辈女性撑起一片真正的天。

然而，女儿们不明白的是，「逃离」总是困难重重，不是每个母亲都有出走的资本。而且，母亲对出走的犹豫可能不仅出于“畏惧”，更源于对家庭成员真诚的爱。父权社会常常将再生产的责任从社会层面转移到家庭内部，尤其家庭中的女性身上。因此，当家人希望在事业上有所发展时，母亲往往会主动承担更多家庭责任以支持对方追求自己的理想，而母亲支持的对象也可能包括女儿（例如帮女儿带孩子）。

这种责任的转移会加剧母女之间的矛盾，使女儿在追求个人目标时被愧疚感缠绕。爱反思的东亚女儿们时刻不敢忘记，自己拥有今天的生活少不了女性长辈的托举。所以，虽然很多女儿的心早已飞向远方，她们却依然会在权衡之后选择压下对自由的渴望，放弃前程以减少母亲的负担，甚至和母亲一起背着原生家庭挣扎前行。

或者，她们也可能狠下心“斩断亲缘”，甚至和母亲站在对立面，成为那个剥削母亲的恶人。总之，和父权社会下的所有女性关系一样，母女关系也很可能演变成一场「零和游戏」。你的自由，总是要以我的「自由」为代价。

更糟糕的是，在这种“零和博弈”中，真正的罪魁祸首——父亲和丈夫——往往会在轻车熟路地扮演起「隐形操控者」的角色。对女儿，他们常用“你妈都是为了你才这么辛苦”的话语来进行情感勒索，让女儿从小到大都沉浸在愧疚感中、进而忘记一切的源头其实在于不公平的家庭分工。

对妻子，他们教训对方“我妈是为了帮你才一把年纪还要操持家务”，用女人间的「婆媳矛盾」掩饰男人对女人的「性别剥削」，从而美美隐身、坐收渔利。而可笑的是，他们自己吸起母亲、妻子、乃至姐姐的血来倒是毫不客气，很少因为剥削自己的女性亲属而感到自责，或是担心自己“不够独立”。

不过，鸡贼的男性也不总是这样疾言厉色。作为父权社会最忠诚的跟班儿，他们真是不能更懂“胡萝卜加大棒”的驭人术了。男人总是用轻飘飘的一句「妈妈真伟大」抵消女性数十年的“吃苦耐劳”，而这恰恰是一种最用心险恶的“捧杀”——他们对女人牺牲个人利益的选择大加“赞赏”却不给予任何实际补偿。男人动动嘴皮子就轻松逃避了给予女人物质回报的责任，自然也就否定了女性劳动的巨大价值。不知不觉间，女性的独立人格被抹杀，她们把「单方面供血」当成了「爱的供养」、哪怕榨干自己全部心力也甘之如饴。

但爱难道不应该是双向、互惠的吗？真正的爱源于真诚与平等，理解和共情，虽然看似平凡质朴，却终究难掩其赤诚热烈的本色。「爱」应该让被爱者感到安全、舒适，让给予者感到欣喜、满足。因此，忍耐不是爱，控制不是爱，牺牲更不是爱。不出于本心的过度付出可能是父权教化下的本能、习惯、是一种自我感动甚至情感勒索，却唯独不会是爱。

因此，尽管觉醒后的女儿依然深爱母亲和祖母，却难以再心安理得地感恩她们的「伟大」。看透剥削的我们开始追问，这样的「伟大」究竟由谁定义，又是谁在从中获利？

2、作为母亲：我把梦想托付给你，也把未竟的人生系于你身

1980年代计划生育政策出台后，大部分母亲都怀揣着满满的期许、迎接那个独一无二的小生命。许多母亲把女儿看作自己未竟梦想的延续，渴望通过女儿的成功来弥补人生的遗憾。她们希望女儿不再重复自己的命运，而是去往更广阔的天地。在《苦尽柑来遇见你》中，爱纯因为与男友私奔而被高中退学，后来又因婚姻和生育放弃了上大学的机会。但她始终期望女儿金明能接受最好的教育。当金明提出自己有机会去日本留学时，爱纯宁愿卖掉母亲留下的房子，也要托举女儿奔赴远方。

然而，这份深沉的爱也有可能演变成对女儿生活的过度干预。妈妈常以“为你好”为由否定和干涉女儿的决定，希望她能像自己希望的那样踏上「正常且正确」的人生道路。母亲一心想让女儿“幸福”，却经常忽视价值观的代际差异。在她们眼中，理想的人生脚本是“找一份稳定的工作、结婚生子”；但对成长于新世纪的女儿而言，这样的模式早已不是唯一的选择。

在经济下行的时代，“体面稳定”的工作变得难以企及，互联网的普及也让女儿看到了婚姻与育儿之外更多元的人生可能。在这个女性主义浪潮兴起的时代，“女人离不开男人”、“生个孩子人生才完整”的规训不再是完美生活的范本，而恰恰是女权主义者要批判和摒弃的“精神糟粕”。

此外，国家统计局人口普查显示，1982至2020年间我国15-65岁劳动年龄人口中，女性劳动参与率从77%下降到55%。由此可知，近几十年间越来越多的母亲选择退出职场、成为家庭主妇。失去公共角色的她们更容易陷入父权家庭的泥潭：丈夫的长期缺席、社会对女性家务劳动的忽视无时无刻不在贬低她们的自我价值。于是，在家庭和社会中都严重失权的她们只能将女儿看做唯一的精神支柱。

因此，女儿经常不得不充当承载母亲负面情绪的容器：听她们无休止地倒苦水、安抚她们的情绪，还要为她们“出谋划策”，而这也是女孩常比男孩早熟的原因之一。甚至在某些家庭中，女儿还要充当“丈夫”的角色、去弥补母亲渴望却无法从父亲那里得到的爱和关怀。这种错位乃至畸形的情感连结使得母女关系变得异常“拧巴”、甚至爱恨交织。

与此同时，母亲们也在女儿身上寻找自我认同。母亲希望女儿作出和自己曾经一样的选择，因为只有这样，自己的人生才不会看起来像是一个“错误”。在《从零开始的女性主义》中，上野千鹤子提到，在她三十多岁仍未婚未育时，母亲曾对她说：“有男人却不结婚生子，难道你要否定我的人生吗？”

漫画家田房永子也曾写道，她的母亲并不是想控制她的人生，而是“为了得到认可才拼命挣扎”。那些充满攻击性的话语、控制欲强的行为其实来源于母亲内心深处的焦虑感——正是因为被家庭和社会“除名”，她们才会将孩子视为自己的“影子”或者“作品”，期盼孩子「走自己的老路却依然能成功」。这样，母亲的人生价值也就得到了证明。

其实也许每位母亲在女儿呱呱坠地的那一刻都许下过一个朴实的心愿：“只要她健康快乐就好。”然而，随着时间流逝，这份简单的期许逐渐膨胀、变质——她们希望女儿成绩优异、事业有成、家庭幸福，希望她能看遍世界、见多识广，又希望她能常回家陪伴自己、成年后依然把母亲当成生活的重心...这样饱含爱意却又无比沉重的期许消耗着母亲的心力，也让女儿感到疲惫、窒息，甚至想要逃离。

可是，满怀期待的母亲和背负期望的女儿都应该记得，最初那个质朴、真挚的愿望才是爱最纯粹的模样。

3、作为祖母：先是母亲，再是自己

对祖母而言，养育孙辈是一次重返旧途的旅程。与母女关系相比，这份跨代关系少了些冲突与期望，多了些温柔和宽容。有时候，正是那一层自然的距离让爱变得更加轻盈。就像《明亮的夜晚》中，智妍搬到偏远小镇与祖母同住。尽管祖母与女儿关系并不亲密，却仍然以温柔又略带疏离的态度接纳了孙女，两代女性由此建立起了互敬、互惠的情感联结。

我小时候曾和奶奶一起生活。她记得我喜欢的饭菜，知道我最好的朋友是谁。搬新家时，她还专门为我留了一个房间，挂上我儿时最喜欢的粉色碎花窗帘。考大学那年，她陪我在外地租房。因为心疼我没日没夜地学习，她每次出门散步总会带回我最喜欢的蛋卷。她也是第一个知道我被大学录取的人。虽然看不懂录取通知书上的英文，却还是抱着我哭了，激动地说：“真的辛苦了。”

奶奶是家里最懂我的那个人。她总能看透我坚强外表下的疲惫和委屈，理解我的倔强与不甘；她夸我努力、聪明，却也不忘提醒我：“要记得给自己放个假。”我常在自己身上看到她的影子：我们都拥有细腻的情感和鲜活的记忆——只要三言两语，就能唤醒彼此共同的回忆，重现那年那月那日的温度、气味和感觉。

小时候她常对我说：“我们是好朋友。”现在我也想对她说同样的话。即使如今我们相隔整个太平洋，我仍会每周坚持和这位“远方的朋友”通话。有时我不禁想，如果我早出生五十年，或她晚出生五十年，我们一定会是最亲密无间的朋友。

祖母那一代人生于二战后，经历了物质短缺、教育匮乏的年代。对当时的女性而言，家务劳动和养育儿女可能是唯一一条实现自我价值的途径。她们勤劳、节俭、对自己格外严苛。到了今天，这种习惯常常被年轻人误解为“没苦硬吃”。但这种选择背后，其实承载着一个时代特有的价值观念。祖母从小被教育：“个人主义是自私的”、“自我”不该存在，于是“儿孙享福就是我的福气”便成了她们的人生信条。

很多祖母害怕生病，并不只是因为身体痛苦，而是担心自己成为儿女的“负担”。这是她们在父权和时代规训下将自我彻底工具化的结果：认为自己一旦对家庭不再“有用”，就会彻底失去存在的意义和价值。做饭、带娃、洗衣、打点亲戚，照顾邻里……这些都是她们赋予自己的价值，也折射出了她们被轻视、打压的心酸过往。

祖母的一生如同一条既定的轨道：年轻时住在父亲家，后来住在丈夫家，最后又来到儿子的家。她从未真正拥有一个完全属于自己的空间。她不懂什么是女权，也不觉得自己需要成为“大女主”。可她感受到的幸福却是具象而实在的：孙女的一个拥抱、孩子渐渐增长的身高、邻里间的一句寒暄都能带给她快乐和满足。祖母的自我价值不在于反叛和对抗，而是深嵌于使自己与她人、生活、自然紧密相连的关系网络之间。

她的生活是那么简单却又那么厚重。正所谓“平常心是道”。在晨光暮色之间、在日复一日的照料和陪伴里、在菜市的喧嚣声和厨房的烟火气中，不知不觉间，她早已用一种古朴的智慧编织出了一个不张扬却温热无比的世界。

结语

三代女人，成长于截然不同的时代，手握不一样的人生剧本，却都在积极地探索生命、感受真情。祖母的爱像一条河，流得缓慢又固执，却让人觉得安稳宁静；母亲的爱像风，急切而强烈，偶尔束缚我的自由，却终究会为我指明前进的方向；女儿的爱像雾，初起时模糊，直到长大后真正“看见”家中的女性，才渐渐变的清澈明朗。

我们曾抗拒母亲和祖母身上的“缺点”，却在某天突然发现自己其实早就继承了这些特质；我们想要离开她们的控制，却又在羽翼逐渐丰满后禁不住回望。在父权社会的轨道中，我们在彼此身上看见自己：是祖母，是母亲，也是女儿。三代女性因爱共生，因共同的困境紧紧相连。而正是在这种命运轮回般的连结中，我们从彼此身上找到了疗愈的可能。

就此搁笔，期待下一次和大家见面！

陌生女人4号 林犀*

二〇二五年四月二十八日

*本文由陌生女人4号主笔，陌生女人1号编辑。

Double-edged swords in the US-China Cold War

A guest post by the excellent Kyle Chan of the High Capacity substack.

The US and China are in a cold war, not a trade war. This is something much bigger than tariffs and trade deficits. It is much bigger than Taiwan or semiconductors. And it began long before Trump or Xi. The US and China are locked in a global contest of power that is playing out along every dimension: economic, technological, military, cyber, soft power, global prestige. Both sides are searching for any tool, any weapon, any piece of leverage they can use against the other—short of direct military action.

There is no such thing as escalation dominance. Trump thinks the US will win in a trade war because China sells more to the US than the other way around. A tit-for-tat escalation on tariffs means the US will always be able to tariff more Chinese goods than vice versa. Adam Posen has recently argued it’s actually China that has “escalation dominance” (a RAND concept in nuclear deterrence) because China has other ways of escalating beyond tariffs, including potentially denying Americans access to Chinese-made goods from smartphones to medicines. However, the reality is neither side has escalation dominance because both sides have already gone far beyond trade measures. If you’re looking at the full range of actions beyond trade tools, there’s virtually no limit to how far each side can go.

The US and China are posturing as if they have escalation dominance, which makes the problem worse. US Treasury Secretary Scott Bessent said on CNBC that China had made a “big mistake” in retaliating against Trump’s tariffs because China was “playing with a pair of twos.” China’s Ministry of Commerce has said that China would “fight to the end.” While there are already signs that Trump is backing down, the confidence that each side feels—or at least tries to project—only fuels a downward spiral of recklessness and emotion-driven bravado.

Double-edged swords

Every weapon in the US-China Cold War is a double-edged sword. Because the US and China are so deeply integrated—both in terms of bilateral ties and as parts of a highly integrated global economic system—any action that one country takes will end up hurting both sides to some degree. The question then becomes: what is the balance of pain? Are you able to inflict more pain on your opponent than you would on yourself? It’s useful to map out the different tools and weapons in terms of the relative costs to each side, as I’ve tried to do in the diagram at the top. Which tools fall into which quadrants?

Both sides are searching for asymmetric weapons where the damage caused to the other side far outweighs the harm to oneself. China believes critical minerals are one such asymmetric weapon (top-left quadrant). The US believes semiconductor export controls are one of its asymmetric weapons (bottom-right quadrant).

There are weapons that would blow up both sides (top-right quadrant). For example, if China cracks down too hard on US companies operating in China, this would have a severe chilling effect on all foreign companies in China. Or, as Trump is now learning, imposing extremely high tariffs on all Chinese goods can have huge costs for US consumers and producers. An act meant to pressure China has backfired spectacularly. CEOs of major US retailers recently warned Trump of possible goods shortages. The US basically placed an embargo on itself.

Tools and Weapons

It’s useful to think about the tools and weapons that China and the US are using in terms of categories of goals:

• Trade tools: Conventional policy tools aimed at shaping trade flows, including tariffs, import licenses, quotas, local content requirements, and other non-tariff trade barriers.

• Competition tools: Policy tools designed to insulate domestic firms from competition and slow down the other side. For example, US-led export controls on semiconductors and semiconductor manufacturing equipment to China are designed to slow down China’s AI progress, among other goals.

• Human rights sanctions: Punitive measures meant to punish the target country for human rights violations. For example, various US bans on solar and textile products due to concerns over the use of forced labor in Xinjiang.

• Defensive national security tools: Defensive measures that are meant to prevent or mitigate potential national security risks, such as the US ban on Huawei telecom equipment or China’s “delete A” (i.e., delete America) campaign to remove US hardware and software from major state-owned enterprises.

• Offensive military degradation tools: Measures aimed at constraining the military capabilities of the other country. For example, China’s export controls on heavy rare earths, which are key inputs for US weapons systems. Or US controls on advanced chips and computing hardware to limit China’s ability to improve its missile systems.

• Pain tools: Tools aimed at causing outright economic or material pain among the population. For example, China reducing purchases of US agricultural goods to cause economic pain for American farmers. Or the US imposing an extra round of retaliatory tariffs meant to increase the economic pain for Chinese producers.

Blurred lines

The lines between trade, geopolitical competition, and national security are becoming increasingly blurred. The actions taken by both sides in the US-China trade war have already spilled over into areas far beyond trade. For example, the surprise arrest of Meng Wanzhou, the CFO of Huawei and daughter of the company’s founder, over alleged sanctions violations was treated by Trump during his first administration as a bargaining chip in negotiations with China. China’s export controls on critical minerals are a move that extends far beyond trade, targeting key inputs into America’s defense industry and power infrastructure. And of course, lurking in the background is an ongoing cyber war, including China’s successful cyber infiltration of US critical infrastructure and telecom networks.

Different policy goals are increasingly mixed together. For example, the Biden administration effectively shut out future Chinese EV imports through tariffs and a national security ban. These actions mixed together several different goals: leveling the trade playing field, protecting US automakers from Chinese competition, and addressing security issues around espionage and even remote control for “connected vehicles.” Mixing tools and goals together might seem like a way to kill two birds with one stone, but it ends up diluting their effectiveness. While this was a problem in past administrations, that pales in comparison to the blind hammer-throwing of this one.

Timing and sequencing

One curious pattern has emerged in all this. Both sides seem to be preempting the other side’s actions by implementing some of these same actions in advance.

• Nvidia H20 chips: For a while, the US seemed on the verge of banning Nvidia’s H20 chips from China. But before this happened, China’s NDRC released new energy efficiency rules that would have effectively banned Nvidia H20s. (The US has now gone ahead and effectively banned the H20 chip for China.)

• BYD Mexico plant: BYD’s plans for building a new EV plant in Mexico were put on hold after Trump got re-elected. Then suddenly in March, China’s Ministry of Commerce jumped ahead and withheld approval for BYD’s Mexico plant, arguing that BYD’s technology might get “leaked” to the US (which doesn’t make any sense given the many BYD plants popping up all over the world).

• Critical minerals: A few days after China imposed export controls over heavy rare earth elements, Trump signed an order to look into imposing tariffs on critical minerals.

One explanation for this pattern of actions is a battle over symbolic control. Rather than getting hit by a ban by the other side, it looks like you have more control when you jump ahead and implement the ban first yourself. It’s like the classic line: “You can’t fire me—I quit.” The end result is the same but the sense of agency switches.

Another factor is control over timing and sequencing. As each country tries to find chokepoints to use against the other side, they’re also trying to patch up their own vulnerabilities. Each country would prefer to do so at a pace and manner of their own choosing. For China’s semiconductor industry, this means retaining access to some foreign equipment and components while gradually substituting in domestic firms for pieces of the supply chain—when they’re ready. For the US, this means gradually reshoring or friendshoring critical parts of its supply chains to reduce dependence on China.

Both countries are trying to avoid powerful shocks that are unexpected and sudden, like the first Trump administration’s export ban on ZTE, which nearly destroyed the company. Even anticipated disruptions can cause near-term pain when they mess up a country’s timing. For rare earths, it’s true that the US can eventually scale up production from domestic and other non-Chinese sources to a certain extent. But this still takes time, and US supply chains could suffer significantly from shortages in the interim, as a set of US agencies have recently warned.

Fear cycle

Lastly, as the US-China cold war escalates and spills over into new domains, each side’s actions increasingly reinforce the other’s fears. China’s rapidly expanding military capabilities, which it often trumpets loudly, and large-scale cyberattacks on US infrastructure feed right into American fears of a growing Chinese geopolitical threat. A Chinese spy balloon floating over the US in 2023 certainly didn’t help matters.

China, in turn, sees a US bent on trying to contain it internationally and suppress its development. [Jordan: well, this is something many in the Chinese system have believed for decades now, and you can find Xi speeches from way back in 2013 saying that “international hostile forces are intensifying their plot to Westernize and split China”] Recent US actions that feed into this belief include efforts to get other countries to do trade deals that shut out China, increasingly stringent semiconductor export controls, and even the Biden administration’s AI diffusion framework, which sought to restrict China’s access to US chips by restricting the entire world’s access to US chips.

These mutual fears may have become too deeply entrenched to roll back. But the US should recognize how its actions may in fact bolster the CCP’s legitimacy and validate Xi’s focus on security and national strengthening. And China should recognize how its growing assertiveness is directly fueling a bipartisan backlash in the US.

Overcoming a Chinese Dual-Use ‘People’s War’ Invasion Fleet

A guest post from Joseph Webster, senior fellow at the Atlantic Council’s Global Energy Center and Indo-Pacific Security Initiative and editor of the independent China-Russia Report. This article represents his own personal opinions.

There has been plenty of coverage of how China could use an armada of civilian ships during a Taiwan contingency (see here, here, and here).

But even if civilian ships aren’t used to transport troops, they could be used as launchpads for PLA drone operations. In a Taiwan contingency, Beijing could mobilize dual-use industrial resources, including its maritime, drone, and battery capabilities, to support military operations.

This article explores how China’s vast civilian maritime fleet — supported by unmanned platforms powered by next-generation batteries and AI — could enable a distributed, real-time radar and sonar sensor network around Taiwan and conduct drone strikes, mine-laying, and other operations.

The Chinese shipbuilding-drone-battery nexus

Just as China has a long-standing military-civil fusion program for science and technology, it also employs a whole-of-society doctrine for military industrial capacity. In particular, the Chinese military’s authoritative study reference on doctrine and strategy holds that a “people’s war” 人民战争 entails “the mobilization and the participation of the whole nation in the war and can maximize the war potential of the nation and countries” 具有实行全民动员、全民参战的政治基础，能最大限度地发挥民族和国家的战争潜力.

China is easily the world’s largest civilian and military shipbuilder, accounting for over half of all merchant vessels constructed in 2023, as measured by gross tonnage. Its fishing fleet is estimated to exceed 560,000 vessels, with its deep-water fishing fleet comprising about 3,000 ships. It has also constructed 50% more tons of military ships over the last decade than the United States, according to analysis from US Navy Captain (ret.) Thomas Shugart.

China’s unmanned drone capabilities are also formidable. Global drone production is concentrated in China: a single Chinese company, DJI, controls 70% of the global drone market, including the first-person view (FPV) drones used widely by both sides in the war in Ukraine. And as seen in Russia’s invasion of Ukraine, cheap, mass-produced, one-way attack (OWA) drones are revolutionizing warfare. While lithium-ion-powered FPV drones have a limited range (about 5 to 20 kilometers), they can be produced in astonishingly large quantities. Michael Kofman estimates Ukrainian annual drone production in 2024 will total between 1.5 and 1.6 million drones, mostly FPV drones. China’s output could dwarf these figures, owing to its greater industrial capacity. Moreover, the distinctions between nominally civilian and high-end military drones are becoming increasingly blurred — China could and would repurpose consumer drone-production lines in the event of a conflict with the US-led coalition.

Alarmingly, China’s quantitative industrial advantages in drones and ships stand to be amplified by qualitative improvements in battery technologies.

Next-generation batteries — like lithium-metal or solid-state — could vastly improve drone range and payload, potentially shifting the military balance of power. Shifting from existing lithium-ion-based battery chemistries to next-generation batteries could improve energy density, enabling drones to fly or swim farther, as well as carry greater payloads. As Brian Kerg notes, it may become increasingly difficult to distinguish drones from a precision-strike munition regime, and the PLA’s “close-in, low-cost, attritable precision strikes at scale” could support a Chinese amphibious landing force.

Not only will better batteries further blur the lines between drones and precision strikes, they will also present significant implications for electronic warfare: advanced batteries can allow drones (and manned systems) to “out stick” — that is, out range — opposition force systems, or enable a “targeting mesh.” As a RAND analysis notes, a targeting mesh’s redundant sensors offers superior performance over a kill chain: “unlike a [kill] chain — which can be rendered useless by the failure of one link — a mesh can retain structural integrity even when multiple elements fail.” A drone network of sensors and electronic warfare systems, powered by advanced batteries, could enable such a PLA targeting mesh.

Given their military implications, including for meshed networks, a quiet but deadly serious competition in advanced batteries is underway. Several Chinese ministries, including the Ministry of Industry and Information Technology, have ordered their top battery companies — including CATL, Geely, BYD, and Beijing WeLion — to work together on next-generation solid-state batteries, while also banning Chinese companies from supplying batteries to Skydio, the largest drone maker in the United States. These efforts may lead to technological breakthroughs for Chinese companies. For example, BYD claims it will begin deploying solid-state batteries in SUVs by 2027. Meanwhile, the US Department of Defense has designated China’s leading battery manufacturer, CATL, as a military company — a move which may indicate that China is developing a diesel-electric submarine with the help of CATL’s advanced batteries.

Drones, batteries, and ships could comprise a “People’s War Fleet”

In wartime, China’s vast and distributed civilian maritime fleet could be repurposed as drone carrier ships, anti-submarine warfare vessels, or radar and sonar collection platforms.

China may deploy drones on civilian vessels, leveraging first-person view, one-way attack drones’ low cost, light weight (often below 3.5 kg/8 lbs), and ease of use. Even civilian fishing vessels could likely carry several dozen drones. At a production cost of only $400, FPV drones can destroy $2 million tanks, with operators able to become proficient in months rather than years. Embedding AI and employing bigger, more advanced batteries on these drones would increase their lethality by making them less susceptible to electronic warfare (albeit while increasing costs).

The Chinese deep-sea merchant fleet is large and growing. Some analyses hold that the China-owned merchant fleet now stands at one-sixth of the world total; China also constructed over half of all new merchant vessels in 2023. To be sure, China’s interest in deep-sea merchant shipping is unsurprising — after all, China is the world’s largest trading nation. But China’s large, growing civilian fleet could nonetheless have important military implications.

To be sure, a Chinese unmanned drone swarm would admittedly face significant challenges. The Taiwan Strait has high-speed winds for most of the year (with spring and especially summer proving to be partial exceptions), impacting drone operations. Similarly, flying in an oceanic, salt air environment can lead to corrosion or necessitate operational adjustments for drones. Moreover, drones launched at sea will be unable to use “visual navigation” or tap onboard cameras and terrain features for navigation. Russia’s full-scale invasion of Ukraine demonstrates that drone swarms continue to face significant risks from enemy electronic warfare jamming, and friendly signal interference.

But drone warfare will evolve. Electronic interference can be overcome, to an extent, with better onboard AI software (albeit at higher complexity and costs, not to mention other AI-related risks). And improving battery chemistries could improve drones’ communications and electronic warfare systems, enhancing the platform’s lethality.

China’s civilian fleet also holds latent subsurface military potential, as it could launch potentially thousands of unmanned underwater vehicles (UUVs). These UUVs could employ active sonar capabilities to monitor the Taiwan Strait and nearby waterways for submarines, as well as conduct mine-laying or mine-countermeasures missions. This is not a far-fetched scenario: China is already using nominally civilian cargo ships to target subsea fiber optic cables around Taiwan. (China’s larger deep-sea vessels are better suited for launching UUVs, which are heavier and bulkier than airborne drones.)

Finally, the Chinese civilian fleet could enable a targeting mesh for the Chinese military by providing real-time radar data. While shipborne civilian radars are weaker than military-grade radars, their open-array radars nonetheless range from 64 to 96 nautical miles; larger ships with greater on-board power and higher mastheads can “see” farther. Next-generation solid-state radars — distinct from solid-state batteries — offer significant performance improvements over traditional magnetron radars. While each civilian vessel has only limited radar coverage, the Chinese navy could theoretically aggregate information from each ship to build a real-time, composite picture of maritime domain awareness, especially when used in conjunction with other platforms like commercial satellite imagery. This information could enable the PLA to track and target coalitional surface fleets.

Overcoming the People’s War threat

While the primary and most serious military challenge facing the coalition remains conventional PLA forces, a People’s War threat adds another potential threat vector. Luckily, many of the capabilities the coalition would use to deter and, if necessary, defeat China’s conventional forces can be repurposed against a dual-use People’s War fleet.

Given the significant risks the Chinese maritime fleet poses in the aerial, subsurface, and surface domains, coalitional militaries should monitor their fleet carefully. As Lonnie Henley wrote in a recent analysis, China struggles to maintain accurate information on its maritime militia — meaning any overhauls to China’s data-management practices may be an early warning indicator of changes in Chinese military tactics. China is unlikely to achieve strategic surprise in any Taiwan contingency. Still, insights into the Chinese civilian fleet’s operations could provide critical hints about the Chinese navy’s tactics in any quarantine, siege, or invasion of Taiwan.

To that end, US and friendly militaries should establish a baseline of Chinese civilian-fleet behavior to identify potential military-related anomalies; they should especially intensify monitoring of civilian vessels operating near sensitive military installations like Guam and Okinawa. And challenging Beijing’s South China Sea claims and highlighting China’s illegal, unreported, and unregulated fishing practices would limit the potentially dual-use Chinese fishing fleet’s size and operational scope, reducing its potential military intelligence collection capabilities and giving coalitional navies more freedom of maneuver in a contingency.

The United States and other coalitional partners must rebuild atrophied shipbuilding capabilities. A CSIS analysis found that a single Chinese firm, COSCO, constructed more commercial vessels by tonnage in 2024 than the entire U.S. shipbuilding industry has built since the end of World War II.

Scaling-up coalitional drone industrial capacity and technological know-how will be critical. Sending a clear, consistent demand signal to industry would help galvanize the sector and enable economies of scale. Given the growing strategic importance of dual-use drones, the United States and other markets should consider emplacing additional tariffs on Chinese drones, helping spur the rise of a domestic sector that could produce items in a military contingency.

Taiwan must invest substantial resources in its own military capabilities. Taiwan must scrap its Public Debt Act, which hampers critical security-related investments. Crucially, greater drone-manufacturing capabilities can ensure sustained production in the face of severe disruptions, while integrated drone operations at the conscript and reservist levels will bolster the ROC’s warfighting potential.

To maintain its military technology edge, the US-led coalition must lead battery innovation. Advanced batteries could give battery-powered and hybrid coalition drones superior range and capabilities, including in electronic warfare and kill-chain operations via a meshed network. Next-generation batteries could unlock directed-energy weapons and ultimately revolutionize warfare through low per-shot costs, unlimited magazine depth, and the ability to engage multiple drones or even missiles. Less than 1% of federal battery funding currently supports next-generation solid-state batteries, according to a Carnegie analysis. Incentivizing next-generation battery research via R&D tax credits, government prizes, and guaranteed DoD contracts could prove fruitful. The United States must also work closely with major international battery players — especially South Korea, Japan, and Taiwan — while ensuring robust domestic manufacturing capabilities.

The United States must pursue both defensive and offensive strategies for technological and industrial advantage in batteries and AI. This includes preventing dual-use technology transfers (eg. advanced batteries, solid-state radars for civilian craft, image-resolution processing used for drones, etc.) and tightening semiconductor controls with allies to curb China’s military advancements. Simultaneously, and given that China has stolen considerable amounts of Western intellectual property, the United States should acquire cutting-edge Chinese-developed technologies, especially in advanced batteries.

Finally, rapid AI development must continue. AI potentially offers sizable economic and societal benefits. Its military benefits are also considerable, as it will make coalitional drones more survivable in drone-on-drone warfare and enable human-machine teaming. Accordingly, the US-led coalition should remove power-sector bottlenecks constraining US development of electricity-intensive data centers required for AI. Cutting red tape surrounding electricity transmission and ensuring an all-of-the-above energy strategy inclusive of natural gas, nuclear, solar (which has significant complementarities with batteries), wind, and geothermal will ensure the United States and others are able to power electricity-intensive data centers. (Open AI’s ChatGPT requires nearly 10 times as much energy as a standard Google search.)

Policymakers must grasp the potential threat of a People’s War invasion fleet, provide substantially more resources to counteract the problems posed by China’s active and latent maritime threats, and act across a disparate, complicated set of interlocking issues, including drones, shipbuilding, batteries, and artificial intelligence. Congress must pass legislation to expand military and civilian shipbuilding, including via the bipartisan SHIPS Act. As ever, working with partners will be a force multiplier for the United States and its military. Overcoming the nexus of China’s drone-making complex, battery capabilities, and potentially dual-use civilian fleets will prove difficult, but the coalition must overcome this challenge.

Mood Music

Of all the policies Trump has rolled out, the two I’m most worried will have irreversible consequences for long term national power are undermining our alliance structure and the attractiveness of the country for high end international talent.

Angela Shen (ChinaTalk’s robotics and biotech analyst) and I published a piece earlier this week in The Washington Post, arguing how America’s student visa system is critical for future science and tech success. They’ve graciously allowed us to rerun it below.

As President Donald Trump is escalating the competition with China, he risks handing it a generation-defining victory by cutting off America’s ability to recruit the best talent in the industries of the future.

In the past month, roughly 1,400 international students and scholars have had their visas revoked or exchange records terminated. Students are being forced out for infractions as minor as dismissed traffic tickets with no link to antisemitism, protesting, criminal charges or anything else labeled a safety threat by the White House. With no clear logic or public explanation behind who is detained or deported, the result is a climate of confusion and fear.

The administration is also threatening the admissions process, with Homeland Security Secretary Kristi L. Noem warning she could block Harvard University’s ability to enroll international students entirely. Existing students are facing additional upheaval from the sudden freeze and elimination of research grants throughout the higher-education system.

This newly chaotic environment jeopardizes a vital talent pipeline that underpins America’s technological edge — and that, if it isn’t fixed, will have deep repercussions for U.S.-China competition for decades to come.

Foreign talent is a core contributor to the scientific discovery and technological progress that have upheld American leadership over the past century. International students make up about 42 percent of STEM PhD graduates in the United States, and rather than “stealing seats,” they are subsidizing Americans’ education by paying full tuition. Their presence generates revenue that sustains academic programs, lowers the trade deficit and funds financial aid that supports increased domestic enrollment.

In recent years, about 75 percent of these students have been choosing to stay. And they have delivered when they do: One study by the National Bureau of Economic Research attributed 36 percent of American innovation (accounting for quantity, quality and value generated from patents, as well as spillover effects) in recent decades to immigrants. One-half of advanced STEM graduates working in the defense industrial base were born abroad, and 60 percent of top U.S. artificial intelligence companies were co-founded by immigrants — most of whom arrived on student visas. And far from taking jobs from Americans, immigrants with STEM education fill skill gaps in critical fields such as health care and technology and occupy complementary roles that support U.S.-born workers.

This edge in building the technology of tomorrow — AI, robotics, quantum computing, advanced batteries — is now under extreme threat from the administration’s seemingly arbitrary crackdown on students. In the face of uncertainty around visas and funding, international STEM students are increasingly reweighing their options in favor of countries outside the U.S. that are doing their best to take advantage of this unique recruiting opportunity. If visa revocations continue, the U.S. might lose some of the best and brightest minds the world has to offer — systematically undermining the future of American innovation at a time when China’s homegrown engineers are already leapfrogging American competitors in key technologies.

China knows what it’s like to have brains drained by foreign schools and employers: Among international STEM PhD graduates from U.S. institutions, those from China choose to stay about 90 percent of the time. Beijing has been trying for years to reverse that trend and is surely celebrating America’s self-sabotaging decision to help Beijing do so.

In a 2021 speech, Chinese President Xi Jinping declared that his country must win the global battle to cultivate “human capital.” China has spearheaded dozens of initiatives to recruit talent, including Chinese workers who left for schools and companies abroad. These programs offer funding, family support and resources for research, aimed at making it easy for Chinese and foreign scientists working overseas to move to China. In the U.S., a group of Republicans in Congress recently introduced a bill to close off American schools to all Chinese students, effectively doing the Chinese Communist Party’s job for it. The bill is a long shot to pass but reinforces the message that China’s best and brightest should look elsewhere.

As of now, international researchers consistently view the U.S. as the more attractive destination for world-class education and opportunities, despite a notoriously complex and sluggish immigration system. But when something as routine as a traffic violation or a trip home could derail the futures of biology and engineering graduate students, trust erodes. The effects will reverberate far beyond the individuals pushed out. Mass visa terminations for arbitrary reasons send a clear message: Even if you try to play by the rules, you won’t be safe. Families considering a U.S. education now have to weigh not just sky-high tuition but also the possibility that a student could have to leave the country before they even earn a degree. When top global talent no longer sees America as a stable, long-term bet — in light of both visa and research funding insecurity — many will vote with their feet.

These concerns should resonate with Trump, who has repeatedly discussed the importance of winning the competition for workers. Just last year, he told the tech-focused “All-In” podcast that the government should automatically offer green cards to foreign graduates of American schools, a proposal he floated in his 2016 campaign as well. In December, he joined tech allies such as Elon Musk in defending the value of H-1B visas from critics inside his coalition. His administration now seems determined to undermine those goals.

In a recent call to revitalize America’s science and technology enterprise, Trump invoked Vannevar Bush, chief science and technology advisor to Presidents Franklin D. Roosevelt and Harry S. Truman. In his pivotal report, “Science: The Endless Frontier,” Bush wrote:

The publicly and privately supported colleges, universities, and research institutes are the centers of basic research. They are the wellsprings of knowledge and understanding. As long as they are vigorous and healthy and their scientists are free to pursue the truth wherever it may lead, there will be a flow of new scientific knowledge to those who can apply it to practical problems in Government, in industry, or elsewhere.

Bush handed America the blueprint for winning the 20th century: Open the doors to talent, fund basic science at scale, and let researchers chase truth. That formula has powered U.S. leadership to a Cold War victory and leadership in critical technologies such as AI and biotech today. And it’s one we should follow into the 21st.

This article followed our longstanding coverage of the importance of talent in US-China competition. Back in 2020, I wrote about how H1B and ICE rulings threaten to deter the brightest and best talent from the world. Since then, we’ve discussed how China has prioritized solving its domestic talent shortages, how foreign talent is the foundation of the US’s strong AI sector, and how leading Party thinkers see America’s open immigration policy as a key threat to China’s tech rise.

由于篇幅限制，本文为第四十九封来信的后半部分。前一部分参见：我爱的女人是一个星球，我爱的星球是一片湖（上）

在与湖建立亲密关系的这几个月里，我开始意识到曾经的自己是多么傲慢。那时的我刚刚摆脱男权社会“专门利他”的规训、急于向别人证明我是个「独立」的女强人。于是，我开始相信“本自具足”的神话，说服自己“我不需要任何人、我自己就能满足我所有的需求”。同时，我把那些为情所困的人定义为“软弱”、“不自洽”、“自我不完整”，并以此满足自己作为“大女人”的优越感。

可是，看似强悍的我却并没有获得我想要的幸福，我变得越来越像拉康定义的“男人”了。正所谓“人定胜天”，男人（The Man）必须证明自己是最强大的，因此才有资格控制和统治一切。他们要把一切都当作为自己的利益服务的工具，而不能真正需要或依赖任何人。于是，男人开始从女人和自然那里攫取资源，满足自己无所不能的幻想，并在榨干她们的价值以后将其一脚踢开。

（注：这里的女/男人更多是社会建构的女/男人，或者说女性/男性气质，而不一定指sex/gender identity。通俗一点讲就类似“精神男人”）

彼时我以为这就是幸福的秘诀，却在努力成为“独立大女主”的过程中变得越来越偏执、孤立，也在不知不觉间丧失了爱的能力。

不肯承认自己脆弱的人是没有办法去爱的，因为爱一个人、爱一个生命必定伴随着风险。爱一个她者总是代表某种「缺失」，爱意味着承认“我需要”。而男人不能依赖别人，不肯承认自己需要。所以，成为真正的“男人”的过程就是放弃「爱的能力」的过程。从这个角度看，男人才是被“阉割”过的、不完整的存在。

我不禁想起戏剧《饥饿的女人：墨西哥美狄亚》中那句著名的台词：“我没有兴趣品尝你那名为「男子气概」的软弱。”

幸运的是，我们并没有太多证明自己“强大”、优越、无所不能的枷锁，因而可以真诚地袒露脆弱，诚实地说“我需要你”。

成为女性主义者不应该意味着失去爱、放弃爱，也不应该代表着掩饰自己的脆弱、否认内心的需求。温柔的湖水告诉我，真正的强大其实是直面自己的脆弱。The most vulnerable is the most powerful.

正因如此，我一直坚信真正的革命必将由女性开始，因为在这个「无爱」的世界里，只有女人有可能懂得爱是什么。

不过，如今我理解的「爱」和传统女性在规训下认为的爱（或者说绝大多数异性恋里的“爱”）有一个最大的不同。那就是爱应该自觉、自愿、平等、互惠。爱不是自我感动式的奉献，不是道德绑架式的索取。爱不是单方面的供血和吸血；爱是相互依赖、彼此满足——是湖教会了我爱的真谛。

从那时起，我开始和人类以外的生命建立互惠的、相互依存的关系。我和室友们在秋末冬初为蜜蜂制作简易巢穴，帮她们捱过难熬的严冬；每年二月，我们外出收集秸秆，以备下个冬天为蜜蜂筑造新巢。

每天出门时，我故意给门留一道缝隙，再把坐垫摆好，这样邻居的三只小猫就可以在串门时睡个好觉。

春天，我们在房子前面的菜地里播种。我们细心照料蔬菜，而她们则在秋天送来沉甸甸的果实作为回报。收获时，我们对土地说：“谢谢你慷慨的馈赠”。

我甚至也和菜地旁的“稻草人女士”建立了一段双向奔赴的关系。稻草人倚湖而立，不分寒暑、不问日夜地守护我们的菜园，而我则用相机记录她飒爽的英姿。

除了欣赏路边的“真花”，我还用心制作手工花，向每片花瓣倾注全部的爱和心血，认真感受她的生命和存在。作为回报，手工花为我的房间增添了一抹生动的亮色。

朋友送我的维尼熊娃娃已经忠实地陪伴了我将近四年。某天突然发现她脖子开线、棉花都露了出来，我便立刻放下手里的事情给她“动手术”，又在缝针时想到她一定很疼，不禁湿了眼眶。

吃过饭的午后，我一厘米一厘米地为地毯“洗澡”，耐心地摘掉她身上的灰尘和毛发，就像从妈妈的毛衣上拿掉线团，又像是轻轻挠小猫的肚皮。

湖、花园、手工花...她们都是生命赋予我的礼物。我越来越深刻地意识到，是我周围的一切造就了我。我是我所有关系的集合。如果没有「关系」，我还能是「我」吗？

我的世界与以前大不相同了。每一样事物都是我的伙伴、姐妹、朋友，我发现自己原来从来不曾孤单过。我和每个偶然相遇的灵魂建立关系，向每一个独特的生命致意。我用心了解知更鸟和每一株小草，就像了解我爱的那些人类一样。

我对青山说“我一直在”、对细雨说“你好”。

我爱上了这个世界，而世界报我以歌。

就此搁笔，期待下一次和大家见面！

陌生女人1号 兔姐

二〇二五年四月二十四日

P.s. 本文标题化用了Paula Gunn Allen在1990年发表的散文 The woman I love is a planet; the planet I love is a tree（《我爱的女人是一个星球，我爱的星球是一棵树》），但内容和她的原作并无关联。

为全球华人游荡者提供解决方案的平台：游荡者（www.youdangzhe.com）
这世界的辽阔和美好，游荡者知道。使用过程中遇到问题，欢迎联系客服邮箱wanderservice2024@outlook.com.

【和放学以后永不失联】订阅放学以后Newsletter，每周三收到我们发出的信号：afterschool2021.substack.com 点击链接输入自己的邮箱即可（订阅后如果收不到注意查看垃圾邮箱）。如需查看往期内容，打开任一期你收到的邮件，选择右上角open online，就可以回溯放学以后之前发的所有邮件，或谷歌搜索afterschool2021substack查看。

截至目前，放学以后Newsletter专题系列如下：“在世界游荡的女性”系列、“女性解放指南”系列、“女性浪漫，往复信笺”系列、莫不谷游荡口袋书《做一个蓄意的游荡者》系列、“莫胡说”系列”《创作者手册：从播客开始说起》,播客系列和日常更新等。

春日好！本期放学以后信号塔由正在荷兰放假（复活节假期）的莫不谷轮值。霸王花也从西班牙来荷兰找我玩，我们还会一起去比利时的两个小镇感受春光。已经第三次来荷兰的霸王花到我家门口和我说：“荷兰真是一个发达的资本主义国家！”我抚掌大笑并深深同意，因为不久前刚去霸王花居住的西班牙某个大城市玩，深入生活就觉得荷兰真好，好的点在春天和夏天尤其突出，因为一切植物复苏，荷兰的好天气也来了，真是满目好风景。当然凄风苦雨的冬日荷兰就难比上西班牙了。

我在不放假的日子每天都在“写作15分钟”并在游荡者平台这篇文章下面打卡，我不打卡的日子就是我放假的日子。平常的日子创作，放假的日子就纵情地享受生活。所以我密密麻麻打卡的日子和空档的日子，都写满了幸福。今天这篇文章是我其中一天写作15分钟写下的关于“幸福”的文章，写着写着中间一度情绪高昂。也邀请你在这美妙的春日里进入到这细密日常的激越！

今天偶然看到一本童书《独角兽魔法传说》中的一页，主角是一个希腊港口城市的女孩，每天看着男人们扬帆起航出海，而她只能待在家里烤面包和照顾鸡，因为他们都说女人女孩做不了水手，不能出海。

女孩会从出海的水手那里听到各种关于海的故事和传说，美人鱼，波塞冬和海中的独角兽，然后女孩说：“听了这么多神奇的故事后，他们还能指望我在家里烤面包？”

于是她偷偷溜出了家，溜上了船，开始去做一个水手，去看看海的神奇，另一种生活的可能。

神奇的故事不一定为真，故事存在的意义也不是真实性。而是它的感召性：让人去出门探索，去感受的未知的calling.人恰恰是因为能听到各种calling，才会觉得生活有奔头，自己活着有一些崭新的地方去奔赴。

理解能力欠奉的人可能会argue:可是就是会有人喜欢在家里烤面包呀？烤面包也可以是很好的生活呀……

有些人仿佛此生都没有学会小学语文中的修辞手法：比喻。也从来没学会小学最基本的阅读技巧：联系上下文。

一个女孩仅仅因为性别就被限制了做水手的可能，只能被迫在家烤面包，这个事情和她是否喜欢烤面包有一毛钱关系吗？以及这么明晃晃的比喻，一本童书中的比喻，是真的那么难懂吗？小女孩都看得懂这是在讲述女性在这个世界因为性别而遭遇的困境，被剥夺的自我的可能性，就总有成年的人，完成过九年义务教育的人跳出来说：可是烤面包也可以很幸福呀！在家养鸡养鸭也可以很幸福呀！做家务生小孩相夫教子也可以很幸福呀！

幸福你大爷！只有你大爷在幸福！

写着写着愤怒起来的原因就是，这个社会充斥着这种掐头去尾欺骗女性或者自我欺骗的言论。总把很多女性迫不得已，被压迫，被诓骗的选择，说成“也幸福”。

这个世界很少有人拿这样的言论去骗男性：辍学不读书也可以很幸福呀！去扫大街当保安也可以很幸福呀，人生不用这么上进！天天在家照顾孩子不上班做个奶爸也很幸福呀！男人在家给全家人每天做饭一日三餐也可以很幸福呀！

男人从来没听过这种疯话，也从来骗不到他们。甚至李安听到这句话的礼貌赞美版，都难受崩溃。他在家做了好几年的饭，一直没有拍电影的机会，也找不到其它工作，靠妻子供养全家，丈母娘怕他心中不舒服，夸他菜做得好吃，说可以给他出钱开中餐厅，他都心中如油煎。

父权制的谎言骗不了男性，却时至今日都还能骗到女性。人想要感受到快乐和幸福非常重要的前提是“自愿”，幸福的秘诀是自由，只要“不得不”，就难有快乐和幸福。只有出门做水手，和在家烤面包，这两个选择都是“任君采撷”时，你才能说你的选择是“自愿”的。当你的其它选项被堵死了，或者其它选择之于你这个性别难如上青天时，你选那个仅剩的选择，整个社会合力让你选的选择，不是忠于自我，而是忠于容易。不是“也幸福”，而是“真可悲”。是不平等的社会加诸于个体的悲剧。个体需要非常强的主体性和反抗性，才能扭转这种“也幸福的悲剧”。

幸福很难的，知足常乐也是骗人骗己的妄语，什么时候人会说“知足常乐”呢？在自己没得到满足的时候来骗骗自己。人从来不会在真诚满足的时候说这句话的，因为真正满足的时候就会享受满足的快乐，不用打着幌子骗自己乐了。

持志如心痛，但是心痛如幸福。人心中有自己真正想做的事，无论是做水手，还是出门去开一家面包店，这些calling在心中常常带来心痛的感觉，就像李安想着拍电影在家做饭的时候的感觉，但是唯有这种心痛存在，才有可能带来一个人，仅仅作为TA自己的幸福，不用劝慰自己知足的真满足，不用诓骗自己也幸福的真幸福。

听了这么多神奇的故事，我怎么还会在家里烤面包呢？

任何故事有在你心中泛起神奇的涟漪吗？那些涟漪有让你有持志如心痛的感觉吗？听完你还要日复一日在家烤面包吗？

也有许多人，听完任何的故事，心中都波澜不惊。它们是淡人，是风吹不动雨打不湿的金钟罩（这个词的原意）和空心人，是尸体。

你不是它们。

我祝福你不是它们，和我祝福你幸福，是同一话的中译中。

如果没有那些让我的生命成为可能、并赋予我生命以意义的「关系」，我就不再是我了。

—— 朱迪斯·巴特勒，当代最著名的后结构主义思想家之一

亲爱的媎妹：

见字如面！

上学期初我搬到一座湖边的房子里住，从此便过上了推开大门就是漫漫湖水的日子。

住在湖旁边，你会轻而易举地爱上她。她好像有着某种巨大而神秘的引力，让我不由自主地想要靠近，一遍又一遍地去湖边游荡，只为确认她的存在、感受她的呼吸和心跳。

和一个人建立亲密关系需要细心的关注和长久的陪伴，和湖亦是如此——白天我在台阶上俯瞰，观察湖的心情如何随着天气变化；或是蹲在湖边看白色裙摆般的浪花，任思绪飞向九霄云外。

夜晚我枕着湖浪入眠，让湖水的温柔细语潜入梦来。

我发现自己的感官被成倍地放大，开始注意到以前不曾留意的种种小事。

秋天，我站在湖边注视夕阳隐没在湖的另一边，感受浸着水汽的湖风一天比一天冷冽。直到某天早上醒来突然发现湖浪声竟然不在了，然后才反应过来：“啊，终于上冻了啊...”

冬天，我戴着厚厚的毛线帽子穿过湖面，实实地踩龙骨一样坚硬的湖冰，用手指轻触湖微微隆起的脊背。

我裹着笨重的羽绒服躺进湖心松软的雪壳，眯着眼细数天空中的寒星。

春天，冰雪消融，我弯着身子贴近湖面、竖起耳朵听那穿透冰层而来的、春雷般涌动的水声。紧接着，湖水回来了——我看到水鸟翩然而至，野鸭在细雨中洗澡，兔子在草丛间如闪电般掠过，松鼠成群结队地在树梢上喧哗嬉闹——以至于树不得不抖动着灰色的枝条向她们大声地抗议。

就是在这样的朝夕相处中，我向湖袒露我所有的心声，也倾听她的心事和回应。我渐渐发现我与湖原来如此相似——我们都是偶然栖居在这颗蓝色星球上的、无比脆弱的生命。我越来越意识到湖和女人命运相连，作为“客体”、作为“资源”、作为待开发的“处女地”。当然，我们也都是危险的、不受驯的“野生生命”（荒野是the wild，而野女人/疯女人是the wild woman）。

于是，我开始思念、担忧、为她牵肠挂肚、为她的细微变化喜悦或伤怀，我知道我们将共享同一个明天。最重要的是，我开始感受那种依赖、依靠，以及内心深处深深的依恋。

我家门口的湖是独一无二的，只有在此处才能寻得。我知道我再难以离开她了——也就是在这一刻，我意识到自己已然爱上了这片湖和那些环湖而生的生命。其实我又何尝不是这样的生命？

在一个地方扎根就是对一方水土做出承诺，就是郑重地对她说：“我会与你建立相互滋养的关系。”湖孕育和供养了如此多的生灵，她倾听我的心事、用缱绻的低语伴我入眠，日日夜夜带给我无尽的爱和包容。而我也会在闲暇时到湖边小坐，花时间与她相处，了解她的脾性，感受她变幻莫测的心情。在寒冷的冬日，我会拿塑料袋到冰面上捡散落在她身上的易拉罐、玻璃碎片和饮料瓶。有时我能看到还有其她人在做同样的事——大多是女性——于是我知道不止我一个人深爱这片湖。

由于字数限制，本次来信只展示文章的前半部分内容。剩余内容将在下一封newsletter中发布～

就此搁笔，期待下一次和大家见面！

陌生女人1号

二〇二五年四月二十三日

Lily is representing ChinaTalk at ICLR this week! Send us a message if you’re going and would like to meet up.

Liberation Day was just the beginning — Trump declared that he will be, “Taking a look at semiconductors and the whole electronic supply chain.” What could happen next and what would be the smart way to go about this?

Welcome back to another CSIS Chip Chat. Joining us today is CSIS’ Bill Reinsch, a longtime Washington vet who co-hosts The Trade Guys podcast, as well as Jay Goldberg of the Digits to Dollars blog and The Circuit, an excellent semiconductor news podcast.

We discuss…

• The conflicting goals of Trump’s tariff strategy,

• What chip designers stand to lose in a worst-case scenario,

• Why tariffs could undermine the enforcement of SME controls,

• How Chinese chip companies benefit from American tariffs,

• Loopholes in the H20 ban, and how the administration could close them.

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

Tariff Anxiety and Trump’s Objectives

Jordan Schneider: We’re recording this interview on Friday, April 18th. What are we looking at with the tariff situation right now?

Bill Reinsch: Tariffs are on the way. That includes tariffs not just on semiconductors, but on a wide range of downstream products. That could be very disruptive to the economy, depending on the specifics. If it has a chip in it, it may be vulnerable. Thanks to the Internet of Things, that’s a lot of stuff — it’s toasters and refrigerators as well as laptops and phones.

The mystery is what he’s trying to accomplish because he’s given out conflicting goals.

One goal is revenge — getting even with the countries that have taken us to the cleaners for the last 50 years.

Another goal is to use tariffs as a negotiating tool to get other countries to remove trade barriers. It’s artfully designed such that we don’t have to do anything. All we have to do is wait for them to make concessions, and then our concession will be not to impose the tariffs that he’s announced. If that works, that’s a good deal.

The third goal, which is inconsistent with the first two, is revenue. They need money to pay for tax cuts. Our current estimate of tariff revenue is $330 billion a year, which is not peanuts. But that’s only true if the tariffs stick. If he negotiates them away, then the revenue goes away.

The fourth goal, which is the most interesting one, is to reshore manufacturing. If you make it here, there’s no tariff. But there’s a time gap — tariffs go into effect now. If you’re building a factory, it takes lots of money and lots of years to put the pieces together. You have to find workers to fill the jobs.

Which is the real goal? It changes from time to time. With chips, it’s particularly hard to tell. It’s not really a revenge issue because most of this stuff is covered by the Information Technology Agreement, so it’s zero tariffs already. It’s probably trying to force manufacturing here, but in the process, it’s going to be one of the largest revenue transfers from the poor to the rich in history because the poor pay the tariffs disproportionately. The tax cuts also go to the rich disproportionately. This will make economic inequality in the United States a lot worse than it already is.

Jordan Schneider: If you’re a major chip designer, what’s going through your mind at the moment?

Jay Goldberg: If you’re a major chip designer, you have two concerns. One is losing access to the China market, because if China’s going to use counter-tariffs, suddenly your products are much more expensive. The other concern is tariffs on Taiwan. I know there are all kinds of side agreements and other things going on here where some chips are coming in at zero tariff, but is that going to continue? It sounds like there are threats to impose more tariffs on Taiwan for various reasons.

Subscribe now

Not all chips are covered. GPUs aren’t covered under the rules — GPUs actually get tariffed. Semianalysis put out a good piece last week showing that there’s a loophole. If you import your tariffs through Mexico as part of a system, then you don’t get that tariff. But if you import the GPUs directly to the US, they’re tariffed differently than CPUs, which hurts my brain and just seems counterproductive. Do you want to do assembly in the US? Well, you’ve just made that more expensive than doing assembly in Mexico. Do you want to win the AI war, whatever that is? You’ve just made that whole process more cumbersome and more expensive.

I’m confused — that’s where I end up on it. But to answer your question, if I’m running Qualcomm or Nvidia or Broadcom, I’m worried that suddenly my products are going to be more expensive somewhere along the line and that I’m going to lose access to 10-20% of my revenue that goes through China.

Jordan Schneider: All right, let’s discuss manufacturers and manufacturing.

Jay Goldberg: Regarding Bill’s question about our goals — do we want to bring advanced manufacturing and semiconductor manufacturing back to the US? If that’s the objective, we’ve already achieved it. We have TSMC building advanced chips in Arizona, so the supply chain is secured. That plant is very dependent on Taiwan’s mothership for R&D.

If you want an American company to have advanced manufacturing of semiconductors in the US, then the solution is simple — give Intel $50 billion. Problem solved. Intel can manufacture chips — it’s not a technical problem anymore, but an economic one. They just need to overcome certain cost curves to manufacture here.

If this were purely a national security concern and we wanted a fully domestic supply chain in case of war, then investing heavily in Intel would accomplish that. However, this approach creates other issues since we generally don’t favor subsidies. The ultimate goal remains confusing to me.

Bill Reinsch: There’s an important distinction you’re making. Biden was more of a carrot guy while Trump is a stick guy. Biden’s approach was to incentivize companies to do what we want. That’s essentially what the CHIPS Act was about, and likely much of what the IRA addressed — tax credits, subsidies, and incentives.

Trump takes the opposite approach. He threatens and pressures companies into compliance. It’s much cheaper for the federal government but not necessarily cheaper for the companies, as you’ve pointed out.

Jay Goldberg: In a modest attempt at bipartisanship, I’ll acknowledge that the Biden administration wasn’t particularly clear about its semiconductor objectives either. Their semiconductor initiatives contained many conflicting goals.

Bill Reinsch: I wouldn’t disagree with you on that.

Jay Goldberg: The current process is quite confused. Our objectives aren’t clear, and most paths forward involve everything becoming more expensive for average consumers. Then we face the question of China’s response, as they have some serious counter-tariffs that are detrimental to all these companies.

Bill Reinsch: The tariff levels are so high that they’ve essentially prohibited most trade between the two countries if they remain in place. My footnote to your comment is that Trump wants a deal and believes he can negotiate one with his “good friend” Xi Jinping.

This causes some concern within the administration because he appointed several people who are “decouplers” — they don’t want anything to do with China. But Trump maintains that he can make a deal. When he says that, as he has a number of times, I think back to 2018 and 2019. We’ve seen this movie before — he negotiated and made a deal, but in my view, he got played.

He entered with a whole bunch of demands and ultimately settled for promises to purchase American goods, which China never fulfilled. When questioned at the press conference about all his initial demands — eliminating subsidies, stopping IP theft, and other legitimate concerns — he responded that those would be addressed in “phase two,” which never materialized. I believe we’ll see a replay of this scenario.

Jay Goldberg: Then they didn’t follow through.

Bill Reinsch: Right now, it resembles a sumo match. You have two enormous opponents in the ring doing a lot of foot-stomping, glaring, and frowning at each other while throwing rice around. Eventually, there will be contact, and they’ll get to the negotiating table.

The only challenge is disguising who made the first call, because neither wants to be seen as the initiator — they both want to appear as the gracious party. An announcement will simply state that the two leaders have spoken and a meeting is scheduled, without mentioning who reached out first.

After the meeting, I expect similar results to before. Perhaps not 145% tariffs on Chinese goods or 125% on American goods, but some smaller figures. Currently, I don’t see any scenario where these numbers drop below 10%, which represents a significant increase for the US, whose average tariff level for decades has been around 2-2.5%.

Jordan Schneider: I’m somewhat disappointed. There’s a prediction market on whether Trump will impose 200% tariffs on China before June, and we’re now down to 19% probability. It spiked to 60% at one point. If we’re going this high, why not go for 1,000%? How about 949%? Let’s really make a statement.

Can China Capitalize?

Jordan Schneider: Let’s take a more positive approach. Bill and Jay, how would you use tariffs to influence the semiconductor industry?

Jay Goldberg: A few weeks ago, the Chinese establishment appeared very nervous. Obviously, I have no insight into what Xi Jinping or senior party members are thinking, but from the semiconductor industry commentary I was reading, there was genuine fear and trepidation.

That shifted the moment the tariffs were extended to the rest of the world. Liberation Day was a positive development from the Chinese perspective. If Trump had maintained his focus solely on China, there was a moment when he could have genuinely affected the balance of trade.

Instead, he imposed tariffs on everyone else, alienating our allies. Then he reversed several tariffs, which undermined our negotiating position. There was a brief window when he actually had a chance at achieving his objectives.

Bill Reinsch: China should return to being worried because negotiations with other countries will likely unfold in a specific way. These nations are lining up to visit the US, and during these discussions, one of our major requests will be for them to match our approach toward China by imposing similar tariffs. While this isn’t specifically about semiconductors, it’s an important ask that will resonate with some countries.

Chinese overcapacity affects several sectors — steel, aluminum, solar panels, electric vehicles, and soon commercial aircraft. Other nations are beginning to recognize how this damages their domestic industries. The European Union’s concerns about EVs are shared by India, Turkey, South Africa, Brazil, Canada, and Indonesia, all of which are taking or considering tariff actions against China for various reasons. Indonesia’s focus is on textiles, which differs from the others.

Many countries have identified the problem and recognized that it’s detrimental to their interests. What’s missing is a coordinated approach. Unilateral action is like squeezing a balloon — the pressure just shifts elsewhere. The primary victim of our actions against China will be the European Union, as products no longer sold to us will be redirected there, which concerns them greatly.

If a collective strategy works, China will face a united front rather than just bilateral conflicts with individual countries, creating a much larger problem for them.

Jay Goldberg: That’s absolutely true. My concern is that Europeans no longer trust the United States to lead such a coalition effectively.

Bill Reinsch: You’re exactly right. We’re alienating our friends more than our enemies. Russia and Iran were exempted from the tariffs. While we don’t trade much with them anyway, it was an odd symbolic gesture. Given how we’re treating our allies, why would they help us? It simply doesn’t make sense, as we’ve already noted.

Jay Goldberg: I recently saw a report about ASML, which produces critical EUV tools essential for advanced semiconductor manufacturing. During the Biden administration, they provided software security keys to the NSA and US government to prevent China from using their most advanced systems. My understanding is that after the new tariffs were implemented, the Dutch government reclaimed those keys, moving us in the opposite direction.

You’re right about the common interests among nations. All the countries you mentioned, plus others, have legitimate concerns about China’s export practices. Eventually, we might align our approaches, but collective action is essential. Otherwise, we’ll remain fragmented.

This sentiment was reflected in Chinese media — initially, they were worried and felt isolated. Then “Liberation Day” arrived, and their tone changed overnight. They became optimistic, with some explicitly stating that the US had overplayed its hand, transforming the situation from “the world against China” to “the world plus China against the US."

Bill Reinsch: That’s a fascinating perspective. I hadn’t considered it that way.

Jay Goldberg: I’m trying not to sound alarmist. You’re right that countries are preparing to negotiate and seek agreements. Hopefully, some of these discussions will develop into collective action against China rather than just a series of bilateral arrangements.

Bill Reinsch: My instinct suggests that Japan and Korea will be the most reluctant participants in such collective efforts, which is particularly relevant for the semiconductor sector.

Jordan Schneider: Yes, the secondary impact on semiconductor export control. We saw the H20 affected earlier this week, but that’s relatively straightforward. Manufacturing equipment requires cooperation from allied nations. While extraterritorial measures might allow for tighter restrictions than previously possible, I have mixed thoughts about this situation.

These countries will be more concerned about broader trade disruptions, making Tokyo Electron’s challenges seem minor compared to potential 20% tariffs on their entire economies. However, this approach creates an unfriendly atmosphere when making additional requests. If these nations are reconsidering their relationships with China, semiconductor restrictions are precisely what will frustrate them most.

Bill Reinsch: China has launched what appears to be its third charm offensive. The United States is essentially signaling to the world that we’re no longer a reliable partner, which has profound implications.

One particular concern, magnified by our implementation approach, is the lack of consistency in our policies. Tariffs are imposed, removed, increased, and decreased; exceptions are announced, then withdrawn, then postponed; deadlines shift, and deals emerge unexpectedly. From a foreign perspective, predicting our next move is impossible, and there’s little confidence that any agreement will endure.

Canada and Mexico provide prime examples. Trump negotiated the USMCA, which he declared the greatest trade agreement in history, yet five years later, he’s dismantling it. Under such circumstances, why would any nation trust our commitments?

Jordan Schneider: The Biden administration managed to bring the Netherlands, Japan, and South Korea on board to some extent, but their primary directive was maintaining good relationships with allies. This approach limited how aggressive they could be on semiconductor manufacturing equipment (SME) restrictions.

With Trump, we’re seeing different motivations at play. First, there’s less concern about allies’ reactions. Second, there’s a stronger desire to confront China. Third, there’s a simultaneous interest in making deals with China. These competing priorities could lead to various outcomes where allies become alienated.

As a result, companies like ASML might stop caring about finding ways to service DUV equipment or sell EUV technology to China. Alternatively, the administration might apply so much pressure through intellectual property restrictions that they successfully block these technologies, regardless of allied opinions — something the Biden administration was reluctant to do. This creates tremendous uncertainty for engineers at companies like SMIC who are trying to plan what technologies they’ll have access to over the next five years.

Jay Goldberg: One fascinating aspect is the competing factions within the Trump administration pulling in opposite directions on these issues. Some favor complete decoupling from China, while others — with Elon Musk being a prime example — remain dependent on Chinese operations.

Musk’s factory in Shanghai is central to his entire fortune. Where do his loyalties lie? He maintains influence in this administration, though perhaps it’s diminishing. Tesla relies heavily on that Shanghai gigafactory for global exports, which likely makes him more sympathetic to China.

From China’s perspective, there are approximately 56 electric vehicle companies competing with Tesla. Some, like BYD, Xpeng, and Li Auto, are beginning to surpass Tesla and would benefit from seeing it hindered or shut down. Strong voices in China oppose Tesla’s presence. If Musk can’t leverage his Washington influence to foster friendly US-China relations, what value does he provide to China? They could simply shut him down.

This situation gives me some hope that a peaceful resolution might be possible.

Bill Reinsch: Last year’s gossip suggested Chinese officials were optimistic because they believed Musk would advocate for their interests. However, it’s important to remember that everyone who has worked for Trump eventually gets sidelined — it’s just a matter of time.

When Trump was inaugurated, I predicted Musk would last six months, and I’m maintaining that prediction. His formal role might expire in May anyway, but the real question is when his influence will fade. Everyone eventually falls out of favor. There’s already speculation here about which cabinet member will be dismissed first, with various theories circulating. It’s become a parlor game.

Jordan Schneider: The one certainty we face is policy swings regarding the CHIPS Act implementation, export control regulations, and tariff structures. This level of uncertainty affects the future of initiatives begun during Trump’s first administration to revitalize domestic manufacturing.

Jay Goldberg: Some believe the tariffs are merely a negotiating tactic that will eventually lead to more reasonable levels without lasting damage. This perspective is misleading for two reasons. First, as Bill noted, we will undoubtedly end up with higher tariffs regardless of negotiations. Second, there are concerning secondary effects.

The implementation of tariffs and Chinese counter-tariffs has created a major opportunity for domestic Chinese chip design companies. Their strengths lie in trailing-edge manufacturing and analog products, with approximately 500 companies competing in this space.

Until very recently, the Chinese semiconductor industry was pessimistic — they faced excessive internal competition and geopolitical constraints. These new tariffs have made products from American companies like Texas Instruments and Analog Devices much more expensive in China, precisely when Chinese companies are releasing massive amounts of analog capacity.

This will propel several Chinese chip companies to positions where they can compete globally. Once they achieve global competitiveness — as we’ve seen many times before — they’ll disrupt foreign competitors due to their cost advantages and low capital costs. This will permanently alter the chip design landscape as Chinese champions emerge onto the global stage.

Without these tariffs, this transformation would likely have taken another five years to materialize.

Bill Reinsch: This is fundamentally a timing issue. When the October 2nd export controls were implemented, our team at CSIS identified three key questions — the effect on US company revenue, the impact on Chinese policy, and the potential “design-out” problem where companies would remove US technology to avoid control restrictions.

Regarding the second question about Chinese policy impact, your assessment is accurate. These controls didn’t change Chinese strategy, as they’ve been moving in this direction for the past decade. However, the controls have accelerated their timeline, forcing us to address these challenges much sooner than anticipated.

Jay Goldberg: The controls have pushed many sectors past the point of no return, which might not have happened otherwise. While this outcome was probable, things could have evolved differently. Now we’ll face permanent Chinese competition in several markets.

Jordan Schneider: Jay, what corporate and policy responses do you anticipate for the trailing-edge and analog sectors?

Jay Goldberg: The situation presents challenges. Companies like Texas Instruments have spent the past five years upgrading their capacity, investing approximately $10 billion to lower their cost structure, enabling them to compete with low-cost Chinese manufacturers. They’ll likely weather this storm.

However, other companies will struggle, particularly European firms like STMicroelectronics, Infineon, and NXP, as well as American companies like Analog Devices and ON Semiconductor. These organizations now face unprecedented Chinese competition in analog, memory, and several other categories. Previously, they could monitor Chinese competitors without serious concern, but now they suddenly confront formidable new rivals.

Jordan Schneider: Is there a solution through tariffs, subsidies, or industry consolidation?

Jay Goldberg: No easy answer exists. Tariffs aren’t effective because demand is primarily coming from China. The bright spot in analog chip companies’ revenues recently has been the Chinese electric vehicle market, and these companies will lose substantial portions of that business. Their growth prospects are diminishing, forcing them to lower their outlooks.

Additionally, these pressures clearly drive Chinese companies toward domestic solutions. To Bill’s point, this trend began around 2018. Chinese domestic chip consumption increases by roughly one percentage point annually. While overall market share remains in single digits, certain sectors show higher penetration.

This trend will accelerate over the next year. Once domestic competitors establish themselves, they secure more funding and cash flow, enabling increased R&D investment. This makes them more competitive and establishes them as permanent industry players in ways they wouldn’t have been otherwise.

Subscribe now

Jordan Schneider: Jay, let’s discuss the leading edge. How do you view Intel and TSMC in relation to Huawei’s efforts to vertically integrate the entire semiconductor ecosystem?

Jay Goldberg: Regarding leading-edge technology, China still has considerable ground to cover. Despite persistent rumors about major breakthroughs, they haven’t quite reached that level yet. Their progress will likely take several more years, though the tariffs have certainly energized their efforts and accelerated the process.

Huawei remains heavily dependent on TSMC, continuously finding new channels to access TSMC wafers in a constant game of whack-a-mole that will continue. Another complicating factor is the apparent reduction in US federal government resources. How can the Bureau of Industry and Security effectively enforce existing export rules with staff shortages while other government branches face similar cuts?

Huawei and other Chinese companies continually establish shell companies and reorganize their corporate structures. This creates a tremendously tricky enforcement environment.

What confused me most this week was Trump’s apparent focus on NVIDIA as an adversary. This seems contradictory — if our goal is winning the AI race, why target the American company leading in AI computing? Hopefully, this is merely negotiating posture, but it’s perplexing.

Bill Reinsch: I agree completely. With Trump, policy often becomes personal. If he dislikes someone, that sentiment transforms into policy, making outcomes difficult to predict.

Jordan Schneider: Visiting Beijing immediately after the H20 ban and meeting with Liang Wenfang seems like an unusual choice. NVIDIA’s future isn’t in China — the administration couldn’t make that clearer. It appears to be a questionable tactical decision.

Jay Goldberg: Jensen Huang even wore a suit and tie rather than his signature leather jacket.

Bill Reinsch: He’s attempting to balance competing interests — isn’t everyone, in one way or another?

Jordan Schneider: Trump could force him to choose sides. Depending on how you calculate it — particularly regarding Malaysia — China represents between 7-15% of NVIDIA’s revenue.

Bill Reinsch: This reminds me of the 2008-2009 financial crisis discussions about banks being “too big to fail.” In this context, are certain companies too important to punish? Trump may recognize NVIDIA as such an organization. His options are limited by the consequences — if winning the AI competition requires NVIDIA, damaging the company creates substantial collateral damage that must be considered.

Jay Goldberg: Precisely. From NVIDIA’s perspective, Jensen Huang must walk a tightrope. They’ve been subtly but consistently criticizing these policies since the Biden administration because they bear the brunt of the impact. NVIDIA and ASML feel these sanctions most acutely, with each new wave costing NVIDIA billions in revenue.

Huang has quietly resisted these measures, and his Beijing visit makes this resistance more visible. His concerns are justified — China has 20-30 GPU companies that, while not currently exceptional, are improving. Additionally, Huawei demonstrates impressive AI capabilities. If Huawei maintains access to TSMC or leverages what SMIC provides, they could become a serious threat to NVIDIA.

As with analog companies, creating a market vacuum inevitably leads to Chinese competitors filling the void. Once established, they become entrenched and impossible to dislodge. NVIDIA has legitimate concerns about this dynamic.

To Bill’s point, what options exist? Undermining America’s leading AI company through tariffs or other punitive measures seems risky, even for Trump.

Jordan Schneider: The truly frustrating aspect is how long the H20 ban took to implement and its failure to address the TSMC loophole and semiconductor manufacturing equipment vulnerabilities. We’ve created a worst-case scenario where we’re inadvertently helping Huawei on both fronts by allowing them to gradually improve and market chips that compete with NVIDIA’s offerings.

Huawei has developed an impressive server rack system that, while not equivalent to Blackwell, is approaching that capability. Their product is specifically tailored for the Chinese market, where power efficiency concerns differ from ours. Instead of receiving a nuanced, optimized policy, we get the lowest common denominator approach. Even with competing objectives, we’re far from achieving optimal results because the regulatory focus emphasizes simplicity and brevity rather than effectiveness.

Jay Goldberg: Let me offer a positive suggestion since you asked for policy recommendations. We should give the Bureau of Industry and Security and the Commerce Department a thousand inspectors to enforce the existing rules. Let’s deploy adequate personnel to investigate the Malaysian companies suddenly building massive data centers.

Jordan Schneider: We could reassign those IRS employees who are reportedly being let go.

Jay Goldberg: Exactly. Shift them over.

Bill Reinsch: The 2025 budget appears to feature either cuts or no notable increases. The only substantial BIS funding increase targets the import side to address Huawei hardware entering the United States. The export control and enforcement sides have received no expansion.

A group of us produced a paper on this issue a couple of years ago, with Greg Allen joining our advocacy for budget increases. We’ve met with congressional representatives, and while our message resonates and generates sympathy, implementation never materializes.

We’re entering an era of dramatic government reduction, as Jay mentioned earlier. Proposing a 25% personnel increase for one agency seems unlikely to succeed, despite its merit.

Jordan Schneider: Let’s conclude with this, Bill. You’ve been in this field for many years, and we’re at a point where thoughtful analysis seems undervalued — yet that’s our profession. Can you offer some encouragement for those of us committed to these issues?

Bill Reinsch: I wish I could, but I’m reminded of the adage, “The more things change, the more they stay the same.”

I served as the staff director of the Senate Steel Caucus for 17 years, largely trying to preserve the US steel industry. This wasn’t part of my portfolio when I joined the government, but more than 20 years later, I returned to it after leaving the National Foreign Trade Council and joining a law firm representing steel companies.

I encountered one of their lobbyists and commented that I had returned after two decades. His response was telling — he said that nothing had changed other than the people involved.

This dynamic applies here as well. For 50 years, it’s been a cat-and-mouse game, which will continue indefinitely. Enforcement officials always operate defensively. The greatest mistake is expecting 100% effectiveness, which is impossible. This requires viewing it as a management problem where you optimize results within constraints.

Consider that Iran’s nuclear development efforts required 100,000 centrifuges. If your enforcement goal is zero, you’re destined to fail. If instead, you aim to limit them to 25,000 units that cost five times more and extend their timeline from three to ten years, those objectives are achievable. However, this pragmatic approach fails to satisfy political demands.

Congressional China hawks want absolute results, like seeing Huawei eliminated, banning all US-origin chips from entering China, or having ASML cease all Chinese operations. Every administration confronts this perpetually present faction that views China as an existential threat.

During the Bush administration, I had a revealing conversation with an NSC official handling this portfolio. While discussing another matter, he abruptly stated that a million Chinese wake up daily thinking about harming Americans. I responded that several Pentagon personnel similarly contemplate targeting Chinese forces — that’s their job. Without hesitation, he replied, “Not nearly enough."

These hardliners exist throughout Congress, administrations of both parties, think tanks, academia, and media. They’re currently more influential than previously. My perspective has been that the optimal policy for both nations is preventing such individuals from assuming control. I’m concerned because several now hold administration positions.

One current discussion topic is whether to implement a complete embargo, which would add little practical impact beyond existing measures but carries symbolic weight.

My outlook is pessimistic, though I hope for more measured approaches. The semiconductor industry has effectively communicated the consequences of excessive controls over the years. However, they achieved far less than they sought during the Biden administration and likely won’t fare better under the current one. Nevertheless, I commend their persistent efforts.

Jordan Schneider: Earlier, we mentioned that Bill’s son teaches religion at a Catholic school. I had AI generate a transcript and asked o3 for Bible quotes relevant to our discussion about BIS export controls. Bill, could you pick your favorite from the list?

Bill Reinsch: “Nothing is concealed that will not be disclosed, nor hidden that will not be made known.” Luke 12:2.

Jordan Schneider: My favorite was: “No wisdom, no understanding, no counsel can avail against the Lord. The horse is made ready for the day of battle, but victory belongs to the Lord.” So we’ll leave export control effectiveness in divine hands.

Bill Reinsch: My son would agree with that.

Jay Goldberg: “Plans fail for lack of counsel, but with many advisors, they succeed.” Proverbs 15:22. I’ll choose that one.

Mood Music

Dylan Patel and Doug O’Laughlin of SemiAnalysis explore Huawei new rack and its broader strategy of leveraging structural advantages to resist, circumvent, and nullify export controls — as well as how the U.S. government should respond.

We also get into chip tariff policy. Semiconductors are exempt from Trump’s reciprocal tariffs for now, but the administration has signaled that chip-specific tariffs are coming soon. Dylan gives his pitch for how tariffs could be used to save American manufacturing, move high-tech supply chains out of China, and even reduce the national debt!

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

Huawei’s Edge and What to Do About It

Jordan Schneider: Dylan, tell us about what's going on with Huawei.

Dylan Patel: Huawei has the Ascend 910B and 910C. It's the same chip silicon-wise, but we've been banging on the drums about this for years now. Systems matter more than individual chips for AI workloads. When we look at Nvidia, one of their three core competencies is networking and systems, another is the chip itself. These are almost equivalent-level components.

What Huawei has done is taken a 7-nanometer chip that is mostly produced at TSMC, some at SMIC, but mostly at TSMC, and put it into a system that consumes more power with more chips, but they've networked it together in such a way — they've cooled it, they've done all the networking — where the performance is actually going to be better than Nvidia's GB200 NVL72 rack. This is the new product that Nvidia is shipping right now in mass production. It's what everyone wants. To be clear, it's really good.

Nvidia's solution is one rack that consumes 140 kilowatts. Huawei has made a system that consists of 16 racks of compute connected all together through optical fiber, using approximately 6,000 optical transceivers connecting the Ascends together. Yes, it consumes 550kW versus 140kW for Nvidia. This is a huge step up in power, but it also brings 2-3x the performance in memory and FLOPS.

When we talk about software to utilize it, understand that Huawei's software is better than AMD's. They support JAX, Python, PyTorch as first-class citizens, and support VLMs. There's a lot of software support that they've already built out publicly. Huawei's engineers are exceptional — their software engineers are arguably just as good as Nvidia's. You can see this in all the products they've built over the years.

Huawei has turned what is effectively a big handicap on chips into a workable solution. They'll use 2-3x the power to achieve 2-3x the performance, but that's acceptable because China can build power infrastructure. The US doesn't build power infrastructure as readily.

This is critical to consider — Huawei continues to receive HBM through this Faraday-CoAsia loophole:

Samsung has come to the rescue, having been the number one supplier of HBM to China through which Huawei has been able to stockpile a total of 13 million HBM stacks which can be used for 1.6 million Ascend 910C packages before any HBM bans.

Furthermore, this banned HBM is still being re-exported to China. The HBM export ban is specifically for raw HBM packages. Chips with HBM can still be shipped as long as they don’t exceed the FLOPS regulations. CoAsia Electronics is the sole distributor of HBM for Samsung in Greater China and they have been shipping HBM2E that is to ASIC design service company Faraday who gets SPIL to “package” it alongside of a cheap 16nm logic die.

Faraday then ships this system in package to China, which is technically allowed, but Chinese companies can then recover the HBM by desoldering. We think they employ techniques to make it very easy for the HBM to be extracted from the package, like using very weak low-temperature solder bumps, so when we say it is “packaged,” we mean this in the loosest way possible.

They've stockpiled a significant amount because the US government telegraphed the bans on HBM for nine months instead of implementing them immediately.

They continue to receive wafers. Allegedly — I haven't been able to validate this personally — Huawei received over $500 million of wafers just for the Ascend alone last year. They could potentially still be receiving wafers from TSMC. Furthermore, as SMIC continues to improve their yields — even if they only reach a paltry 20% — they could produce millions of Ascend chips because we continue to sell leading-edge equipment to SMIC to produce wafers. We continue to sell leading-edge equipment to CXMT to produce wafers. Yes, there are some restrictions against them, but these restrictions have loopholes.

Given all these factors, Huawei will be able to ramp production unless the US government improves its strategy. They should have focused their attention on banning the equipment before addressing the H20. Now they're just taking revenue away from Nvidia while still allowing China to import advanced technology. ASML recently stated that China is buying a lot more high-end equipment than expected.

There are significant issues with the government's strategy. They're doing things backwards. Credit to them for banning the H20 — perhaps that was necessary — but they got the order of operations completely wrong, and many loopholes were left by the Biden administration.

The Trump administration now has the chance to be tough on China in ways the Biden administration claimed to be but wasn't fully committed to, though they haven't taken decisive action yet. We'll see if the Trump administration decides to be genuinely tough on China. Their rhetoric is strong but their actions have been weak so far. They recently laid off one of my favorite people in the NSC because Laura Loomer claimed he was a "deep state agent." This is not a joke — this is serious. There are many concerning decisions being made, and I want the administration to do better. I'm happy to advise them on what they need to do.

Doug O'Laughlin: This is not just ASML. All of the suppliers have been saying there's been incremental pull-ins on demand for semiconductor capital equipment specifically in China.

Dylan Patel: It's equipment, subsystems, and especially metrology equipment, which is barely controlled and completely a US industry. There are chemicals that could be used to cut off SMIC or the Huawei-associated fabs, but they won't do that.

Doug O'Laughlin: People don't fully appreciate that Huawei is vertically integrated. Huawei is dominating the Chinese domestic ecosystem. They steal NARA and AMEC tools as well. There's a tool they had that's identical to a DISCO tool we saw, and there's an identical KLA tool. The equipment manufacturer SiCarrier is also functionally an arm of Huawei.

Dylan Patel: Let's be clear about SiCarrier and Huawei. Huawei bought about $8 billion of equipment last year. They're simultaneously running wafers through their tool and imported tools side by side and calibrating their tool to improve. They're continuously getting better and better. It's amazing how well they're doing this.

Doug O'Laughlin: What's crazy is that Huawei and Microsoft have the same number of employees.

Dylan Patel: Pound for pound, Huawei is so much better than Microsoft.

Doug O'Laughlin: Microsoft does make a lot more money, to be clear. But Huawei is massive, and they're doing hardware, software, and multiple layers of hardware networking. What is Huawei not doing?

Dylan Patel: Huawei is even doing electric vehicles, which are fire. (YouTube review, but bilibili spocon is way more fun).

Doug O'Laughlin: Jensen Huang has said before: Huawei is “single most formidable technology company in China.”

Jordan Schneider: Can you explain the vertical integration and what SiCarrier is aiming to become?

Doug O'Laughlin: SiCarrier makes tools. They're like Applied Materials, ASML, and KLA. But Huawei and SMIC are heavily involved.

Dylan Patel: SiCarrier has already bought about $2.5 billion worth of equipment for an HBM fab that they're building for SMIC/Huawei. They’ve built leading-edge 7 nm and 5 nm fabbing tools. They have a new form of non-volatile memory which should beat DRAM in cost and NAND in performance. There's an advanced packaging Huawei fab. The US government doesn't consider any of that to be part of Huawei. But if you use basic logic, they're clearly Huawei.

Jordan Schneider: Anything else on Huawei?

Dylan Patel: It's important to recognize what they did on the networking side as well, which was really impressive. They built their own switches. They built their own NVLink equivalent, which is impressive on its own. But then on top of that, they built their own optical transceivers.

Every single optical transceiver that goes into a Google TPU pod is made in China by a Chinese company. Approximately 70% of optical transceivers that Nvidia uses are designed in China and shipped to all these servers. Most of them are built in China as well. There's a significant optical supply chain issue that America has that needs addressing immediately.

China has their own optical supply chain, and it's remarkable because they've become the first to productize this technology called LPO. Doug and I have been discussing this for exactly two years now. It essentially involves taking an optical transceiver and removing the most expensive chip while still making it work. It's cheaper, lower power, and technically lower latency. The challenge is that it's somewhat less reliable, but with good engineering, you can solve that problem. Huawei figured it out before anyone else.

Optimizing Tariffs

Jordan Schneider: What’s going on with our semiconductor tariff situation?

Dylan Patel: If they are intelligently designed, the semiconductor tariffs could do a lot of positive things. But they could also do a lot of negative things.

Jordan Schneider: How would you design semiconductor tariffs?

Dylan Patel: First, if you import a subcomponent into the US and then export it out of the country, you should be able to get a refund on the tariff you paid on that subcomponent. This shouldn't be limited to just semiconductors either — we should have a policy that you can get import credits for anything you export out of the country. That fixes a lot of the tariff issues. This is done by some countries, but not by America.

Second — and let's frame this as an electronics tariff, not just a semiconductor tariff — any sub-materials or subcomponents should be excluded from the tariff initially. Take consumables, for example. We don't make wafer sputter targets in America; everyone just imports them. It would be nice if we made some in America, but we don't want to increase costs for American factories that rely on those imports.

Instead, we should implement a phased-in tariff. If it takes two years to build a sputter target factory in America, we should announce that a tariff on sputter targets will go into effect in two and a half years. That way, no producers get obliterated. This needs to be communicated very clearly so that everyone knows it's coming. We should do this for as many subcomponents as possible.

Next, the tariffs on final assembled goods should be much higher than the tariffs on sub-components and sub-assemblies. We should impose heavy tariffs on iPhones even if it's politically unpopular. Instead of excluding them, iPhones should be hit with a 135% tariff if they're made in China or a 25% tariff if they're made in another country. But the sub-components, such as the display from South Korea, should either not be tariffed or be tariffed at a very low rate. That creates a waterfall effect where it becomes profitable to do the assembly in America or USMCA countries.

Finally, we should reclassify goods. Right now, motors and robotics are in the same classification. That means no one is going to manufacture robotics in America. Under the current policy, if you import a motor in order to make a robot for export, it's not considered a “substantial transformation,” and thus you can't get any exemptions. Whereas if you import a chip and then use it to make a computer or a PCB board, it's a substantial transformation.

We should definitely impose steep tariffs on China and not tariff other countries nearly as much, because many of these tariffs should work by moving supply chains out of China and into Vietnam, Indonesia, Malaysia, India, Mexico, Canada, as well as the U.S.

Share

Jordan Schneider: Why should we want assembly to happen in America?

Dylan Patel: Assembly is actually not that labor-intensive anymore. When we talk about automation, we're never going to get automated factories in America if the factory doesn't exist in America in the first place. When they move the factory to America, they're going to automate it much more.We need assembly in places other than China. Specifically, we need assembly in America because the process becomes less labor-intensive and more capital-intensive over time, and America should not be at a deficit when it comes to capital intensity.

We can increase our tax base massively if we have all of the highly capital-intensive activities happen in America. We need to have that small set of skilled labor operating massively capital-intensive factories.

There shouldn't be a natural advantage for other countries versus America in these areas, but there is because we don't have any supply chains here. The main thing we need to do is import 100,000 Taiwanese citizens into America and build every fab in America. We need to import around 100,000 Korean people and build memory in America.

If you look at the labor cost of an Intel fab or a TSMC fab, it's minuscule. For a gigafab — a fab that makes over 100,000 wafers a month — there are only a handful of those in the world. Power costs just as much as the people, and power costs twice as much in Taiwan. For all intents and purposes, a fab should actually be cheaper in America. But we have supply chain issues with gas delivery, tool delivery, tool installations, and all these other things.

A hundred-billion-dollar fab has about 10,000 people working in it. That's nothing. It's not about the jobs — it's the fact that it's here in America and it's now contributing to the tax base. Fabs are an example of ridiculous capital intensity for very low labor. That labor is intelligent and very skilled, but we can just import them.

We should fix the trade deficit by having a larger talent deficit. These highly capital-intensive aspects of society are really something America should have, and we don't. We should be building $100 billion factories like TSMC is in America, eventually, hopefully.

The same should happen across all these supply chains. Now fabs are extreme, but assembly is going to get there. A lot of manufacturing is going to get there because robotics is really hitting its golden age. There's a lot that robotics and AI are going to do that makes things even more capital-intensive and even less labor-intensive.

This is better on audio. Listen now on iTunes, Spotify, or your favorite podcast app.

Mood Music

Read more

今天说一下中文世界流行的两种思维方式：非此即彼和爱走捷径。这两种思维，听起来很简单，但它们正在悄悄塑造我们看世界、理解制度、判断历史的方式，而且，严重限制了我们对“现代文明”的真正理解。

什么意思？你只要说点什么，马上就有人把你贴标签。比如你说提高民众的精神素养很重要，他立马说你是不是觉得制度不重要；你说川普有问题，他立马说你是民主党或者共产党；你说民主党有些政策缺乏常识，他就说你一定是川粉。

这就是典型的“名词概念奴隶”——他们不听你说了什么，他们只听你用了哪个关键词，然后决定你是敌是友，非黑即白。

第二种，就是我们太喜欢走捷径。

中文世界很多人看着日本、韩国、台湾的发展，心里想的是：哎呀，他们不就是先走一步发达了，早一点现代化了吗？我们直接跳过那些复杂的制度建设、精神文明、社会文化演化，直接“弯道超车”，一脚油门干进发达国家？

但现实是，任何一个成功跻身发达国家的后发国家，都经历了一整套“现代文明化”的进程。这不只是高楼大厦、5G、量子科技那种“物质层面”的现代化，还包括精神世界、政治制度、社会行为方式的深层转型。

可惜，中国过去几代人，热衷的是投机取巧，只要GDP、高铁和房地产，再顺手牵羊搬运点发达国家的技术，别的都不管。结果就是，物质现代化上去了，精神文明被“阉割”了，制度建设停滞了，社会行为变得越来越“功利而粗暴”，从企业到个人，一直内卷到卷不动了。

也可以说，这其实就是“只要现代化，不要文明化”。于是我们看到，成为世界第二大经济体，成了这个模式的巅峰；但从巅峰下来，也只是这两年刚刚开始。

甚至连很多向往民主的人，也幻想能靠捷径“飞升”。他们看着日本、韩国、台湾的民主发展，就想哪天中国也能“天降民主制度”，一下子实现民主转型。

这几年，中文圈还流行一个新说法，说日本能变成民主国家，是因为麦克阿瑟帮了一把，说明治维新其实失败了。

听起来很有戏剧性，但这其实是一种新的“救世主幻想”——换个角度讲，天降伟人、系统外挂的升级版。

如果连明治维新都叫失败，那恐怕人类历史上没有哪个后发国家的改革算得上成功了。说这种话的人，往往既缺乏对日本近代史的了解，也反映出他们根深蒂固的心理——既想现代文明，又不想走那条艰难的文明之路。

日本现代文明化的先驱，福泽谕吉早就对这种人做了精准描述。他说：“没有独立精神的人，一定依赖别人；依赖别人的人一定怕人；怕人的人就会阿谀奉承；时间久了，就不知羞耻，面对不公不敢发声，面对强权只知道屈服。这种人，如果释迦牟尼和孔夫子再世，也没救。”

这是福泽谕吉《《劝学篇》中的话。《劝学篇》虽然是本只有几十页的小册子，却塑造了现在日本人精神世界。在明治时代，这本小册子可以说是颗思想炸弹，它改变了整个日本对“学习”、对“文明”、对“独立”的理解。当然中文世界不乏把思想炸弹，读成心灵鸡汤的人。浅薄无聊的人可以把任何思想读成心灵鸡汤。

《劝学篇》有句话：“天不生人上之人，也不生人下之人。”

听着是不是很耳熟？这不就是“人人生而平等”？没错，福泽谕吉正是把西方“个人独立”“平等”的观念，用当时日本人能听懂的话翻译出来。他说：人的贵贱不是出身决定的，而是靠“学习”和“自立”获得的，学习上进，就会变得高贵，心灵封闭，不学习，结果就是低贱。

在福泽看来，一个人要获得真正的独立，不靠祖上、不靠政府、不靠社会的恩赐，只能靠自己的知识、能力和判断力。

他举了很多例子，比如说，有些人虽然穿得光鲜亮丽、出身高贵，家里有钱，但完全没有自己的主见，只会随波逐流。这样的人，哪怕有地位，有钱，其实也不自由，是“精神的奴隶”。不是高贵，而是低贱。

而一个真正有知识、有判断力的人，真正的现代文明人，即使过普通人的生活，甚至贫寒，也能独立、自信地活着。

“天不生人上之人，也不生人下之人。”这可以说是日本的独立宣言，不是国家的独立，而是国民人格的独立。

福泽早在150年前就明白一件事：如果一个人、一个民族，没有独立思考的能力，只是像工蚁一样工作，靠别人的市场卖货，就算物质再富裕，也终究离现代文明很远。

福泽的高明之处就在于，他不仅看到表面的政治问题，而是看到更深层的人格塑造问题。读书、明理、学以致用，最终是要落实在你身上的。如果你没有福泽说的那种“独立自尊”的人格，哪怕真来了个麦克阿瑟，都没法救你。

孟子也讲过，普通人总是“待文王而后兴”——意思是，总等着遇到伟大时代，有人来带我起飞。但如果你没有那个内在的独立人格，就算文王真的来了，也带不动你。

说到底，只要中国社会想往好一点的方向走，国民建立独立人格，是怎么绕都绕不过过去的一道坎。

美国有个很有洞见、但英年早逝的媒体人 Andrew Breitbart 曾说过：“政治是文化的下游。”

什么意思？你这个社会的精神状态，决定你最终会选择什么样的制度。精神世界腐烂的人群，弄不出有质量的民主制度，即便天降民主制度，也会把它玩成烂泥一滩的“劣质民主”；如果继续走独裁路线，那也是选那种最武断、最低级、最任性的版本——不是说服你，而是“镇住你”、“压服你”。

很多年前，我跟朋友聊过一句话——中国的当务之急，不是实现民主——那里没有民主的社会和人群土壤，而是避免独裁劣质化。那时候他们都觉得我说得太悲观。但那很快就变成了现实——不好的独裁被更坏的独裁取代。

中国现代史就是一出不断重复的舞台剧：取代坏的往往不是好，而是更坏。说的更具体一点。别说实现民主了，就连“质量说得过去的独裁”都保不住，通常撑不过一代人就退化了。

中国也不是没人争取民主，努力的人不少。但现实是，一百多年过去了，民主没有实现。反观周边，日本、韩国、台湾，甚至蒙古，都不同程度实现了民主转型。

这是不是说明，中国可能真的缺乏“民主的土壤”？这是个很沉重，但也必须直面的事实。

如果你觉得，民主比人生中其他任何事情都重要，那你就去争取。但那是一项水滴石穿的事业，可能只有耕耘，看不到收获。

如果你觉得，过日子更重要，也无妨。那就继续原地生活，随国家的大船在风浪中起伏，知其不可奈何而安之若命。古人也有过这样的选择，那也是一种人生。

但不论选择哪种生活，要做个现代文明人，独立的人格、清晰的思维，都是必选项。政治可以烂污，但人格不能烂污。

Ray Wang is a Washington-based analyst formerly based in Taipei and Seoul. He focuses on U.S.-China economic and technological statecraft, Chinese foreign policy, and the semiconductor and AI industry in China, South Korea, and Taiwan. You can read more of his writing on his Substack: SemiPractice or @raywang2.

Key Takeaways

• Nvidia’s H20 GPU with HBM3 had become the most sought-after accelerator in China amid rapidly increasing inference and computing demand. Prior to the new restrictions, shipments were projected to reach 1.4 million units in 2025.

• CXMT is now only 3-4 years behind global leaders in high-bandwidth memory (HBM) development, aiming to produce HBM3 in 2026 and HBM3E in 2027 amid notable technological improvements in DRAM.

• CXMT still faces major roadblocks — these include U.S. export controls on lithography and other equipment, a volatile geopolitical environment, limited access to global markets, and the uncertain pace of technological development against market leaders.

Greater Demand for HBM and Nvidia’s H20

In December 2024, the U.S. released new export control packages targeting Chinese access to high-bandwidth memory, or HBM, and various types of semiconductor manufacturing equipment, including tools essential for HBM manufacturing and packing. The new rule also added over 140 Chinese chip manufacturers and chip toolmakers to the Commerce Department’s Entity List.

The new rule was designed to further constrain China’s AI development by leveraging the chokepoint on HBM, ultimately controlled by three companies around the world — SK Hynix, Samsung, and Micron. The restriction around SMEs, on the other hand, aims to limit China’s ability to develop its own HBM.

HBM powers almost all of the AI accelerators that train large language models. It has become even more important since the rise of reasoning models and inference training, where memory bandwidth and capacity play a vital role.

The Chinese AI accelerators are no exception when it comes to reliance on HBM. For example, Huawei’s latest AI accelerators, the Ascend 910B and upcoming 910C GPUs, are mainly equipped with 4 and 8 HBM2E, respectively, mostly sourced from Samsung before the December 2024 restriction went into affect, with some sourced after the restriction. Similarly, other Chinese GPU makers such as Biren, Enflame, and VastaiTech are likely using HBM2 and HBM2E from either SK Hynix or Samsung. Biren’s BR100 GPUs that launched in 2022 incorporate 4 HBM2E, and Enflame’s DTU released in late 2021 uses 2 HBM2.

Amid the rapid development of advanced reasoning models in China — including DeepSeek R1, Alibaba’s QwQ-32B, Baidu’s Ernie X1, Tencent’s Hunyuan T1, and ByteDance’s Doubao 1.5 — China’s demand for inference training and overall compute has been accelerating. This trend began as early as late January, and in turn is driving demand in the Chinese market for accessible GPUs with the most advanced HBM, namely the Nvidia H20.

The H20 was by far the most advanced and accessible foreign GPU available to the Chinese market. While Huawei’s Ascend 910B — the domestic alternative — offers performance roughly on par with the Nvidia A100, it delivers only about 40% of the H20’s performance in a cluster configuration. Its latest GPU Ascend 910C, is competitive in both computational and memory performance. While 910C’s scale of adoption in China’s AI industry remains unclear, it is gaining more attention. Beyond raw metrics, Chinese firms continue to prefer Nvidia hardware for now due to their engineers’ deep familiarity with its mature and widely adopted software ecosystem, as well as its superior reliability and efficiency in large-scale cluster environments backed by reliable supply chain partners.

While the performance of the H20 is about 6.7 times less powerful compared to Nvidia’s flagship H100 in terms of computational performance, it does provide larger memory bandwidth and capacity. Thus, the H20 was preferred for inference training over the H100, a key factor driving its demand in China.

Multiple media sources have recently reported the rising demand for Nvidia’s H20, and several supply chain checks by the author confirmed this trend as early as the beginning of February. That marks a shift from my earlier supply chain check based on data from mid-January, which suggested a significant decline for H20 orders likely due to concern over a potential ban on the H20. We should also not rule out the possibility that the fast-increasing demand for H20 was driven by the looming concerns of H20 restrictions since as early as the late Biden administration.

According my calculation in early March through downstream supply chain checks (advanced packaging and chip testing), Nvidia was on pace to ship roughly 1.4 million H20s in 2025. China could have obtained about 600,000 H20s by June based on a separate calculation by the author. Notably, since the H20 started shipping in Q2 2024, China has obtained more than 1 million units.

Subscribe now

The rising demand for inference and computing power also aligns with the projected annual spending of China’s top cloud service providers. Companies like Tencent, Baidu, and Alibaba are expected to spend a combined $31.98 billion in 2025—a nearly 40% increase from 2024.

Nonetheless, the H20 was officially banned effective last week. According to Nvidia’s recent 8-K filing to the SEC, the company states that the U.S. government has informed the company that selling the H20 — or any chip matching its memory bandwidth, interconnect bandwidth, or both — requires a license. In other words, Nvidia will not be able to sell H20 or any more powerful GPUs to China as it is unlikely to obtain a license for the Chinese market.

Chinese Memory Advancements

The surge of demand for Nvidia’s H20, the critical role of HBM, and the growing emphasis on both reasoning models and inference training all ultimately point to HBM’s strategic importance to China’s AI sector.

Certainly, the Chinese government and industry are well aware of the importance of HBM in the midst of an increasingly unfavorable regulatory and geopolitical environment around computing resources and associated AI hardware. For China, there is a strategic urgency and necessity to develop its indigenous HBM to shake off the existing restrictions from the United States and enhance their capability for developing AI.

Knowing HBM’s strategic role in China’s AI ambitions amid escalating U.S.-China tech competition, it is essential to map China’s HBM development as precisely as current data allows.

My best guess is China’s HBM development trails market leaders by roughly only four years amid increasing export controls — a narrower gap than previously estimated or widely expected, despite ongoing challenges and uncertainties.

Share

In the second half of 2024, reports indicated that CXMT — China’s leading DRAM and memory maker — has begun mass production of HBM2, placing it roughly three generations behind market leaders, which have been supplying HBM3E 8hi and 12hi (the most cutting-edge HBM in the market to date) to leading AI chip vendors such as Nvidia, AMD, Google, and AWS since 2024. My retrospective analysis of the HBM roadmaps of these four companies in December suggested that CXMT lags behind the market leaders by approximately six to eight years, with various challenges to overcome.

A six-to-eight-year lead — a rather reassuring gap for policymakers in Washington and industry leaders in this space — may no longer hold true given the fast-evolving industry developments.

The latest sources suggest Chinese memory makers have improved their HBM technology faster than previously projected. As of today, CXMT is reportedly working on HBM3 and planning for mass production in the following year. This narrows the gap between CXMT and HBM leaders to about four years. Moreover, the firm also plans to announce HBM3E and push for mass production in 2027, according to Seoul-based Hyundai Motor Securities. If true, the gap will be three years instead of four.

At SEMICON China 2025, held from March 26 to March 28, over 1,400 domestic and international semiconductor firms gathered alongside senior Chinese government officials. There, much of the media spotlight focused on Chinese advancements in lithography and other semiconductor manufacturing equipment (SME) like the SiCarrier.

What was not covered, however, was Chinese memory advancement. Analysts attending the event expect rapid advancement in HBM through 2025, noting that HBM3E is the primary target specification for domestic HBM firms. While this does not indicate that China is on the verge of successfully developing HBM3E, the fact that domestic firms are actively developing around this specification implies meaningful progress in Chinese HBM technologies. At a minimum, it signals that China’s memory industry is working on the most cutting-edge HBM — a development that aligns with the note earlier.

If CXMT manages to bring HBM3E to market by 2025 — or even 2026, which the author views as unlikely — it would mark a major milestone in China’s push for semiconductor self-sufficiency and send shockwaves through both the global memory industry and policy circles. This scenario, however, is not entirely out of reach. SemiAnalysis projected in January 2024 that “CXMT’s HBM3E for AI applications could begin shipping by mid-2025.”

If CXMT rolls out less advanced HBM3 late in 2025 or 2026, even that would be surprising to many given the pace of progress thus far. It is important to remember that, unlike its legacy memory competitors, CXMT has only been established for 9 years, and its HBM2 only entered mass production last year. Not to mention the fact that the firm has been impacted by a series of export controls for several years.

From a technical standpoint, CXMT appears increasingly capable of producing the DRAM die for both HBM2E and HBM3, building on its progress in DRAM. The company is currently able to manufacture DRAM at the D1y and D1z (17 nm - 13 nm) node — technologies that are used in these two generations of HBM. Another leading technology consultancy, TechInsights, confirmed in its January analysis that CXMT is capable of manufacturing DDR5 at the D1z node (approximately 16nm). The density of CXMT’s DDR5 is comparable to that of leading global competitors in 2021 — Micron, Samsung, and SK Hynix — though the chip exhibits a larger die size and an unverified yield rate.

CXMT’s R&D team is likely developing sub-15 nm DRAM nodes, specifically the D1α and D1β (14–13nm), which are essential DRAM nodes for HBM3E. Although CXMT will likely face major challenges in developing D1α nodes without Extreme Ultraviolet Lithography (EUV), it is not impossible to develop D1α DRAM nodes without EUV. In 2021, Micron debuted its D1α DRAM without the use of EUV, paving a potential track for CXMT to duplicate.

Taken together, the author believes the more realistic assessment is that CXMT is currently developing HBM3 with the expectation for mass production to begin in the first half of 2026. For HBM3E, it remains too early for now to make a decisive call given its progress in DRAM and limited information on this specification.

It is worth noting that if CXMT develops HBM3 or HBM3E, careful evaluation of its overall performance and compatibility with large language model training will be essential. Past experience shows that not all HBMs within the same generation are created equal. Samsung’s HBM3E 8hi and 12hi have struggled to pass Nvidia’s qualification test as a supplier for its high-end GPUs over the past two years.

One might ask why CXMT isn’t pursuing HBM2E, which appears to be the logical next step on its technology roadmap. The likely reason is market timing: most domestic GPUs are already equipped with HBM2E, meaning limited commercial opportunity would remain by the time CXMT’s version reaches mass production.

Given Nvidia sold over one million H20 chips with HBM3 (or reported “H20E” with HBM3E) to Chinese AI firms before the ban, Chinese GPU firms must continue to compete against Nvidia’s product in the short and medium term. As such, it makes strong business sense for CXMT to prioritize advanced memory technologies like HBM3 and HBM3E, supplying domestic GPUs with more competitive HBM.

Strategically, staying competitive in the fast-moving AI chip space requires CXMT to pursue a leapfrogging strategy — aligning its products with the memory demands of domestic GPUs and ASICs to secure both domestic relevance and potential global competitiveness. For example, Huawei’s Ascend 910C — and future iterations — will almost certainly seek to upgrade from the 910B’s HBM2E to HBM3 or HBM3E, improving its memory performance and overall competitiveness.

The Multi-Dimensional Challenges

To be sure, there are multi-front challenges awaiting CXMT and other Chinese memory firms despite the improvement.

First, the restrictions around semiconductor manufacturing equipment will continue to hinder CXMT’s development. Although CXMT stockpiled enough semiconductor manufacturing equipment for HBM and DRAM production — likely sufficient to sustain operations through 2026 or 2027 — both existing export controls will still limit its ability to develop and scale advanced DRAM and HBM production in the coming years. For example, the December export controls restricted equipment critical to HBM manufacturing and packaging processes — including tools for through-silicon via (TSV), etching, and related steps. On top of that, the maintenance personnel from U.S. semiconductor equipment firms embedded at CXMT have been instructed to leave the company amid the tightening restrictions, affecting its development in DRAM and HBM.

Second, while the author outlines a potential path for CXMT to advance below the 15nm node without EUV, it is likely that adopting EUV will become inevitable for the development of cutting-edge DRAM and HBM, which is the case for other memory giants. Without EUV, CXMT could face challenges similar to those encountered by SMIC in recent years, in a way that struggles to improve yield, die size, and scale production. This choke point could continue to be the key roadblock for CXMT’s pursuit of cutting-edge DRAM and HBM.

Third, CXMT’s access to the global HBM market will likely remain limited for at least the next few years, capping its role in the global AI hardware supply chain. With a multi-year technology gap, its HBM offerings are unlikely to be adopted outside China, where buyers have access to more advanced alternatives. Moreover, ongoing U.S.-China tensions and existing and potential restrictions will deter foreign firms from adopting CXMT’s HBM for their AI accelerators, even if its products become technically competitive, due to fears of geopolitical fallout.

That said, if CXMT manages stable, scaled production of low to mid-end HBM at highly competitive prices, it could at minimum press the gross margin for other HBM players in the global market and secure some market share.

Fourth, the Entity List will limit firms’ commercial activities. Since 2020, the U.S. government has placed 768 Chinese entities on the Entity List, including industry major players such as Huawei and Chinese NAND leader YMTC. While CXMT is still notably unlisted, the author understands that there has been constant discussion in Washington about adding CXMT to the Entity List since January, posing a major risk for CXMT in the coming months and years.

Together, these hurdles could reshape the gap between CXMT and industry leaders in the years ahead. Still, if CXMT maintains its current momentum, its technological progress could further narrow the gap with leading players in the HBM market.

Lastly, it is uncertain whether CXMT can keep pace with industry leaders by consistently refreshing its product lineup generation after generation. Companies like SK Hynix are moving aggressively, launching HBM4 this year and planning HBM4E, which aims to fulfill the need of Nvidia’s rapidly evolving product line. Can CXMT keep up amid the current technological gaps and challenges? I will leave that question open for debate.

*The author would like to express sincere appreciation to those who provided valuable feedback on this piece, including Lennart Heim (RAND), Sravan Kundojjala (SemiAnalysis), Kyle Chan (Princeton University), and Sihao Huang (Oxford University).

*Acknowledged Limitation: This article does not delve into several critical processes in HBM manufacturing and packaging, such as through-silicon via (TSV), bonding, and related steps, which the author acknowledges should be included in the discussion of China’s HBM development. That said, the author believes these areas may present fewer technological hurdles for Chinese firms compared to the more complex challenges discussed in this analysis.

The author also acknowledges the inherent difficulty of projecting China’s HBM trajectory, given the dual constraints of limited public data and the rapid, often opaque nature of technological advancement within China’s semiconductor ecosystem. In recent private discussions with Korean analysts covering the memory sector, we echoed shared concerns over this persistent information asymmetry and the limited transparency — factors that inevitably affect the precision of any external assessment.

That said, this analysis is formulated based on both credible public sources and private insights, benchmarked against known technical progress and industry developments. While not exhaustive, it provides a meaningful perspective on China’s evolving position in advanced memory technologies and the broader implications for global semiconductor dynamics.

Read more

为全球华人游荡者提供解决方案的平台：游荡者（www.youdangzhe.com）
这世界的辽阔和美好，游荡者知道。使用过程中遇到问题，欢迎联系客服邮箱wanderservice2024@outlook.com.

【和放学以后永不失联】订阅放学以后Newsletter，每周三收到我们发出的信号：afterschool2021.substack.com 点击链接输入自己的邮箱即可（订阅后如果收不到注意查看垃圾邮箱）。如需查看往期内容，打开任一期你收到的邮件，选择右上角open online，就可以回溯放学以后之前发的所有邮件，或谷歌搜索afterschool2021substack查看。

截至目前，放学以后Newsletter专题系列如下：“在世界游荡的女性”系列、“女性解放指南”系列、“女性浪漫，往复信笺”系列、莫不谷游荡口袋书《做一个蓄意的游荡者》系列、“莫胡说”系列”《创作者手册：从播客开始说起》,播客系列和日常更新等。

大家好，本期Newsletter由西班牙的霸王花木兰轮值。

前段时间，莫不谷来西班牙找我玩，在西班牙待了整整十天，每天都是狂吃狂走，离开的前一晚，我们还在美丽的海滨城市Alicate愉快地游玩。那天忘记是什么话头，开始和莫不谷敞开心扉聊天，过去没有说的话，好奇想问却犹疑埋在心里的话，自己内心深处恐惧害怕的话，和我的眼泪一起，像是水龙头开阀一样哗啦啦流淌出来，从晚上八九点聊到凌晨三四点，聊了整整七个多小时。

从影响睡眠的角度来说，夜聊很伤身，聊到凌晨两三点人是很难继续入睡的，这里主要是莫不谷受苦，但是从另外一个角度来说，夜聊对于身心健康大有裨益，这里主要是我受益。因为坦诚和鼓起勇气发问让我松快很多，莫不谷给我的反馈也时常让我有恍然大悟的惊讶，原来如此的深思和不敢相信的疑惑，很多细节和对话我记得不太清楚，但是现在过去了十几天，夜聊后那种对自己认识了解更深，耳清目明的总体感受直到现在也还没消退。前两天我看到一个帖子深有同感，说黄圣依参加完浪姐，用新的目光审视自己的生活，开始考虑离婚。评论区有留言说“是不是女星们宿舍里夜聊，越聊越觉醒”。夜聊的魅力大概是晚上人更容易卸下心防、伪装和逞强，心和心有机会更近距离彼此看见，如果还能有精神层面的交流，人会进入比“冥想”“心流”还要沉静清晰的世界。

说到这里，大家也许会好奇我们究竟聊了啥，这次夜聊很私密，也没有任何录音，不过此前在莫不谷提议下我剪辑的《莫路狂花今夜不设防：人如何不糊弄和痛恨自己，并找到自己的渴望呢？》，就是一期女性夜聊的电话录音播客，很多人听完之后觉得像是做了一场线上Therapy。在莫不谷出发来西班牙的前一天，我们又打了一次电话，意外又录制了一期播客，深入讨论了如何避免过上荒谬的人生，糊涂的人生和自我欺骗的人生，预计在5月31日上线，到时候大家可以收听这期精彩非凡的内容，一起感受下女性围炉夜话带来的头脑风暴和精神充电。

（莫路狂花，一起去世界游荡）

就像我前面说的那样，Alicate这次夜聊中很多对话和细节我都有些记不太清楚了，但是其中有一个我深有感悟和体会的小事想要分享出来。因为我发现这不仅是我自己会遇到的问题，我认识十多年的朋友也遇到相同的问题，可能是女性普遍会遇到的问题。

2024年4月30日，是我拿到离职证明完成离职手续的第一天，今天2025年4月16日，我不上班已经快满一年了。一两年前，我还在为是否离职犹豫不定，上班痛苦，不开心，但这样的日子还能过，反正一睁眼，一闭眼又是一天，但离职是否就不痛苦？毕业以来从未离职过的我离职后生活会陷入何种境地？我真的能面对无着无落的生活吗？那时候我妈和我一起生活在上海，劝我不要离职时说：

“我的孩子我知道，你没有莫不谷那样的才华，也没有她的韧劲，过不了她那样的生活。”

听到我妈这么说的时候，我心里无名火蹿地冒出来，生气冷硬地说：“你为什么要这么说，你是在威胁我吗？在我看来，这句话对我没有丝毫帮助，而且我觉得你在威胁和恐吓我。威胁我不知好歹，不了解自己几斤几两，早晚会因为这样的选择后悔吃苦。”

依稀记得母亲那时候满腹委屈和生气，觉得她是全世界对我最好，最不会伤害我的人，但居然被我这样回复。

我记得自己嘴硬说：“做出决定的是我，承担责任的也是我，你可以不必支持理解我的决定，但请不要拖我的后腿。我知道自己要面临什么风险，也知道自己的恐惧怀疑，我不需要你的提醒，也不需要你指出我的恐惧和担忧。”

坦白来说，那时我的生气更多的是自尊被戳破的愤怒，因为不得不承认她说的对，所以我生气，我愤怒。我比她更清楚地知道自己无法像莫不谷那样写出令人触动叫好的文字，我了解自己不仅不会迎难而上，反而常常想要放弃，我害怕辞去这份给我社会认同和物质保障的工作后，我无法凭借自己的才华和能力生活，当发现自己身无所长的时候陷入全面崩溃，我不知道如何面对自己生活的坠落。

和母亲的对话以双方的痛苦、愤怒、难过和落泪结束。母亲说，父母是世界上最不会伤害我的人，也是无论如何最爱孩子的人。我很困惑，因为如果相信这句话，相信父母的动机和立场，我就需要谨慎考虑离职这种冲动且危害自己的决定。但不相信这句话，又如何理解母亲在上海日常生活中对我无微不至的照顾呢？下班在家的我，是不会有积极、勤快、周到这些只出现在上班中的品质，人一整个懒洋洋，和巨婴差别也不大。父母会帮我把水果洗干净、削掉皮、切成块，再插上牙签，和湿巾、垃圾袋一起送到我的床边，等我吃完再帮我收拾整理。

我最终决定不听取父母的建议，不能让父母干涉我生活的考量是，既然我的愤怒和生气更多是对自己，就不必让父母承受无名之火，我恐惧，我害怕，我担心，但我做出了选择，我就自己来承担后果。继续上班的日子好像还能过，但我的情绪和身体都在告诉我，我过不下去了，忍受不了形式主义，忍受不了照本宣科，忍受不了天天洗脑，忍受不了目之所及的人不能被好好当作人，猫狗不能被好好当作生命的痛苦和悲剧，忍受不了已经觉醒却还不行动对自己的憎恨，我的耳朵像是发烧一样又痛又痒，我的大脑像是上锈一样迟钝缓慢，我的胸口像是堵住了石头一样憋闷难受。父母以爱和以关心为名劝解的想法，对我不仅没有帮助，还会让我忍不住迁怒。

另外的原因是，我能从父母的担心害怕中感受到深植在她们心底的生存焦虑，母亲害怕我放弃金饭碗后过上流浪的生活，更像是她自己的恐惧，挣扎拼命了一辈子希望不再挨饿被欺负，想要争口气过上有尊严的生活，却眼睁睁看到女儿要放弃令人艳羡的生活。因为我看到了她们的恐惧，所以我更不能让这种恐惧影响我，我不想一辈子在恐惧的阴影下生活。

做决定的过程是漫长、反复犹豫、焦虑和痛苦的。但决定一旦做出以后，人会比想象还要快地适应新生活。大概在去年十月，我环游世界127天结束后回到了上海，准备西班牙签证的申请。有天晚上，我一边散步一边和几乎像发小的朋友电话聊天，不知道怎么，聊到了她说，母亲和她说过：

“莫不谷和霸王花的生活很潇洒精彩，但你知道，你是过不了这样的生活的”。

这熟悉的话语让我立刻警觉，马上问朋友，那你是怎么想的？她平静地说：

“母亲说的对，我也这么想。”

我很讶异，我说，为什么你不生气？不愤怒？我的母亲和我说了同样的话，但我不服，我就算是心服我口也不服。

我曾以为，只有我内心在恐惧害怕在焦灼忧虑，只有我在面对父母对我才华能力的担忧、不信任。和朋友聊天后才发现，我不是一个人。同时发现，因为我不服，即使怀疑我也做出了选择，是我最终的选择和决定让我的生活发生了变化，走上了从未设想过的道路，看到了从未想象过的世界风景。就好像此前莫不谷提议去非洲Safari游荡时，原计划一起要去的有好几个朋友，大家因为各种各样的原因没能成行，其中一个朋友因为母亲害怕安全哭着挽留没能参加，从没去过非洲的我依着刻板印象想到条件可能很艰苦，卫生条件会很差，也许路上也会有风险，甚至我的领导同事都让我落地非洲报平安，我还是去了。因为那是唯一让我还期待生活，期待第二天日出的提议。我曾经畅想过，与其行尸走肉地活着，不如去一趟非洲大陆，亲眼看看电视里的动物世界，这样就算死了，我也没白来这世上一遭，做鬼我也骄傲。如果那时我也充分理解父母的担忧和顾虑，认同父母对我作出的判断，我想，我这辈子都很难有机会看到更大的世界。

当我终于鼓起勇气在Alicate和莫不谷分享过去这件小事，心里轻松释怀了不少。莫不谷并没有“与任何人比较”的想法，她只是选择了做她自己，但当我选择不止要当“拍手鼓掌”的观众，还要“上台表演”的时候，便下意识地会和身边亲近的朋友对比，也因为对比后认识自己没有才华而心虚害怕，还会因为自尊心无法坦诚这一点，更不可能坦诚我的父母对我作出这样的判断。大概是来到西班牙给了我坦白的力量和勇气，莫不谷说：重要的是你最终做出了选择。

虽然我做出了与我的朋友不同的选择，但我们在某种程度上都认同了父母对我们才华能力的判断和建议，在和莫不谷深入聊天后，我想指出的是，这个判断是错误的。

前段时间莫不谷推荐我读了一本书《活出生命的意义》，书里说，生命对每个人都提出了问题，我们必须通过对自己生命的理解来回答生命的提问。因此，否认自己生命的独特，不相信自己生命的独特是错误的。如果不承担起自己生命的责任，要么就会去做别人做的事（随大流），要么就会去做别人希望自己做的事（集权主义）。

我的父母和我说，我没有才华，过不了莫不谷的生活，这是错的，让我内心相信这句话，更是大错特错。朋友的父母和她说，她过不了莫不谷和我的生活，这是错的，让她的内心相信这句话，也是大错特错。不能因为父母不相信自己的生命是独特的，就不愿意相信自己孩子的生命也是独特的。不能因为父母不相信自己孩子的生命是独特的，就理解、认同父母的错误。错误的信息就是错误的，糟糕的理念就是糟糕的，我们不能被错误和糟糕所影响。

以及，我也必须承认的是，我确实是过不了莫不谷的生活，我的朋友同样也过不了莫不谷或者我的生活，因为我只能过我自己的生活，我的朋友也只能过她自己的生活，因为生命给莫不谷、给我、给我的朋友，给每个人都提出了问题，我们只能担负起自己生命的责任。但是，莫不谷能够探索世界的辽阔和美好，探索自我的可能性，我也能，我的朋友也能，这世界上任何一个女性都能。

所以我想和阅读这篇文章的你分享，我们不要随大流做别人做的事，也不要去做别人希望自己做的事，即使这别人是你觉得无比亲近的父母，我们只能去做正确的事情，去做自己想做的事情。

写在后面：

想了一想发现这篇文章还没回答一个问题，就是我内心的恐惧和担忧有没有落地？在5月31日上线的莫路狂花第2期播客里，大家可以找到许多答案和启发。这里我想借用波伏瓦的一句话，“离职和离开故土不是“万灵药”，只有当女性知道如何以积极地方式利用她们的自由，她们才会获得解放，但为了发现自己的可能性，离职和离开故土往往是一个必要条件。”

（我要在我认同的环境里生活）

（竹杖芒鞋轻胜马，谁怕，一蓑烟雨任平生）

为全球华人游荡者提供解决方案的平台：游荡者（www.youdangzhe.com）
这世界的辽阔和美好，游荡者知道。使用过程中遇到问题，欢迎联系客服邮箱wanderservice2024@outlook.com.

【放学以后文章&书籍&其它】

解锁放学以后《创作者手册：从播客开始说起》：https://afdian.com/item/ffcd59481b9411ee882652540025c377

解锁莫不谷《做一个“蓄意”的游荡者》口袋书：
爱发电：https://afdian.com/item/62244492ae8611ee91185254001e7c00微信公众号：《放学以后After school》（提示安卓用户可下载“爱发电”app，苹果用户可把爱发电主页添加至手机桌面来使用，目前爱发电未上线苹果商店）

Newsletter订阅链接：https://afterschool2021.substack.com/（需科学/上 网）

联系邮箱：afterschool2021@126.com （投稿来信及合作洽谈）

为全球华人游荡者提供解决方案的平台：游荡者（www.youdangzhe.com）

小红书：游荡者的日常

同名YouTube：https://www.youtube.com/@afterschool2021

同名微信公众号：放学以后after school

欢迎并感谢大家在爱发电平台为我们的创作发电：https://afdian.com/a/afterschool

播客收听平台：【国内】苹果播客（请科学/上网）、爱发电、汽水儿、荔枝、网易云、小宇宙、喜马拉雅、、QQ音乐；
【海外】Spotify、Apple podcast、Google podcast、Snipd、Overcast、Castbox、Amazon Music、Pocket Casts、Stitcher、Radio Public、Wordpress

Read more

Does America still have what it takes to stand up to China? Does short-term military readiness trade off with long-term strategy? What does the US need to do today to stay competitive for the rest of the century?

‘’ is the author of Breaking Beijing, a Substack examining the military dimensions of US-China competition. Tony’s Substack goes deep on subjects you didn’t know you needed to understand, like Arctic policy, and takes a refreshing step back to look at great power competition holistically. Tony wrote Ex Supra, a sci-fi thriller about a near-future US-China war.

We discuss…

• What it will take to win the 21st century, and what America needs to prioritize in the short, medium, and long term,

• Why investing in education, basic science research, and foreign aid pay dividends in military readiness,

• Why Washington is short on coherent China strategy,

• Taiwan’s impact on global nonproliferation efforts,

• How AI could change warfare, even if AGI can’t be considered a “wonder weapon.”

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

A Blueprint for Competition

Jordan Schneider: Tony, why don’t you share as much autobiographical information as you feel comfortable?

Tony Stark: Tony Stark is the nom de plume that I’ve used for years. I’m a China policy guy by both research and practice, with some background in tech research. I served in the US Army infantry, formerly active duty. I worked on Capitol Hill and in OSD, and now I’m in the private sector trying to help the US win the 21st century.

Jordan Schneider: Let's talk about the endgame for U.S.-China relations — you called it Plan Noble. When people talk about the endgame for U.S.-China relations, what should they actually be talking about?

Tony Stark: What they should be talking about is how to make the world unsafe for the Chinese Communist Party while making the world safer for Americans and Western-style democratic order. That is the ultimate endgame. Anything else where you talk about trying to depose the regime, or regional areas of control between democratic and authoritarian powers, doesn't actually solve anything, and it doesn't put you in an advantageous position. You're either ceding too much ground, or it's overreach. The goal is to make the Chinese Communist Party feel unsafe to follow their policy goals out in the world beyond their shores.

Jordan Schneider: Can you talk me through your decade-by-decade framework for the competition?

Tony Stark: 2025 to 2030 is your rough near term. That's your immediate threat. That's where the investments that you've already made or perhaps chosen not to make are directly impacting your ability to operate from a military standpoint in peacetime or wartime.

What you're trying to do now is focus on investments that you can produce and get off the production line in the next one, two, three years. You're focusing on maximizing that production, showing steady state investment to industry, both in terms of workforce and production. You're starting to invest in things that are attritable, because if you're only investing in exquisite systems, you are not really able to plan for a longer fight in the event that you get into one.

There are things that you have to do to ensure that we're still competing in the 2030s. This means prioritization to ensure deterrence in and around Taiwan. It means starting to build relationships throughout Southeast Asia, so if there are other contingencies you have to worry about, you could start laying the groundwork there.

Additionally, what are the basic R&D investments that you have to start to do from an AI side, quantum, synthetic biology? It starts today. The discussion in China policy over the last five to ten years has been, “We should have done this ten years ago. We disinvested of all these things in the 1980s and 1990s, and this is where it got us.” Now you can think about that from a forward-looking perspective — what do we need to keep and buy today such that in 2040 we’ll be saying, “Thank God we invested in that”? Now is really your last opportunity. You have to do it now.

Beyond 2030, that's when you start to see payoffs from large-scale industrial investment, education investment. Investments you start today start to pay off. 2030 to 2040, the big one is AUKUS. You're looking at smarter machines. You're looking at new weapons systems that are coming online that might be in initial rate production today, and you're looking at all of the new doctrine coming out across the forces that is actually getting their reps and sets internally for training and now being able to be demonstrated.

People usually use the 1980s example of the Abrams, various fighter jets, etc., that, between that and the combination of new doctrine through the ‘80s of air-land battle that culminated in the Gulf War, that's what we're trying to pursue for the 2030s. You’re getting to that point where not only have you managed to stave off destruction today, but you are also prepared for a higher-end fight with an even more capable People’s Liberation Army in the 2030s.

In the 2040s, your early R&D bets start to see payoffs. Consider this as buying in on a company that has maybe five employees and then might expand out to a real-sized corporation by then. You’re doing the government investment equivalent of that for any sort of R&D project or force design. Quantum and many of the positive sides of synthetic biology are likely in that late 2030s, 2040 category.

Through 2040, you have to give yourself the maximum wiggle room to account for external events, changes in budget, etc., while still having this guiding principle of what we are doing is to keep the CCP contained, to make the world unsafe for their operations abroad. That gets us to mid-century.

Jordan Schneider: In order to make sure we don’t find ourselves in a World War III or happen to lose it over the next 30 years, what is the first thing that folks should be thinking about, focusing on, and shoring up?

Tony Stark: You have to be able to build, acquire, and deploy things today that work. Where some of the scholars and leading thinkers — Rush Doshi, Elbridge Colby, and others — get torn up in debating this is whether to prioritize short-term or long-term, and you can’t just choose one. You have to do both. That’s the very difficult reality.

You need to hold that five-meter target — “What if Xi really wants to move in 2027, 2028, or 2029? What do we have to do today? What can we feasibly do today to prevent that? How do I do that in a way that still allows me to invest in the long term?"

Yes, you have this sprint in front of you, but if you burn all of your energy in the first mile of a 26-mile race, you will lose. Simultaneously, if you just go at an easy pace, at your own pace, irrelevant to the competition in your marathon, where is your competitive spirit? You’re definitely not going to be the first at the finish line.

What does that specifically look like? Unmanned systems and munitions. Munitions are probably the biggest investment you can make, because magazine depth is just an exorbitant challenge. At the high point of Ukraine operations, they were burning 60,000 rounds of artillery a day. The Pacific fight involves a different set of munitions.

Really, it’s about ramping up production and giving the industry that signal of, “You are reliably going to get investment from us for the next few years. This is our priority.” A lot of our tech works. There’s not a whole lot of areas where we’re saying, “If we had a new main battle tank by 2027, we’d win the fight.” The Abrams is pretty good. But I would like about 1,000 more long-range anti-ship missiles. Everyone would. That’s obviously capped by actual production rates, but I think you get the idea.

Jordan Schneider: Maybe going one step up from that — you’ve got to want to do the competition. You wrote a piece back in 2023 entitled “Where Did All the China Hawks Go?” We’re recording this March 21st, 2025, and it is still very unclear just what this president’s stance towards China and Chinese territorial aggression is going to be.

You can lose before it even starts. We’ll get into the acquisitions stuff and force structure stuff in a second, but you can lose at a systems level in two ways. First, you don’t even show up, and second, you stop being the system that you initially thought you were in the first place.

This is a global beauty contest as much as it is a competition between the US and China. If America is a less attractive partner for ideological reasons or for reliability reasons, then that’s a problem. The US and China together add up to less than 50% of the world’s GDP. There are a whole lot of other countries that are potentially up for grabs, that are going to choose which way to lean over the coming decades.

As nice as magazines are, it’s nicer to have the entire industrialized world on your side versus on the other side. I go a level up personally when thinking about this problem.

Tony Stark: 2015 is a good place to start looking at China Policy because that’s when the South China Sea island development and various hacking operations hit their peak, in the Obama administration.

That was the start of when, aside from us wonks, people needed to start paying attention. The Trump administration focused on deterrence by denial. There was a mix there where President Trump was friendly with Xi Jinping, but there were also tariffs, and those in the DOD under Jim Mattis and others were trying to figure out how to fight in the first and second island chain — how to show up to the fight.

The Biden administration took a different approach. They had to continue some of the Trump administration’s legacy of being able to fight in the first and second island chain, but they also said, “We’re going to do industrial policy. Not only are we going to show up to the fight, but we’re going to show that America is strong enough to survive that long fight, that competition.” That’s the CHIPS and Science Act and parts of the infrastructure bill — what does that 15-20 year investment look like for us?

But simultaneously, they didn’t really like the idea of having to do this from a foreign policy perspective. It was not particularly convenient, and that was a challenge.

Here we are today in a world where we still don’t have a solidified China strategy. On top of that, there are not many champions of actual hawkish China policy in DC or around the country. You brought up “Where Did All the China Hawks Go?” It’s not particularly a blue or red problem at this point.

There are motivators on both sides where there’s less interest in China as a competitive space, whether that’s because of wanting to focus on domestic policy, the Western hemisphere, commerce, or making trouble with other folks, rather than saying, “This is the priority."

That’s where we are. We are adrift, and this is a really bad time to be adrift. I mentioned before that we should have done this 10 years ago. It wasn’t good to be adrift 10 years ago, and now we’re about to be adrift on a raft going into a hurricane. This is going to get really bad if we don’t figure this out.

Jordan Schneider: The urgency is less from an “I think Xi is going to invade in 2027” perspective, but from a comprehensive national power perspective. The Chinese military in the 1980s, 1990s, and 2000s was not a serious challenger on paper to the US and its allies. That is different now because China is richer, they have modernized, and they are able to project force in new and interesting ways at scale. You would have to go back to the early Qing dynasty to have a relative power comparison to what we’re looking at today.

It’s not a bad thing that a billion people in China are richer than they used to be. However, we’ve all learned this very horrific lesson of what autocrats can do with Putin in Ukraine. Even if you think Xi doesn’t necessarily have bad intentions over Taiwan, building a deterrence capability ensures that whoever’s leading China over the next 30 years understands it would be devastating for them to start a high-intensity conflict. This seems like a reasonable insurance policy for the US to invest in.

Tony Stark: Over the late Cold War and definitely over the last 30 years, we’ve become accustomed to being able to pick and choose the fights we enter or how we engage in them. With 9/11, we didn’t pick that fight, but afterward we maintained this idea that we can go anywhere, be anywhere, and simultaneously withdraw if we want.

Some leaders in Washington might decide, “We don’t really want to pick a fight with the Chinese.” That’s understandable. We didn’t really want to pick a fight with the Nazis either in 1937, and we still ended up having to confront them. Wars happen because of fear, honor, or interest. The enemy gets a vote, and if they decide that fear, honor, or interest requires them to either fight us, challenge us, or push us around, then they will do so.

Elbridge Colby said, “Taiwan is not an existential problem for us.” Not in the sense that losing Taiwan means the end of humanity. But it’s like hanging off the edge of a cliff and having one hand slip. We’re already at that precipice. If you lose Taiwan, yes, you lose face, and you lose TSMC. But there are also 24 million Taiwanese who would fall under the boot of a regime that believes they should not exist.

This situation triggers concerns for every ally or potential ally in Asia and Europe that the US might not be there to back them up. “The authoritarians are on the march, and this might happen to me.” That either leads these countries to surrender or to acquire nuclear arms. Nuclear proliferation is something you don’t want to pursue. That’s where the existential problem for America transitions to an existential problem for humanity — when you start worrying about nuclear proliferation. The more fingers with access to nuclear buttons, the higher the risk of nuclear conflict, which nobody wants.

Jordan Schneider: The question is whether we’ve already crossed that threshold with what we’ve seen coming out of Trump’s diplomacy over the past few months.

We’re going to do a whole show on global nuclearization in response to America’s treatment of NATO. It’s really dark.

Tony, you have a recent piece about the evolution of combined arms warfare. What is your mental model when trying to evaluate what changes could be coming to the battlefield, and how to invest around them?

Tony Stark: Everyone focuses on what we see in Ukraine, and I think much of that is applicable. When discussing a potential US-China conflict, you also have to understand that Chinese capabilities are technologically superior to the Russians in many ways. You probably have to increase that threat assessment significantly. This isn’t to make the Chinese sound invincible, but to understand the substantial difference between Russian and Chinese technology in certain areas. The Chinese are undoubtedly learning many lessons from the Russia-Ukraine conflict.

Subscribe now

Regarding the future battlefield, we’re evolving from traditional combined arms warfare that has existed for the last 150 years. Combined arms warfare integrated radio, control, precision fires, and different military branches at both joint force and service levels to synchronize effects and damage the enemy. This involves coordinating artillery while infantry and tanks maneuver, with tanks and infantry working together rather than in separate formations. That’s the basic concept for the audience.

Now, we’re taking all that learning from the last 150 years and applying it to unmanned systems that, at a rudimentary level, can think and perform tasks autonomously at a narrow level. The challenge is determining what’s best for humans to do, what’s best for machines to do, and where they work in concert.

For example, the Ukrainians still need soldiers and manned tanks on the front lines, but there are particularly dangerous areas where they operate, such as the combined arms breach — blowing a hole through enemy defenses. The enemy knows you’re coming, likely knows where you want to breach, and has artillery, wire, mines, and other defenses in place to prevent this. You don’t want to be the human in that situation, even if you’re in a tank. I could send you countless videos from Ukraine demonstrating why. But if you can send machines to do it, you reduce your exposure while accomplishing the rest of your mission.

That’s the first part of robotics and unmanned systems as enablers for shooting, moving, and communicating. They allow you to shoot with better precision, communicate more effectively, and move either faster or in more dispersed ways. These are the fundamentals of warfare.

Moving to the shooting aspect, Ukraine uses what they call one-way attack UAVs or even USVs in the Black Sea. These rudimentary war robots make decisions independently or are remotely controlled, functioning to collect information, transmit it to personnel behind lines, and target and strike. This represents the next iteration of artillery, close air support, and long-range fires.

The final component is maneuver — infantry, tanks, and other forces pushing through, fighting through the enemy, taking and holding ground. This is the hardest component because it requires relying on those previous capabilities — collecting information, shooting, etc. — and combining everything into one package that must be survivable. While machines are more expendable than humans, they’re still expensive and time-consuming to build.

The challenge is creating something that can survive in harsh environments. Cold, heat, rust, and moisture annoy us as humans but don’t typically kill us, unless you’re in extreme conditions like the Arctic. However, machines have specific thresholds, often lower than humans, for what they can operate in. How do you ruggedize equipment while ensuring it keeps pace with the rest of the formation? How does a ruggedized unmanned vehicle with a 50-caliber weapon maintain pace with tanks moving at 30-40-50 miles per hour, while remaining survivable, maintaining targeting capabilities, and withstanding being thrown and bounced around with all its internal electronics?

Essentially, you’re asking for the equivalent of a ruggedized iPhone or laptop to survive being bounced around on high seas, in mud, in swamps, etc. That’s extremely challenging, especially to get it to do all the things you want it to do. There’s a lot progress in the startup world and even among major defense contractors, but this represents the current battle laboratory we’re witnessing.

Jordan Schneider: This is a good counterpoint to the narrative that AGI is going to change everything once and for all.

In Watchmen, which is a great graphic novel that was turned into an excellent HBO series everyone should watch, Dr. Manhattan is a superhero that America deployed in 1974. He shows up in Vietnam, and the war ends five minutes later because he’s essentially a Superman-type figure. America’s ability to control him simply wins wars without discussion.

It seems implausible that in the near or medium-term, AGI is going to create a gap in capabilities at the level of the US versus Iraq in 2003.

Tony Stark: To that point, as a reminder to everyone, please do not connect the nukes to the AI. Please do not do that.

Regarding the Dr. Manhattan concept — your AI is only as good as your data. We don’t actually know what the equation is to get us to AGI. You’re not looking for the Higgs boson where you think, “I believe it’s in this range and if we get there, I just need to look in the right space.” It’s as much a philosophical question as a scientific one.

I would discourage those who claim that AGI will solve battlefield problems from strategic decision-making to having some “wonder weapon.” That’s not happening.

In the next 10 years, you’ll see incremental gains from AI in tactical decision-making, data processing, the ability to find, fix, and finish targets, as well as logistics.

Logistics are probably the most significant use case for AI because humans are really bad at efficient logistics.

Jordan Schneider: Let’s talk more about the “wonder weapon” concept. Historically, these were terrible ideas. Hitler kept thinking he would invent one weapon that would win the war, and it didn’t work. He didn’t even know he should have been focusing his resources on nuclear weapons instead of three-story tall tanks and V2 bombs.

The enemy has a vote. The enemy can copy and counter what you do once you deploy it. War is much messier, and there is no single trump card that can help the US beat China.

Tony Stark: Exactly. Some of that perspective comes from how we teach history, either at the popular level or undergraduate level — “The allies got the bomb, and then we beat the Japanese.” In reality, many other things happened before that to get us in that position.

I really like your point that Hitler didn’t even know the actual wonder weapon he should have been building. That’s a perfect comparison to AGI. I assure you, if you try to chase AGI by itself, you will not reach your goal.

If you focus on very narrow applications — and I say narrow in the AI context, not meaning you just do one thing — if you focus on applied tasks that will actually have effects at the tactical and operational level, as you start to aggregate those systems, you’re learning valuable lessons. You’re proving what does and doesn’t work, and that’s not just about the machine, but about how to develop AI itself.

Jordan Schneider: Foreign aid is a rather spicy topic lately. We’ve seen USAID seemingly hitting a giant destructive reset button. Tony, what’s your pitch for spending money in a broadly aid-focused sense in foreign countries?

Tony Stark: I wrote an article a while back called “Fighting the Four Horsemen.” The Department of Defense fights war, which is one horseman. Foreign aid fights the other three — pestilence, death, and hunger — which cause pain and are often drivers of war.

Do our foreign aid programs get everything right? No. I can’t think of a government or corporate investment plan in history that got everything right. Everyone has a pet project they want to work on, or they simply make bad guesses. Your data won’t be perfect, and you have to plan for that.

What does foreign aid buy us? It buys goodwill, which sometimes matters when you’re a soldier stuck behind enemy lines and somebody remembers, “The US government fed me during a famine.”

It also allows us to prioritize the things that matter. If we prevent a famine in Africa, then we don’t have to deal with the economic and possible military fallout of that famine. From a purely realist standpoint, this prevents us from having to dedicate additional resources when we need to focus on the main adversary.

At its most basic level, that’s what foreign aid does for us. Obviously, it does many other things — it fights viruses at their source, prevents outbreaks, and if outbreaks occur, prevents diseases from spreading through early identification. All these things matter when discussing how we create stable institutions at home and a stable, prosperous economy for more Americans. As we’ve seen, viruses and ecological disasters hinder those efforts. You’re fighting many threats abroad through foreign aid so they don’t come home.

Jordan Schneider: I couldn’t have said it better myself. The Office of Net Assessment is being canceled. Hopefully it’s just a reset button and won’t disappear forever. But I think there are parts of this new energy in Washington which aren’t even anti-intellectual — they’re anti-thought. You should do things because they seem “based,” as opposed to actually putting in the time to think about costs and benefits.

Tony Stark: US history throughout the Cold War is littered with examples of times when we simply didn’t understand the ground state of things, and we either made situations worse or ignored them. I can think of any number of coups or civil wars we got involved with in the ’50s and ’60s that might have gone differently if we had actual people on the ground. That’s an example of where information matters. You cannot simply look at the globe and decide, “I want to do that.”

The broader case for education is that an informed population is more independent. America is a nation based upon choice, opportunity, and independence. An educated civilian populace leads to prosperity — that’s simply the equation.

I’m not saying that everyone needs a college degree or a master’s degree. I don’t agree with that, and that’s not how the economy works. You need people in trades. You need people who join the military at 18 and go to college later in life.

When I talk about education, everyone focuses on higher education because that’s usually their most recent memory, that’s the fun part, and that’s where you get into politics because you’re 18 or older. K through 12 education is largely neglected.

Jordan Schneider: Can you illustrate with the anecdote about Army recruits with literacy issues?

Tony Stark: As a former Army infantryman — I enlisted after college because I decided I needed to earn my way to leadership — within the first two weeks on the ground at Fort Benning, you go through what is basically a leadership training course. What it really involves is going from station to station with your group of recruits.

You read about a particular Medal of Honor winner and try to complete team-building challenges. You can probably imagine what that’s like — an outdoor course where you need to get across the fake lava by putting planks together or something similar. When you inevitably fail, you do a lot of push-ups or burpees.

The drill sergeants make the recruits read the Medal of Honor citation. What struck me was how many recruits struggled to read their own history. That’s your lineage as an infantryman — the people who came before you and did great things — and you can’t read it. If you can’t read it, how can you understand your place in the world? How can you analyze things on the battlefield? How can you make decisions in complex modern warfare?

People might argue that in the 1800s, half the recruits couldn’t read. It’s a different story now when you have to operate drones and tanks. You need to be literate and able to make complex decisions.

A Marine Corps report came out a couple years ago stating that the ideal infantryman is around 26 years old with a bachelor’s or master’s degree. The population can’t support that, and frankly, I don’t think you actually need that. What you’re looking for is a more well-educated population through K through 12.

Jordan Schneider: When you see people with degrees from prestigious institutions expressing thoughts and logic that demonstrate brain rot from short-form media and Twitter — it’s evident that their content consumption has shifted dramatically over the past five years. That’s a scary transformation to have witnessed closely over recent years.

Tony Stark: Don’t get me wrong — I enjoy funny Instagram reels too. The point is that reading from an early age teaches critical thinking because you have to read, decipher, and learn about the world.

Never let anyone convince you that humans aren’t built upon curiosity and research — we are. If we give people the right tools, they will research sciences and social sciences rather than QAnon conspiracy theories.

Jordan Schneider: What are your dream pieces of legislation?

Tony Stark: I would like to see multiyear procurement for weapon systems. I know there are Congressional limitations on spending beyond two years, but multiyear procurement, especially for long-lead items, would be beneficial.

I would also like to see legislation on refining our economic warfare capabilities. I wrote a piece a long time ago about a Department of Economic Warfare. That doesn’t necessarily need to be created, but it’s clear that each economic warfare component — whether it’s the Bureau of Industry and Security or the Office of Foreign Assets Control (OFAC) in Treasury — advocates for policies from their own perspective. That’s the nature of Washington.

This makes it difficult when facing a massive economic competitor like the PRC. You have people in some departments saying, “Our priority is investment,” while others say, “Our priority is hunting down terrorists.” This leaves us asking, “What tools do we reliably have to fight the great economic competitor that is the PRC?” I’d like to see legislation addressing that.

Jordan Schneider: Any book recommendations?

Tony Stark: I think Rush Doshi’s The Long Game is valuable. I don’t agree with everything Dr. Doshi presents regarding his assessment of the broader threat matrix of the PRC. However, if you want to read the party in its own words — because I’ve encountered many people in the DC policy community who ask, “What are the Chinese saying?” — there’s a book right here that documents it.

You can read it if you choose to. Literacy, again, is important. That’s one recommendation.

Number two is Spies and Lies by Alex Joske, because everyone should better understand how our main adversary chooses to interact with people from an influence perspective. I found it to be a fascinating book. I’ve heard there were other case studies that were going to be included but weren’t for various reasons. So if someone reads that and thinks, “This is only 10 cases,” I assure you there are more.

You have to read Chris Miller’s Chip War because the semiconductor problem, like most supply chain problems, is incredibly dense. He presents it in a way that’s very accessible and goes beyond simplistic views like “TSMC makes chips in Taiwan and America wants control of that.” You need to understand why this is so important and the potential economic fallout if we lose TSMC in any fashion.

Jordan Schneider: I’ve read all three of those books. Regarding Chris’s book, it’s been out for two or three years, and it’s frustrating that no one has written a “Chip War for Robotics” or “Chip War for Biotech.” This book sold many copies. We need the industrial and global national security history of more industries.

Tony Stark: I would pay so much money for a “Chip War for Synthetic Biology.” For the Ex Supra sequel, I’ve consumed a vast amount of books on synthetic biology. Textbooks aside, most of them are very preachy. I was listening to one audiobook that said, “The Chinese might do some bad things in Xinjiang, but who can tell?” Clearly, they’re trying not to get their research canceled in Beijing.

I would really like to see that because it’s such a fascinating debate you can have with anyone in any policy community about how synthetic biology might impact you. Unlike Chip War, which is a very technical view of the technology, synthetic biology is darker. One book posed the question, “What if one day it’s immoral to not modify the genes of your baby?”

Building a Career in Writing

Jordan Schneider: All right. Tony, do you mind telling everyone your age?

Tony Stark: Yes, sure. I'm 29 going on 30.

Jordan Schneider: Awesome. When did we start Breaking Beijing?

Tony Stark: Two and a half years ago. Before that I've been writing on and off since I was in college.

Jordan Schneider: When I talk to students and student groups, oftentimes people ask me, "Jordan, what's your advice for getting into policy?" I say, "Start a Substack. If you're scared about using your own name, you can do it anonymously, and it'll still lead you into cool places, doing interesting things.” Here, we have the one and only Tony Stark, the perfect example of this. What advice do you have for folks debating whether or not to put their writing out in public?

Tony Stark: Write your own stuff. First, don’t use AI, but more importantly, write what you know and what you’re interested in. Don’t chase buzzwords.

If you’re thinking about stuff in your own way, that’s where your writing is going to come out best. It’s going to show as your own brand from a personal marketing standpoint, and you're also going to have a better sense of control over what you should and shouldn’t write.

Jordan Schneider: The most fun you will have in doing a policy writing job is when you get to choose what to write about. The best version of this job is you picking the topics that you're most interested in, reading about them, and then writing in your own voice and style.

This is the cool and scary thing about this field — it’s not like immersion lithography, which you cannot do unless you are at TSMC or Intel. You can write about policy just by sitting at your computer reading and writing. You shouldn't gatekeep yourself.

The other thing you said, Tony, which is really important, is that you want this to be the most fun part of your week and a repeated game that you play. You’re probably doing this on nights or weekends. It has to be exciting to you for a reason other than 4D career chess.

Tony Stark: Absolutely. I even tried, once or twice, to force myself to write on current events, and you just don’t get good quality work, especially if you're trying to use it as a way to promote yourself.

Secondly, reps and sets matter. I got really good at writing because I wrote frequently for work and in my personal life. Not all of it saw the light of day, but just writing and getting that feel for that, getting that feedback, going through editing — that’s what's going to make you better. It’s just like physical fitness.

Jordan Schneider: Can you talk a little bit about the doors writing opens, even if you only have, say, 2,000 readers?

Tony Stark: Before I wrote on Substack, I wrote on Medium, and I wrote threads on Twitter. I don’t think people understand how much policymakers, and especially staffers, are online and reading that stuff. They’re looking for feeds of information. They’re looking to understand what’s next, what’s current. That’s the space that you have to plan.

It will be a slow burn at first. I’ve had high-ranking commanders reach out to me. I’ve had high-level politicians reach out to me. I've been hired now at two jobs at least partially because of my writing.

It doesn’t happen to everyone, and it's not going to happen immediately, and you should pace yourself for that. Work on focusing your craft. But yes, it can happen.

Jordan Schneider: As a hiring manager, what you want to see is a body of work and sustained commitment and excitement to the topic, not just writing about critical minerals because it's in the news this week. The best way to prove you’re interested in something is to show that you've been thinking about it critically. It matters less if you are right or wrong, but just that you are being rigorous and analytical in your thinking on a topic is the thing that gets people excited.

Tony Stark: Another very important part is that you have to know when you are wrong. Maybe your writing cadence was wrong, how you phrased something was wrong, or a concept was wrong. People who stick to their policy viewpoints despite being proven wrong repeatedly, even if it's something very niche — the audience doesn’t like that. They want to see that you can iterate.

Jordan Schneider: By the way, when you're in your 20s, you're not like John Kerry running for president in 2004. No one is going to care whether or not you flip-flop on something. The idea is just to show that you are thinking and continuing to think about whatever it is you're interested in.

In your writing, you are manifesting the sort of doors that are going to be open to you in the future, and it's just better to do that about the stuff you’re passionate about than the stuff that’s current.

Tony Stark: Absolutely. You don't want to box yourself in, because all of a sudden, you become that one guy or girl that has the expertise, and you didn’t want to have that expertise. Now you’re stuck somewhere you don't want to be. It’s better to just follow your passions.

Jordan Schneider: Any other words of wisdom?

Tony Stark: Write what you know, don't chase other peoples’ ideas, and don't be afraid. You are already ahead of your peers because you are making the attempt to write.

Jordan Schneider: Here’s an open invitation — if you write five Substack articles, Jordan and the China Talk team will give you feedback. That is the new policy. DM me to have me review your writing.

Let’s close with discussing your book, Tony. I don’t read a lot of this genre...

Tony Stark: I don’t either. That’s my actual confession. People ask me, “What are your top 10 sci-fi books?” I get to five and then I’m stuck. I love science fiction, but I think I can predict where most books in the genre end because they’re so repetitive.

Jordan Schneider: The strongest elements for me were the out-of-the-box but still grounded military scenarios — warfighters in outrageous geopolitical or technological situations where there’s a cyborg assassin on your tail or drone swarms. Not drone swarms in a hand-wavy way, but as described by someone who was an infantry officer who has done the reading and can paint a grounded yet novel and provocative vision of what the future of war might look like. I’m curious, in your first book and now going into your second, how did you think about what visions of future warfare you wanted to portray?

Tony Stark: When I was in college, before I joined the infantry, I attended a defense tech conference where they were showing off various early-stage technologies. There were all these talks about the future of war, high-end conflict, gadgets, and push-button warfare. The following year, I found myself, almost to the day, trying to dig a ranger grave — a very shallow defensive position — in the backwoods of Fort Benning with a broken shovel. That experience embodies my perspective on future warfare.

I wrote something similar in the combined arms piece I recently published:

In 2025, a hypersonic missile can fly thousands of miles to strike a single target built on the backs of days’ worth of intelligence collection, analysis, deception, and SOF-enabled targeting behind enemy lines…and simultaneously, a few blocks away one guy can beat another guy to death with a shovel in the same war for the same piece of ground.

If you want to understand the combat operations in Ex Supra before reading it, that’s very much it. You have this high-tech fight, but the bloody, muddy reality is that you’re still going hand-to-hand. People are still dying in the mud. No amount of shiny technology is going to change that.

Jordan Schneider: Mike Horowitz sent me his History of Military Innovation thesis, and S.L.A. Marshall’s Men Against Fire was on it. I hadn’t read it beforehand. It’s such a strange book because the author is known for fabricating stories and exaggerating his battlefield experience. Yet, this book does an excellent job of capturing that John Keegan The Face of Battle approach in the World War II context, which feels very relevant to Ukraine and to other theaters today.

The reality is that despite the many drones in Ukraine, people are still sitting in trenches on the front line. Wrapping your head around the fact that this aspect is unlikely to disappear from warfare anytime soon — regardless of how advanced the Next Generation Air Dominance (NGAD) program is or how sophisticated robot technologies become — is important. There is a human element to warfare that has been with us forever and will likely remain for centuries to come, for better or worse.

Tony Stark: Another aspect of considering future warfare is not only what technology you think will work, but what technology you think will fail. I make this point about rail guns repeatedly in my story — that they somewhat work and somewhat don’t. Even the best high-end technology has shortfalls.

Jordan Schneider: One of my favorite World War II stories concerns American submarines whose torpedoes were simply broken. The trigger mechanism didn’t work until around 1944. All the submarine captains knew it and would report back saying, “Our trigger isn’t working. You need to fix this.” For whatever bureaucratic or acquisition-related reason, they kept being told, “No, you just need to be closer. You shot at the wrong angle. It’s your fault.”

This illustrates the future of war — having submarines capable of destroying aircraft carriers or battleships, but taking America three years to get its act together to truly leverage this technological advantage. Things don’t work as expected, learning happens at lots of levels, and AGI won’t solve these problems. We still need people writing unconventional analyses like yours, Tony, to help us conceptualize and think through the future.

Subscribe to Breaking Beijing and check out Tony’s book, Ex Supra, which is available now on Amazon and in independent bookstores near you!

Breaking Beijing

The Blueprint for How We Win the 21st Century

By Tony Stark

This is a guest piece by JS Tan, a PhD Candidate at MIT’s international development program, researching the political economy of innovation in the US and China with a focus on cloud computing. He was previously a software engineer and writes on Substack here.

Alibaba’s recent commitment to invest $53 billion in AI and cloud infrastructure — surpassing its total AI and cloud spending over the past decade — marks a major bet on the future. This investment signals not only Alibaba’s confidence in AI-driven growth but also its recognition that cloud computing will be a central pillar of the digital economy. Once battered by regulatory crackdowns and slowing profits, Alibaba is now doubling down on AI infrastructure, aiming to secure its place in China’s rapidly evolving tech landscape.

However, AI’s transformative potential — especially its diffusion across industries — ultimately depends on the strength of the underlying cloud infrastructure. Whether Alibaba’s aggressive AI push and massive data center investments will translate into higher profit margins for the company or a more competitive Chinese cloud sector as a whole remains an open question. And even though Alibaba is best positioned to play a critical role in AI's diffusion in the country, what’s missing from the FT’s analysis is a broader political-economic assessment of China’s cloud industry and the institutional factors that strengthen, or in some cases, constrain the country's cloud ambitions.

This article explores why low-margin compute and a non-existent SaaS ecosystem leave China behind in the cloud — and why this might not matter much to Beijing.

How Far Behind Is China?

To assess the political economy of China’s cloud sector, we must first understand the global competitive landscape. The U.S. cloud industry is dominated by three major players: Amazon Web Services (AWS), Microsoft Azure, and Google Cloud. China’s cloud industry follows a similar pattern. Alibaba Cloud and Tencent Cloud lead the market, mirroring Amazon and Google’s dominance. However, the Chinese industry lags far behind. A key measure of this competitive gap is revenue. By this measure, U.S. cloud providers dominate by a wide margin — AWS alone generates more revenue than the entire Chinese cloud sector combined. The disparity is also evident in data center capacity. The U.S. accounts for 51% of the world’s data center capacity, while China lags behind at 16%.

A major distinction between the U.S. and Chinese cloud sectors is the role of telecom firms, which have more experience, when compared to internet companies, at building physical infrastructure. Unlike the U.S., where no telecom company has successfully entered the cloud market — AT&T attempted but later sold its assets to IBM — China’s cloud sector includes firms with deep expertise in telecommunications. Huawei, China Mobile, China Telecom, and China Unicom have all invested heavily in cloud computing, leveraging their existing network infrastructure to expand into the sector.

Revenue and infrastructure alone do not fully explain the structural differences between the two cloud industries. A deeper distinction emerges when breaking down revenue by cloud service type: Infrastructure-as-a-Service (IaaS) vs. Platform-as-a-Service (PaaS).

IaaS consists of the basic building blocks of computing — compute, storage, and networking — delivered as virtualized infrastructure. Because IaaS is more about lifting existing IT operations onto the cloud, it requires less business transformation and is easier for firms to adopt. PaaS, by contrast, is more software-driven and involves deeper integration into enterprise operations. It is more profitable for cloud providers since it enables higher-margin services, such as AI platforms and business analytics.

While both the U.S. and Chinese cloud providers derive most of their revenue from IaaS, U.S. firms are far ahead of Chinese ones when it comes to revenue from PaaS. To this day, Chinese cloud firms are disproportionately reliant on low-margin IaaS, making its cloud providers more akin to commodity infrastructure providers rather than enablers of productivity-enhancing digital transformation.

Why Has China Fallen Behind?

A fundamental weakness of China’s cloud sector is low IT spending. A 2014 McKinsey report found that Chinese firms spend only 2% of their revenue on IT — half the global average.1 Historically, businesses have been reluctant to invest in IT due to high costs and uncertain returns. IT spending — whether on enterprise software, better infrastructure, or cloud services—is intended to improve productivity, but such gains aren’t immediate or guaranteed. Successful cloud adoption depends on structured data, formalized business processes, and standardized workflows — conditions that many Chinese firms lack.

Moreover, China’s low labor costs and weak labor protections reduce the incentive for businesses to invest in IT infrastructure. When labor is cheap, firms can simply hire more workers instead of upgrading their IT systems for uncertain productivity gains. As a result, many Chinese businesses use cloud computing only for basic, low-cost functions (such as website hosting) rather than as a platform for productivity-enhancing digital transformation.

These trends are reflected in industry composition. In the U.S., high-IT-spending industries—banking, finance, and business services — drive enterprise cloud adoption. In China, however, the economy is dominated by low-IT-spending sectors such as manufacturing and construction. These firms see little reason to invest in cloud-based enterprise software, further slowing adoption.

China’s cloud firms also lack two critical supply-side advantages. First, China has a weak enterprise software ecosystem. Because cloud revenue ultimately comes from other businesses, enterprise software is critical as it acts as a stepping stone for customers to adopt the cloud. Take Apache Spark, an analytics engine for big data and machine learning. While Spark itself is open-sourced and thus free to use, its widespread enterprise adoption in the U.S. has been driven by companies like Databricks, which help customers adopt the technology and even provide a "managed" version of the open-source software. And for customers that use Spark, moving to the cloud becomes a no-brainer, especially because Databricks has deep partnerships with Microsoft Azure and Amazon Web Services that make the transition to using Spark in the cloud seamless. In this way, companies like Databricks act as on-ramps to the cloud, helping businesses migrate workloads efficiently.

China lacks such a software ecosystem. As a result, Chinese firms not only have fewer incentives to migrate to the cloud but, when they do, they are more likely to use it for basic, low-margin IaaS rather than the higher-value PaaS offerings that drive profitability for U.S. cloud providers.

Second, China lacks professional IT service providers that guide enterprises through cloud adoption. In the U.S., a vast network of professional consulting firms — such as Accenture and Deloitte — plays a key role in guiding enterprises through this transition. These firms not only assist in moving IT infrastructure to the cloud but also ensure secure implementation and encourage the adoption of cloud-based software in daily operations. Acting as intermediaries between cloud providers and enterprise clients, they help streamline cloud adoption and drive demand for higher-value cloud services. In China, however, such professional services hardly exist, leaving businesses with fewer resources to navigate the complexities of cloud migration.

A Low-Value-Added Trap?

Recognizing these challenges, the Chinese government has stepped in. Since 2011, policymakers have allocated billions in subsidies to strengthen domestic cloud providers. The most ambitious initiative is “Eastern Data, Western Computing,” launched in 2020 to relocate energy-intensive computing tasks (such as AI model training) to China’s energy-rich western provinces while keeping latency-sensitive workloads (like AI inferencing) closer to users in eastern coastal cities.

This national-scale effort to coordinate cloud infrastructure with the country's energy capacity is no small task. To a large extent, what makes this coordination possible is the fact that China’s cloud ecosystem isn’t comprised solely of publicly traded private firms. Instead, the presence of state-owned enterprises (SOEs) such as China Telecom, China Mobile, and China Unicom allows the government to direct investment into infrastructure projects, even when short-term profitability is uncertain.

Beyond offering more flexibility in coordinating national-scale industrial policy, SOEs in China’s cloud sector encourage greater emphasis on low-margin IaaS over higher-value PaaS. While U.S. cloud providers have successfully moved up the value chain into PaaS, China’s cloud sector remains disproportionately reliant on basic infrastructure services, limiting its profitability and ability to compete globally.

Part of the reason is because China Mobile, China Telecom, and China Unicom are hardware-first enterprises with deep expertise in telecommunications infrastructure but limited experience in developing software-based cloud services. Their focus remains on expanding data center capacity rather than building out cloud-native software ecosystems. At the same time, IaaS — particularly compute infrastructure — is seen as a strategic resource, only one level removed from energy itself, leading the state to prioritize expanding capacity rather than fostering a robust software ecosystem. Just as control over energy production ensures industrial stability, control over compute capacity secures technological self-sufficiency and resilience in an era where data processing and AI development are becoming critical economic and geopolitical battlegrounds.

This leaves PaaS development primarily in the hands of commercially oriented firms like Alibaba and Tencent. However, these efforts have been punished as both the state and market has favored IaaS over PaaS. The result is a cloud ecosystem that is highly capable of providing cheap computing power but struggles to offer the advanced services that drive higher profitability and deeper enterprise adoption.

This is exacerbated by a fierce price war among China’s cloud firms over the past two years. In 2023, Alibaba launched the largest price cuts in its history, slashing core product prices by 15% to 50%. Within a month, Tencent, JD.com, China Mobile, and China Telecom followed suit with their own reductions. In 2024, Alibaba slashed prices again, prompting another round of retaliatory cuts from competitors. The presence of SOEs and privately held firms in China’s cloud portfolio has intensified these price wars. Unlike publicly traded cloud firms, which are under constant shareholder pressure to generate returns, SOEs benefit from state-backed patient financing, allowing them to sustain deeper price cuts for longer periods without the same risks faced by profit-driven private firms.

In Sum

China’s cloud industry is at a crossroads. On one hand, state intervention and expanding infrastructure could help China narrow the gap with the U.S. On the other hand, persistent price wars and an overreliance on IaaS — coupled with weak PaaS development—risk locking China’s cloud sector into a low-margin commodity industry. On top of that, China’s underdeveloped PaaS offerings will make it even harder for its cloud providers to compete on a global scale.

To the Chinese state, however, this may very well not be the goal. Data — and by extension, compute—are seen as critical factors of production, akin to land and labor, making the cloud too strategic to be left entirely to the private sector. If this remains the case, China’s cloud sector will remain structurally distinct from its U.S. counterpart — more state-directed, possibly less profitable, but ultimately aligned with national priorities rather than besting the U.S.

Read more

Read more

America, are we cooked? 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 @overshootMatt Klein, author of Trade Wars Are Class Wars and substack.

We discuss…

• The chaotic implementation of new tariffs and their impact on international alliances,

• Whether China will be able to capitalize on this opportunity,

• Why the US dollar hasn’t strengthened in response to Trump’s tariffs,

• Whether allies will choose nuclear proliferation over an unreliable USA,

• The resilience of America’s strengths despite governance failures.

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

Great Changes Unseen in a Century

Jordan Schneider: Peter, how do you feel the country is doing?

Peter Harrell: I actually don’t think we are cooked yet. There are enormous underlying structural strengths here in the United States that make me quite bullish about the future of America.

I should acknowledge that you caught me during a week of vacation in Southern Utah, where the weather is gorgeous and the rocks are magnificent. It’s hard to see us being cooked this week. If you ask me next week when I’m back in the office, I might have a different view on the situation.

Kevin Xu: We’re not cooked yet, but that probably implies some cooking is happening as we speak.

One line that got me thinking came from Josh Wolfe of Lux Capital, who runs a firm that I really admire. Early in the pandemic, he said, “Failure comes from failure to imagine failure.” For the sake of saving America or ensuring we’re not cooked, it’s probably useful to think hard about what “cooked” looks like so we can avoid going there — the aversion mental model of Charlie Munger.

This is something I think about more these days than before. That’s not to say we are cooked. There are many things in my personal and professional experience that suggest America remains special, if not exceptional, for all the right reasons. My current mindset focuses on imagining potential failures precisely to prevent them from happening.

Matt Klein: While I agree that we are not yet definitively cooked, we are definitely cooking. I concur with Peter that there are many long-term structural reasons to be optimistic about the United States economy and society. However, we have an administration that seems determined to attack all those strengths simultaneously. We’ve only been at this for about two and a half months.

If we were looking solely at the tariffs announced last week, I would agree that while they have negative consequences — clearly demonstrated in market reactions — that’s not the main reason to be concerned. What’s revealing is how these specific tariffs were chosen and rolled out, which is revealing of the overall governance process of this administration.

The Washington Post did thorough reporting on how the process actually occurred. Various teams of staff at different agencies — USTR, Council of Economic Advisers, Commerce, and others — were simultaneously developing methodologies for justifiable, reciprocal tariffs. Substantial staff work was being done. They were asking US companies abroad about their challenges and conducting extensive research.

Yet the final decision, made just two or three hours before the announcement, abandoned all that work in favor of the most simple-minded, least rigorous approach — a formula based on essentially two variables: exports and imports of bilateral goods with each country. This approach isn’t justifiable theoretically or in any other way. After people independently deduced what had been done, administration officials repeatedly denied it before eventually admitting the truth.

What does that tell you about everything else? This comes after we’ve already seen attacks on scientific research, extreme hostility toward foreigners entering the country, and a changed approach to international relations that includes threatening some of our closest allies. The tariffs imposed capriciously on Canada and Mexico are particularly troubling, considering this same administration negotiated and ratified a trade treaty with them just five years ago.

Looking at the broader context — the attacks on the rule of law, for instance — raises serious concerns. You can’t put a GDP value on the rule of law, but historically, the confidence that people can safely prosper in this country as long as they don’t harm others has been one of America’s structural strengths. When you add everything up cumulatively, you can understand why people might be worried.

Is the market reaction purely about the tariffs? Or is it a realization that all these things are adding up and potentially creating serious costs? That’s the big open question. We haven’t crossed the point of no return by any means, but there’s reason to be very concerned about whether we’re “cooked” or not.

Peter Harrell: Matt, don’t forget they also tariffed the penguins. In addition to tariffing the penguins, we then saw Howard Lutnick go out and justify it on the basis that maybe they would somehow get involved in rerouting trade from China or some other such place. This whole story of tariffing the penguins and then claiming you had a rationale because you’re worried the penguins will get into the transshipment business really doesn’t inspire confidence in current trade policymaking out of the administration.

Matt Klein: Right, there’s a decent chance they actually used a chatbot to come up with the exact tariff formula. Someone did a test — if you asked the free version of ChatGPT or Claude, how would you come up with it? They basically came up with exactly what the administration said, including basing it on internet domains as opposed to customs jurisdictions. Hence, the penguins.

Peter Harrell: Let’s also keep in mind this is an administration that has canceled most of its paid subscriptions to major outlets. Presumably, that is why they’d be using the free version of ChatGPT rather than paying for at least something that might have calculated the formula correctly.

Jordan Schneider: How weird can the next four years get? To what extent is this reversible by a Democratic Congress two years from now or a new president four years from now?

Peter, you were part of the “Let’s Make Friends With Our Allies Again” team in the early years of the Biden administration. What was that experience like?

Peter Harrell: I should begin by saying that I actually believe there are enduring, long-term American strengths. We have a very entrepreneurial culture that, by and large, values education quite highly. We have a pretty hard-working work ethic here. Even if the Trump administration is cutting federal R&D spending and grants, there’s a tremendous amount of private R&D in this country.

We start with a number of very strong underlying societal, business, and commercial strengths. The question is really, against these underlying societal strengths, how much damage can things like the very chaotic tariff policy we are seeing do?

I break that down into two different buckets. One is economically, how much damage can it do? Here, I don’t think we’re at the end of the story. We’re recording this on Monday morning. Maybe the market reaction is still getting through to Trump and they will begin to change course.

Particularly on the tariffs, I’ve argued for a couple of months now that this is all illegal. The statute that they are using for these tariffs simply doesn’t authorize them. As of last week, we are now seeing litigation against these tariffs, and we are more likely than not going to see the courts begin to curb some of these tariff powers. He’s still going to have a lot of tariff power even if the courts do curb them, but the courts will likely insist on some rational order being brought to the tariff agenda.

But you asked about the global damage. Aside from the economic harm, there’s also a geopolitical set of actions here. It’s really angered the Europeans, Japanese, and South Koreans. How does this impact our alliance structures?

There’s likely to be some durable damage to our alliance structures. Our Western European allies aren’t going to pivot to China — they understand it is not in their economic or social interest to join themselves fully at the hip to the Chinese. That would hurt their economy too. They’re going to need to raise some barriers of their own to China to prevent their own industries from being hit.

They will see Americans as less reliable allies going forward. That doesn’t mean they won’t cooperate. The analogy I’ve been using is that in four years, post-Trump, whether it’s another Republican or Democratic presidency coming in, many of our core allies will view it as, “Well, you just went through a divorce with us. We’d be happy to come up with a more amicable custody arrangement and see if we can be on better terms at parties, but we’re probably not going to move back in together."

This will be an enduring hit to our alliance structures, though it doesn’t mean that a future administration won’t be able to get a good degree of cooperation on areas of mutual interest.

Matt Klein: I would just add that this is happening at the exact same time, particularly with Europe, where we’re doing a complete heel turn on foreign policy grounds and essentially abandoning a lot of what they consider to be their core strategic interests. It’s not just the economic approach that would be offensive to them, but when paired with the approach to Ukraine and Russia, and the possible withdrawal of American commitments to NATO, and of course, the threats to annex Greenland from Denmark — viewed collectively, this creates some long-term damage. The damage extends to Canada too, which has historically maintained a very close relationship with the US for well over 100 years, and that relationship now seems to be severely damaged.

Jordan Schneider: How should we read the rhetoric we’re hearing from Mark Carney? Here’s a quote from him:

The global economy is fundamentally different today than it was yesterday. The system of global trade anchored on the United States that Canada has relied on since the end of the Second World War, a system that, while not perfect, has helped to deliver prosperity for our country for decades is over. Our old relationship of steadily deepening integration with the United States is over. The 80-year period when the United States embraced the mantle of global economic leadership, when it forged alliances rooted in trust and mutual respect, and championed the free and open exchange of goods and services is over.

Peter Harrell: There is clearly both a tactical and strategic part to Mark Carney’s comment that the relationship is over. Tactically, Carney is in the middle of a political campaign in Canada, so there’s a domestic political dynamic. He wants to send a signal domestically that he will stand up to Trump and won’t take being belittled the way he thought Governor Trudeau was constantly belittled by Trump.

He’s also signaling to the Trump administration, “We are prepared to push back. We’re not going to take this lying down. While we would like a negotiated solution, we are going to stand up.”

But there is also a deeper strategic point he’s making to Canada — “While the United States is very likely to remain our most important trading partner and our most important security partner, we here in Canada need to begin hedging our bets. We can’t be dependent on the United States to the extent we have been over the last 20, 25, or more years."

He’s sending a message that Canada isn’t going to cut everything off with the US — he does want to get a deal — but Canada will be looking to increase its trading connectivity with Europe and not just rely on the US market. Canada probably isn’t going to be reflexively in favor of agreeing to many of America’s asks when it comes to China. Canada will take its own interest with China into account. They don’t want to be flooded by Chinese exports either, but they might be looking for Chinese investment if their trading relationship with the US is weakened. They need to figure out where they can get capital. I hope they don’t go down that road, but Carney is beginning to signal he’s got to figure out what their long-term economic interests look like in a world where they see the US as a less reliable partner.

Matt Klein: The takeaway is that this represents a long-term cost — a long-term cost to Canadians, a long-term cost to the US. This is a standard textbook situation: you want to have the closest relationships with your immediate neighbors because distance is a challenge. It’s beneficial to have integrated manufacturing, finance, and other sectors between Detroit, Windsor, Toronto, and other border regions. If the conclusion is that Canadians think it’s better not to maintain that integration, then everyone will be worse off than in the alternative scenario.

Jordan Schneider: Kevin, do you think China will capitalize on this?

Kevin Xu: The Chinese economy still has many of its own problems that are in the process of being addressed. These include the property bubble and a weak consumer economy that they’re trying to stimulate.

Headlines are already emerging about stimulus measures, ratified from the National People’s Congress in March, now being accelerated from a deployment perspective now that the tariff numbers from both the US and Chinese sides are known. China needs to do many things for itself in this precarious situation.

There’s always this reflexive narrative — every time we consider whether the US is doing well or not, we automatically wonder if China will immediately fill that international space. I am personally very skeptical of any country being willing or able to fill the international role the United States has occupied for the last 30-40 years if the United States steps back.

What other countries have observed is that being the hegemon, being the global superpower that handles a lot of responsibilities, can be a thankless proposition. The tariffs have demonstrated to all countries that retreating to take care of your own national interests is now the fashionable approach. It’s acceptable to have no hegemon for perhaps the next 10 years. Everyone takes care of themselves.

Share

Whether that’s the economically optimal outcome, we all know the answer. However, politically speaking, that may be the most preferable path for all countries’ leadership, including China. There might be anecdotal evidence where China fills spaces or programs where USAID used to operate but no longer does. From the Chinese perspective, they’ll be very tactical about how they fill that vacuum and let the rest of the world, or the chips, fall where they may if we are indeed entering this new world order — or lack of order — which I believe we are stepping into right now.

Matt Klein: I’m curious what you make of Xi Jinping appointing the ambassador envoy to the European Union in Brussels — someone who had previously gotten in trouble with Europeans for suggesting the Baltic States didn’t really have a legal right to exist.

Jordan Schneider: This seems odd considering it would be a great time for the Chinese government to improve relationships with Europe, especially given what the US is doing. What do you think was the rationale for that?

Kevin Xu: First of all, we give Chinese foreign policy making far more credit than it deserves. There’s a narrative that portrays their actions as part of a masterfully executed 20 or 30-year plan, but that’s completely misinformed. Every foreign policy, regardless of political system, has a domestic component. Being strong and pro-China from a foreign policy perspective still has a lot of appeal.

We should be careful about over-interpreting the strategic nature of many of these moves. This could simply result from random internal political dealings — perhaps this official was up for a position, and this was what was available to him. His previous statements may not have factored into any grand strategic preference from the American perspective.

The EU-China relationship is probably the most interesting and dynamic one to watch. There isn’t an obvious trajectory where, because the transatlantic alliance has deteriorated, a Eurasian alliance will suddenly emerge. We’re in an era where most nations are primarily looking out for themselves.

Jordan Schneider: It’s worth recalling that China also has a recent history of bungling major policy tests with the COVID lockdowns. Xi is getting older, he’s a large man, formerly a smoker, and there’s no succession plan yet. That’s the risk people forget when they view China as an ocean of stability.

The CCP’s track record for power transitions is not particularly excellent. Sooner or later, we’ll witness that system trying to work around this issue, which could have big downside risks. For all the jokes about Trump’s third term, I don’t think we’ll go that far in America.

Peter Harrell: As we observe the market turmoil and chaotic tariffs, it’s easy to criticize Trump’s policies. Trade policy specifically is unlikely to yield positive results either in the near or long term. However, there is a smarter way to reform American trade policy. Matt has written about this over the years. A more thoughtful approach to American trade policy could actually benefit the United States without causing such chaos.

While we’re currently experiencing maximally chaotic Trump policies, there will likely be some initiatives from Trump and Congressional Republicans that will leverage American strengths and prove beneficial. As someone who served in the Biden administration, I acknowledge the criticism of our ability to actually build things over the last four years. While some criticism is exaggerated, I believe there is merit to the idea that we need a deregulatory agenda to facilitate more construction in the United States.

If Trump succeeds in reducing barriers to physical construction at federal, state, and local levels — whether for manufacturing facilities or infrastructure — that would be positive. Additionally, while I’m skeptical of tax cuts heavily favoring the wealthy, some stability in our tax regime could be useful. We obviously need to address this issue this year.

So while I’ve been critical of Trump’s trade policy and other aspects of his approach, including on Ukraine, I believe we may see some more valuable policies emerge over the next couple of months.

The Moron Risk Premium

Jordan Schneider: You need competent staff to create legislation, executive orders, and regulations. If we’re experiencing a complete policy lobotomy for the next four years, does that mean even the potentially positive initiatives in the Trump agenda won’t be able to flourish?

From 2016 to 2019, there were relatively few crises. With COVID in 2020, Trump might deserve a B+ grade, all things considered. Five years later, we’re forgetting about some of the more questionable pandemic suggestions. But if this is how the administration functions during a relatively calm period, four years provides ample time for outlier events to occur. The administration’s ability to navigate and respond to external shocks — let alone self-inflicted ones — has significantly diminished in my estimation based on recent weeks.

Kevin Xu: I’d like to expand beyond the Loomer situation specifically, which I know you discussed in a previous episode. This relates to what Matt mentioned earlier about one of my personal indicators of the “cooked or not cooked” divide — the arbitrariness in our civic life.

One of the enduring strengths of the American system isn’t just the rule of law broadly, but specifically due process. The fact that any decision, large or small, follows a process that’s impersonal and largely apolitical. Whether you’re getting your Social Security check or renewing your driver’s license at the DMV — which may be inefficient but treats everyone identically — this extends to the other extreme regarding what influences the hiring and firing of important government officials.

Having served as a political appointee, as Peter has too, we understand we serve at the president’s pleasure. Being fired isn’t catastrophic for us. However, civil servants are different for very specific reasons. When institutions that traditionally uphold due process are systematically undermined, that pushes us toward the “cooked” side of the spectrum.

What Laura Loomer has illuminated as a circumvention of due process troubles me most — not specifically who was fired, which I don’t have a personal assessment of, but the process represents the biggest warning sign.

Matt Klein: This is consistent with how they’ve been rounding up people who are either legally in this country or whose status may be uncertain, then immediately deporting them to places like El Salvador before they can prove their status. You could do that to a U.S. citizen, and if you act quickly enough, they can’t prove their citizenship either. That pattern is concerning.

Kevin Xu: Exactly right. There’s anecdotal evidence around me now of people who carry extra documents despite being American citizens when traveling, such as going on a cruise. They think, “I don’t look the way some people might expect, so I need extra documentation to ensure I can return to my own country.” At the ordinary citizen level, this erosion of due process has already taken root.

Peter Harrell: Regarding the layoffs — setting aside how the recent NSC layoffs occurred — let’s consider the dismissals of career staff across agencies. This will adversely impact the Trump administration’s ability to pursue goals that many of us on this podcast and those listening would actually support.

A couple of weeks ago, BIS Undersecretary Kessler, who oversees export controls at the Commerce Department, spoke at a conference about ramping up export controls enforcement on China — something I would support. However, the administration has now laid off a large number of BIS export control staff. As a straightforward American, I struggle to understand how they expect to increase enforcement while eliminating most of the staff.

This will have two impacts — the more important being the effects on Americans who believed they were entitled to procedural rights only to discover they don’t have them. Additionally, the Trump administration itself will encounter barriers to pursuing even broadly supported goals because it won’t have the necessary staff to implement them.

Jordan Schneider: Matt, do you have any leading economic indicators for us?

Matt Klein: The most leading economic indicator would be asset prices. Academics refer to this as both a mirror and an engine. On the other hand, it tells us people are obviously less optimistic about the outlook for various reasons and more uncertain. On the other hand, high and rising stock prices have been a support to both consumer spending and business investment.

When companies have higher stock prices, they feel more comfortable hiring people and investing in capital expenditures. They’ll now be less comfortable doing those things, which could have self-reinforcing real economic effects. Consumers who look at their asset portfolios and think, “I’m doing well, I can afford to spend more of my income,” will be less inclined to do so.

Taxes are due next week, and people who make quarterly estimated payments will need to write a big check. These payments are based on what you earned last year, but you have to pay from what you have now. This will be unpleasant for many people.

The hard data we’ve had on inflation, unemployment, and other indicators have been acceptable, but that information is already outdated. The most recent employment data from mid-March was actually quite good in some ways, but much has happened since then. Most other data we have is from February, making it difficult to assess the current situation.

Survey data have been quite pessimistic, though with variation — some surveys look worse than others. The economic impact remains unclear at this point. The situation appears concerning, but that’s somewhat anecdotal and easy to say when stocks drop 15% in three days.

Subscribe now

Jordan Schneider: Peter, do you have a nuclearization take for us?

Peter Harrell: The perceived lack of reliability of the US as a security partner will encourage a number of US allies and partners to take steps to increase their own self-defense capabilities across different vectors.

In recent weeks, several Eastern European Baltic nations have begun moving toward withdrawing from the treaty that bans landmines. They’re clearly signaling that, given Russian aggression and America’s potential pivot on Ukraine, they need more defensive weapons, including historically criticized ones like landmines. We’re already seeing evidence of countries revisiting aspects of their defense posture in light of these new geopolitical realities.

It would be quite rational for countries like Japan or South Korea to begin quietly considering what developing an independent nuclear capability might look like as a deterrent against Chinese or Russian aggression. Whether or when they will take that step remains uncertain, but it would be rational for them to begin contemplating it.

Similarly, a nuclear debate in Europe is emerging, but it will likely begin quietly before gaining momentum.

We are already seeing it, indeed. The question there is whether the issue will be developing an independent British and French nuclear umbrella that extends across Europe, or whether Germany needs its own independent nuclear capability. As Matt notes, we’re already witnessing that debate.

Matt Klein: Right. The Polish government has said that they now want to develop nuclear weapons.

Peter Harrell: Whether they’ll go nuclear in the next four years is uncertain, but we’ll likely see progress toward nuclear capabilities in Europe.

Kevin Xu: Returning to economic indicators, something that perplexes me — a factor on my own “cooked or not cooked” indicator — is the dollar’s value. Since the recent policy announcements, the dollar’s value has actually decreased slightly, which contradicts the theoretical best-case outcome of these policies.

The theory suggests that as trade barriers rise, the dollar should also strengthen against other currencies. This would shield some erosion of American consumers’ purchasing power while putting the government in a favorable position to refinance debt at lower rates, thereby reducing the deficit. However, this hasn’t happened yet. I’m curious about your interpretation of the dollar’s reaction. I have my own theories, but I’d like to hear your thoughts first.

Matt Klein: You’re right that the dollar’s movement was unusual and different from what many had anticipated. During the first Trump administration, officials claimed tariffs wouldn’t be as inflationary as critics thought because currency values would absorb some of the impact. Looking at what happened in 2018-2019 when they implemented tariffs, the dollar did rise to offset some effects. Earlier this year, when tariffs were announced on Canada, Mexico, and China, the dollar appreciated against the Canadian dollar, peso, and yuan, which seemed to confirm this pattern.

The textbook reasoning suggests that tariffs make investments in the United States relatively more attractive for selling to the US market compared to investments abroad. This should create financial inflows into the United States that strengthen the dollar. Yet that’s not what happened this time. Even with very large tariffs — essentially increasing our overall tariff rate from roughly 2% to 25% — which theoretically should make investment in the United States much more attractive, the dollar declined.

This outcome aligns with concerns about economic instability. Whatever theoretical attractiveness might come from these trade barriers has been offset by other factors. If investors believe the tariffs won’t be permanent or are uncertain about what the final levels will be, why would they commit to long-term investments?

There are arguments both for and against tariffs potentially boosting investment in the United States, but most would agree it depends on a commitment to policy consistency. This level of uncertainty is destructive to investment, compounding the fact that the tariffs themselves will negatively impact many existing investments.

Peter Harrell: What happened in the UK in fall 2022 provides an instructive parallel. During the very short-lived Liz Truss premiership, they introduced what was called the “mini-budget” — though there was nothing mini about it. They implemented major changes to taxes and spending that many considered irrational.

Part of what concerned observers was that they bypassed normal procedures established for presenting a budget in the UK. They have an Office for Budget Responsibility that’s supposed to provide independent forecasts, which they essentially ignored. As a result, British interest rates spiked dramatically and, despite rising interest rates, the pound sterling declined.

This prompted comparisons to an emerging market currency crisis. Someone coined the phrase “moron risk premium” — meaning the additional risk factor wasn’t about inflation outlook or deficits, but about the competence of the people in charge.

Something similar may be happening here. Whether the dollar will decline significantly from this point is impossible to predict, but the fact that it didn’t strengthen as expected, is indicative of something having happened

Jordan Schneider: The comparison of America exhibiting emerging market economy-level policymaking is perhaps the most succinct way to tie all this together. The problem is we haven’t had many global hegemons who transitioned from being functional to deeply dysfunctional.

Peter Harrell: The trade policy approach is truly striking. If the Trump administration had announced several weeks ago that they planned to launch investigations over the next six to twelve months — giving them authority to increase tariffs on Europe unless it provided better market access for our tech firms and reduced its car tariffs — the reaction would have been different. If they had focused on targeted tariffs to re-shore specific sectors we care about, economists would debate the pros and cons, but we would have seen an orderly process with a reasonable chance of achieving policy success.

Even if there were economic costs, we wouldn’t be witnessing these wild market reactions. But that’s not the approach they’ve chosen. Instead, they seem determined to turn the world upside down overnight and then deal with the consequences.

What’s particularly notable is the rhetoric over the past week describing this as “medicine” the American economy needs to adjust to. Historically, political leaders talk about economic “medicine” when facing a massive debt crisis or when the economy is in free fall, requiring tough choices to turn things around. You rarely see leaders advocating shock therapy for an economy that’s actually growing reasonably well. Typically, adjustments are gradual. This represents a historically unprecedented approach to economic policymaking.

Jordan Schneider: That’s true, although some people in the administration legitimately believe there was either an existing crisis or an imminent one. Whether you or I agree, that perspective might be informing their approach.

You’re right about the hypothetical alternative — an Earth Two possibility where we had a sane, rational, logical approach to restructuring trade relationships. But that’s not the world we’re actually in. Any closing thoughts?

Kevin Xu: Not cooked.

Peter Harrell: I agree, not cooked. We’re going to get through this.

Matt Klein: We’re not done yet. For all the damage we’re doing to our international relationships, if the US was able to become friends with Vietnam by the 1990s, maybe it’ll take some time, but we can be friendly with Canada again.

Jordan Schneider: Alright, I guess that means some Canadian outtro music as an olive branch to our friends from the north.