What will Trump mean for Asia and democracy? ​​To discuss we have on , who made time for an in-person interview with Managing Editor Lily Ottinger during his recent trip to Taiwan. He runs the Noahpinion substack and is the author of an upcoming book on the revival of the Japanese economy.

We discuss…

• The goals of Silicon Valley’s pro-Trump constituency, from deregulation, to tariffs, to China policy,

• Whether Elon is standing up for Taiwan behind closed doors,

• Whether Taiwan, Japan, South Korea, and Poland need their own nuclear weapons,

• How Taiwan could bargain for independence with Chinese leaders post-Xi,

• National health insurance as a potential solution to China’s aggregate demand problem,

• A Georgist perspective on China’s real estate problem,

• Why China’s demographic issues are overstated,

• Recommendations for Taiwan’s economic development.

What follows are excerpts from the interview — but we recommend listening to the full podcast for maximum fun. Here on iTunes or Spotify.

From Silicon Valley to D.C.

Lily Ottinger: Let’s talk first about presidential powers in the second Trump administration. You’ve already written about the restraints on Elon Musk — I’m interested in a similar question, which is, what do you think are the constraints on Trump from within his new coalition?

Noah Smith: Trump lost a lot of his old allies. Rudy Giuliani and company were discredited by their involvement with January 6th or various legal efforts to overturn the 2020 election. Trump is also very old, and he appeared to be running on fumes when Elon Musk and the tech right swooped in to bail him out. Trump might have won the election anyway just because people were so dissatisfied with Biden — but certainly, the tech right seems to be the most influential faction within Trump’s team right now. That doesn’t mean they’re omnipotent, however.

Lily Ottinger: What do you think the tech right will spend their influence on?

Noah Smith: The number one thing they’ll spend influence on is things that will make the business climate better in America. They are business people. You can call that corruption if you want, but it has a long history in America, China, Japan, and elsewhere. Keidanren’s influence within the LDP is not dissimilar.

Honestly, if you want my personal editorializing, we do need some of those things. There’s a lot of deregulation that needs to happen. It’s become clear that deregulation is the next frontier of economic policy, and it’s something we were unable to do for many, many years. Even Reagan was unable to do it. Carter was the last real deregulator. I guess Clinton deregulated finance, which blew up and killed the appetite for deregulation. Now we absolutely need deregulation in America. The tech right is going to push for that. Democrats will unfortunately resist it, but they will lose, and that’s good. I’m a Democrat, but the progressive love of regulation for regulation’s sake is just strangling America. Unfortunately, this is going to come with a small side of some kinds of deregulation that shouldn’t be done. Financial deregulation is often bad, but it allows rich people to cash out very quickly on their asset appreciation.

I’m highly optimistic that the tech right is going to get good deregulation done, but I think it’s going to come with a side of some bad financial deregulation.

Lily Ottinger: Do you think the tech right will oppose tariffs?

Noah Smith: That’s a really good question and I don’t know the answer to that.

Lily Ottinger: No insights from the parties in San Francisco?

Noah Smith: Everyone at the parties is saying nice things about tariffs, and that means nothing. They’re saying nice things because they feel like they are part of this winning coalition, and Trump says tariffs, so tariffs. But when it comes down to whether they want tariffs that actually impact the component sourcing for your businesses, things might get different very quick. You might see some push to make the tariffs symbolic behind the scenes. I’m speculating, but anything that hurts tech businesses is something I wouldn’t bet on happening. Now, one interesting thing about tech people is that, Elon is a hardware guy, but a lot of tech people work in software. Software doesn’t source a lot of components from overseas. The software people aren’t necessarily going to oppose tariffs because they don’t get hit by tariffs.

Lily Ottinger: They do get hit by restrictions on immigration, though.

Noah Smith: Exactly. That’s why you saw this titanic fight over high-skilled immigration, especially Indian immigration, on the right around Christmas time.

Lily Ottinger: You also wrote that export controls are going to be the litmus test for whether or not Trump has what it takes to stand up to China. Do you really think Elon will push Trump to sell out?

Noah Smith: It could be that Elon Musk just loves China and thinks that authoritarianism is great and he will rule the world as one of the three great authoritarian leaders in a new Metternich system that will crush global wokeness, which is something someone suggested to me the other day.

It could be that Elon Musk is just using China temporarily and knows that we need to oppose them. He notably hasn't turned over any SpaceX technology to China. For Tesla, it was clear China was going to just win on electric cars on its own. It could be a pragmatic move to say, “Taiwan is part of China,” because that’s what you have to say to do business in China. The NBA has to say that. It could be that he really resents having to say that — I know I would if I were him. It could be that he is preparing a counter-strike that will restore American power vis-à-vis China.

I don't know what his objective function is, and I don’t claim to know. As for what other people want — the pro-China faction is all finance. The finance guys have been downgraded in the Trump administration. Tech is very anti-China in general because China shuts out tech. China shut out Google, Facebook, and the like, and the tech sector rightfully fears China.

Finance just wants to make a quick buck off of investing in China, although the degree to which they think they can do that is falling rapidly. I talk to private equity guys — two years ago, they were saying, “What’s with this de-risking nonsense? What’s with this decoupling stuff everyone is talking about?” They were pouring money into China. Now you see all these news stories about private equity being kicked out of China, trying desperately to get their money out. It turns out that a modicum of foresight is useful in the investment world, so the finance guys are becoming less pro-China.

The reason Trump might sell out to China is personal. China, by all accounts, changed the TikTok algorithm to promote pro-Trump content. If TikTok is a pro-Trump media platform in America, Trump will not want to ban it. Jeff Yass, a billionaire who owns a lot of TikTok from the American side, contributed a ton to Trump.

Trump can be bought. Trump is a deeply corrupt individual and always has been. China can use its levers to control Trump as a man, as a person, in ways that it couldn't control the US Right or the conservative movement, and in ways that it couldn't control Elon Musk as a man, probably.

I doubt that Elon Musk can be controlled by China so easily. I don’t buy the idea that Elon Musk will just do anything to salvage Tesla Shanghai. To find out whether Trump is going to sell out to China, watch those export controls.

Lily Ottinger: Living in Taiwan, people often share their thoughts on American politics with me. Taiwanese people in general seem to feel pretty okay about Trump’s election. They’ll say things like, “Trump is crazier than Xi, which will prevent China from invading,” or they’ll point out that Marco Rubio has a record of standing up for Taiwan.

I’m not as convinced. Do you think this unpredictability is going to be an asset in foreign policy? Or do you think that Trump is now predictable to adversaries based on their experiences in the first administration?

Noah Smith: Unpredictability is a complete and total asset. But it is not the only thing going on. The corruption is a negative. Trump can be bribed. That’s bad. But the unpredictability is great because China’s leaders do not understand America at all. Their models of America are even worse than the models in the minds of Taiwan or Japan or countries that know us better. They’re totally in the dark.

The Case for Allied Proliferation

Lily Ottinger: Let’s talk about proliferation. You’ve argued that Japan, South Korea, and maybe Poland need their own nuclear weapons. Do you think the case for that applies to Taiwan also?

Noah Smith: No, because they can’t. If Taiwan had gotten nuclear weapons in the 1960s, that would have been great. America stopped them from doing it and therefore doomed Taiwan. Ukraine should get nukes if they can — that would stop them from being conquered.

We live in a world in which great powers once again see fit to conquer smaller countries. Russia thinks it’s their right to conquer Ukraine. Later, they’ll try to conquer Moldova and the Baltics, and maybe Poland, although they might just be content to externally bully Warsaw Pact countries.

That’s a big break in the world. Xi Jinping definitely wants to take over Taiwan, part of India, part of Japan, part of Okinawa, and certainly various parts of other countries he has designs on. If Russia suffered some kind of collapse, China would likely seize part of Russia — the area north of the Amur River that used to be Qing territory — “for safekeeping.” They already mark some of that territory as China on government maps.

Xi Jinping is an expansionist, though maybe not as much as Putin. That’s why Japan needs nuclear weapons yesterday. South Korea needs nuclear weapons yesterday. It’s insanity that they’re not getting them. That single policy change will stabilize East Asia more than anything else. Nothing the United States does is as powerful as those countries having nukes — nothing. Even if the United States maintained a full security commitment to protecting every inch of territory and fought to defend Taiwan, Japanese and South Korean nukes would still matter more.

Lily Ottinger: Are you worried at all about a preemptive violent response from China once it’s clear Japan and South Korea are preparing to go nuclear but before the bombs are actually completed? Do you think the CCP would try a preemptive strike or a Stuxnet-style hack?

Noah Smith: They can try the hack — it’ll fail. Those countries already have all the technology and materials to go nuclear very quickly. Those countries should exert lots of effort and care toward making sure their nuclear programs don’t get sabotaged and ensure facilities aren’t networked so they can’t easily be hacked.

If Taiwan goes nuclear, the invasion will start tomorrow, or China will use missile strikes to stop Taiwan from proliferating. But if Japan and South Korea go nuclear, China will not attack them. My guess is that China would not be able to stop them from going nuclear, and this would stabilize the situation in East Asia and ensure North Korea will not attack and conquer South Korea — and ensures China would not support North Korea if they tried.

Subscribe now

If Japan has nukes, Japan’s territorial integrity is assured. Great powers do not attack nuclear-armed states. They can sabotage them, erode their power, try to compromise them, and use all sorts of gray zone warfare to try to compromise them, but they do not invade them because they’re too scared of nukes. The downside is too large.

When discussing nukes, Japan and South Korea should build up hundreds of nukes — not five nukes. They should build hundreds, maybe even thousands, depending on how many China builds, but certainly as many as China has. For China, it’s not worth the risk of Qin Shi Huang’s tomb being vaporized. Ultimately, Japan and South Korea would lose a nuclear war because they’d get obliterated, but mutual assured destruction is a powerful deterrent.

Poland should probably get nuclear weapons at this point — it’s looking more like support for Ukraine is going to vanish. Poland said they would get nukes unless they were admitted into NATO. But if the NATO guarantee is gone because Trump won’t answer an Article 5 summons and won’t fight the Russians under any circumstances — Poland needs nukes. If the deal’s off, get the nukes!

No one is going to attack nuclear-armed states like North Korea or Pakistan. Iran won’t attack Israel in a way that would get them nuked — instead, Iran uses proxies to attack Israel, which are not doing well lately, and Russia just uses gray zone warfare against the European countries. India and Pakistan have basically calmed down.

Japan and South Korea need nukes yesterday. They need them right now. There’s no ambiguity. When discussing Japan and South Korea getting nukes, this does not mean American nukes. They should not simply station American nukes on their soil like Germany does. They need their own launch codes, their own nukes, 100% control of their own nuclear weapons. This will ensure their independence.

This isn’t 100% guaranteed protection, because you could still get such a madman in China that they would launch nuclear war preemptively. That could happen. Maybe Hitler would have done that — though I don’t know if we’ll get someone like Hitler ever again. But Xi Jinping would not attack a nuclear-armed state.

Taiwanese Independence and Xi’s Legacy

Lily Ottinger: Let’s say China attempts to invade Taiwan and fails spectacularly for some reason. Do you think Taiwan should declare independence after declaring victory?

Noah Smith: Yeah, why not? Taiwan won’t get another chance. I can’t see much of a downside because China already tried to invade at that point. Everybody likes a winner, so Taiwan would get the maximum support for its independence bid.

Lily Ottinger: I read a paper about this recently that said the opposite — it argued that the United States should tell Taiwan not to kick China while they’re down.

Noah Smith: If China’s leaders were reasonable, there would be room to strike a bargain.

1. China’s leader declares, “We’ve decided to grant Taiwan conditional, temporary independence.”

2. Taiwan declares independence as Taiwan, not as the Republic of China, and stops claiming to be the legitimate government of China.

3. Taiwan amends its constitution to say, “Even though Taiwan is currently an independent country, Taiwan is also a part of China and will reunify with China eventually, and the government of Taiwan is obligated to hold reunification talks with China every five years, indefinitely.”

Then they just do those talks every five years. Taiwan would support that bargain because it preserves the status quo. China would support it because it would allow them to claim that Taiwan concedes that they are part of China and agreed to be reincorporated, even as they grant Taiwan formal independence. America could help sweeten this deal by agreeing to withdraw military forces from the area, and Taiwan could agree to stay neutral and not formally ally with any foreign power. It’s the Finlandization of Taiwan.

Now, Xi wouldn’t go for this. Jiang Zemin might have gone for this, but those days are done — though they may come again.

Xi Jinping isn’t that great at running China. He’s good at wielding power over the CCP and being in control, but he’s not actually very good at running things.

Once he’s gone, and if someone competent takes over, there’s a possibility they could make this deal with Taiwan. The personal failings of Xi Jinping are much more of a factor in China’s problems than people from China would admit openly or than people outside China realize. Yet, Xi has the trappings of an effective person because he has taken credit for all the great things China has done through the efforts of others — through the entrepreneurs who built Huawei, BYD, DJI, and Tencent, and through the leaders who came before him.

The elites that Deng Xiaoping and his chosen successors, Jiang Zemin and Hu Jintao, picked to run the party were highly competent people. All those people built this incredibly powerful, incredibly effective Chinese state. Xi Jinping has taken credit for that and is prepared to spend that inheritance down — in an analogous way to how Putin is making war in Ukraine using tanks and artillery that the Soviet Union built.

Subscribe now

Putin has made us appreciate the Soviet Union more than we did. Yes, the Soviet Union was dysfunctional. Yes, it collapsed in the end. But before that, it built the greatest land army that the human race had ever seen. Putin is spending that army to devastate Ukraine incompetently, taking a few centimeters of Ukrainian territory for massive casualties and getting all his tanks and artillery from the Soviet inheritance blown up.

Xi Jinping is similarly spending the institutional inheritance built up by Deng Xiaoping, Jiang Zemin, Hu Jintao, Wen Jiabao, Li Keqiang, plus the private entrepreneurs. All these great figures of modern China built this foundation, and Xi Jinping is prepared to spend it down.

That’s a tragedy for the people of China. America’s engagement policy toward China didn’t fail as badly as most people now believe. China liberalized in many ways under Jiang, Hu, and Deng. They were still an authoritarian state, but they were a much more liberal authoritarian state where you could be a female Mao impersonator and no one would care. You could make “Jiang Zemin is the toad” memes. You had a pseudo-free press, civil society, and local elections.

Americans don’t understand gradations of democracy. They’re so far removed in history from their own democratization process in the 1700s and early 1800s that they don’t remember what it was like to build those institutions. They couldn’t see it happening in China. All they know now is “friend of America or foe” — that’s how they determine what’s a democracy and what’s not.

China was headed toward greater democracy, analogous to how Japan under the Taisho Emperor was headed toward democracy in the twenties. In Japan’s case, that progress was destroyed by civil unrest and right-wing cults. But in China, they went back toward authoritarianism for a simple reason — they got the wrong guy in power. That’s why I can’t go to China — because I say things like that. I’m not a China bear. I’m a Xi Jinping bear.

East Asian Healthcare and China’s Aggregate Demand Problem

Lily Ottinger: Let’s talk about China’s lack of aggregate demand. You’ve written that public healthcare spending is one form of stimulus that could help fix this problem. Would you like to explain that thesis?

Noah Smith: China’s healthcare system is patchy. One of the many ways in which China is like America is that they fear that a universal healthcare system will make the government too expensive and will make people decadent and complacent. Americans think this too. They’re wrong. But China and America both fear having the kind of healthcare system that other rich countries have implemented.

Freeing people up to spend because they don’t have to take care of grandma is this incredibly liberating thing. Social Security during the Depression was a big stimulus in America. Using the state to take some of the burden of elder care off of people’s hands allows people to go out, spend money, and boost aggregate demand.

I understand why China isn’t doing this, although it’s a bad reason. It’s the most obvious way to have the government boost consumption. Old Chinese people have very, very austere tastes because they grew up really poor. Have you ever seen that video where they have Chinese people and their American kids eating Panda Express?

The American kids are like, “Ugh, so disgusting, so low class, blah, blah, blah.” The Chinese parents grew up poor, so they are like, “Yum, calories!”

They have these very austere tastes. They don’t consume a lot. Giving a golden retirement to the Chinese boomer generation as a thank you for all their service to the Chinese state would be a really nice thing to do, and would free up younger Chinese people to consume on their own because some of the burden of elder care would be alleviated. China’s leaders, especially Xi Jinping, who is recognizable as sort of like a 2007-era Fox News dad, probably won’t do this unfortunately.

By the way, despite not being a “China hand,” I feel like I’ve managed to do a decent job predicting and understanding Xi Jinping himself — he’s basically a conservative boomer.

“Software isn’t real technology. It’s not a physical product. Only manufacturing is real technology.”

“Why do we have all these girly men on TV? We shouldn’t have girly men on tv. And now we have the gays? Why?

“You’re playing too many video games.”

“If the government pays for you, people won’t work for themselves and take care of their families like they ought to.”

“Stimulus — it’s a temporary sugar high. It’ll make people get decadent and consume more.”

Lily Ottinger: Do you have strong feelings about why the Japanese model of healthcare is the right one for China, as opposed to say, the Taiwanese model?

Noah Smith: I don’t know how Taiwan manages to have such cheap healthcare. Would you care to educate me?

Lily Ottinger: Taiwan started with a public health insurance plan for specific types of workers. They did a big audit based on the health consumption of all the people in that healthcare pool to set the first reimbursement prices once they started transitioning to a universal health insurance system.

The big difference between the Japanese system and the Taiwanese system is that in Japan, you have to pay a certain percentage — around 30% — as your copay for any health services.

Taiwan uses a prospective payment system to set copays. No matter what intensity of treatment you get, you pay a set fee to see the doctor regardless of whether you are having a hypochondria episode or are actually on death’s door.

Subscribe now

In Taiwan, the premium calculation is a payroll tax, which is split between people, employers, and the government.

Noah Smith: There’s just a set percentage no matter if you’re rich or poor? That’s cool — they have means-tested healthcare premiums.

Lily Ottinger: The Taiwanese government then sets the reimbursement rate — there’s a set amount that the government’s willing to pay hospitals for any service. If the hospitals negotiate with suppliers to get a better deal, they get to keep the rest as profit. In Japan, the government sets the prices. Providers can’t charge more, but they can’t charge less either.

Noah Smith: Right, Japan does price controls.

Lily Ottinger: Yes. In Taiwan, everything on the care side is privatized. The negotiations with suppliers become public eventually — every few years, the government does an audit to figure out where they can lower the reimbursement rates. If one city on the island gets a good deal from a certain supplier, that deal becomes public to the entire island.

Noah Smith: Japan is like that too, in some ways. The price control system is more rigid, but the national health insurer basically figures out what it can afford to pay. Medicare negotiating with people would be like this too, if we did that. The problem is this can hurt innovation if you do it. What you should do is address how much profit the suppliers are making, not how much they’re charging.

If you wring money out of the suppliers by forcing them to forgo innovation and be short-termist, then they’re free-riding on American medical innovation. You need to base your haggling on their profits, not on their prices. If they are investing a lot in innovation, that should mean they get to charge higher prices. That makes sense as long as they’re not using mislabeled innovation expenses to secretly pay themselves out, which they’ll try to do — they try to do the R&D tax credit right now.

Lily Ottinger: One more interesting thing about the Taiwanese healthcare system is that if the government suspects providers of inappropriate profit-seeking, they can pull the rug out from some of their revenue sources by suddenly repealing the prescription requirement for certain drugs. You don’t need a prescription for parasite medications here. You don’t need a prescription for arthritis medications, or inhalers, or birth control — whereas in Japan you do.

On Taiwan’s Development

Lily Ottinger: What recommendations do you have for Taiwan’s development?

Noah Smith: The constant threat from China has inhibited Taiwan from finding a new economic model. They haven’t been able to think in the long term and do long-term planning because they’ve been focused on this short-term threat from China. That’s unfortunate.

There is a way to dual-purpose economic development and resist the threat from China, which is to become an arms manufacturer. People have noticed that the United States can’t make anything anymore. Recently, there was a story about how Taiwan wants to be America’s new drone manufacturer. Taiwan doesn’t have much — it doesn’t really have a car industry, but it does have an electronics industry. Taiwan knows how to make stuff with batteries, and they’re able to do it.

Taiwan knows how to make drones. Taiwan as the arsenal of drones, the arsenal of batteries, and the arsenal of everything electric that you don’t want to source from China is the obvious play. Japan has ignored all this stuff. South Korea has ignored most of this stuff — Kia is building EVs, but they’ve really dropped the ball. They turned out to be more like Germany, wedded to heavy industry. But Taiwan has the same sort of electric-first orientation China does.

Taiwan needs to go hard for all the stuff China’s going for. Maybe if they still thought of themselves as the Republic of China, they would have already done this just because China did it. Nobody except China is building drones, nobody except China is building batteries. EVs might be a bridge too far. Get all those laptop contract manufacturers — Asus, Acer, Quanta, and all those companies — to make drones and batteries. Combine defense manufacturing and defense exports with electrification and electric manufacturing technology.

Lily Ottinger: What’s your favorite thing about Taiwan?

Noah Smith: My favorite thing about Taiwan is how laid back it is. I have never seen a country this chill. Yes, they’ll mess up your order sometimes. Yes, people mill about aimlessly in the train station. The rules I mentally apply from living in Japan just don’t work here. But it’s just a really sweet place, and people are just really chill. I hate the idea that China would blow up a place this chill — that’s just a crime.

大家听到这期熟悉问候的时候，这个世界已经迎来了2025年，北半球的大部分地区都已经进入到了深冬。不知道你听到的时候，你所在的地方有没有下雪？我所在的荷兰已经下过了雪。在这个冬季，我们决定Gap休息。

凛冬已至，严寒加剧，人如何度过这样的时间呢？我们决定休息一下，待到春暖花开再回归。不知道是不是开创了播客史公开宣布Gap的先例，但是我就是满怀激情地发出了这样的倡议。

希望所有处在同样的情况的朋友，也能勇敢且理所当然地Gap一下！当然，我们还要在这一期发布2025年的征稿，期待我们利用Gap的时间，能把这些2025年的征稿主题一起用富裕的时间想一想！

2025年播客征稿（截稿日期 2025年4月1号）

没关系，不知道麦克阿塞是谁也无妨。

总之你在2025年的年初收到了这个电话：“我们麦克阿塞基金会，决定给你一份为期5年的奖金，资助你做任何你想做的事情。”

是的，没有任何附加条件，你想做什么都可以。奖金一共80万美元，分五年逐步发放，你无需提出具体计划，也无需报告如何使用资金。你拥有最大程度的自由支配权。

现在重点来了：这5年，你想用这笔钱做什么呢？

曾经有人接到了这个电话，它第一天都非常兴奋，但从第二天起，一整周都情绪低落，这是它人生中唯一一次情绪低落的时候。它是《枪炮，病菌与钢铁》的作者贾雷德·戴蒙德，彼时它还未开始写书，仍在做医学和观鸟的工作。它情绪低落的原因是，它意识到这通电话实际上是说：“贾雷德，你是一个很有才华的人，我们给你5年既有自由又有报酬的时间，希望你好好利用这一自由，做点重要的事情。你的人生到目前而言，都在撰写关于胆囊和新几内亚鸟类的学术论文，没能发挥出你的潜能！”

那么你呢？你现在在做自己真正想做，发挥出你的潜能的事情吗？你被什么原因阻止了你做这些事情呢？给你5年的时间和80万的奖金，你想用它们来做哪些对你来说真正重要的事情呢？这也是放学以后“心愿清单”系列在2025的变身返场。

我从今天就开始想了，这样当麦克阿瑟打来电话时，我就已经准备好了，希望你也是。

欢迎你来投稿分享你对上述问题的答案，也让自己的声音被麦克阿瑟听见！

我们每个普通人，在绝大多数时候，都是世界，历史和时代的承受者。承受着“兴，百姓苦”，也承受着“亡，百姓苦”。时代的沙砾砸在我们的头上就是愚公也难移走的山，一滴水溅到我们身上就是精卫也难填的海。说好了“水能载舟，亦能覆舟”，这次我们来倾覆一下！或者修补一下！

做一天世界的产品经理，不再承受，勇于发起！不考虑任何现实中实现的条件，不考虑资源，成本，和投入和能成功的可能性，你只需要负责提出需求！

成立一个组织或者崭新的社群，为痛经问题提供彻底的解决方案，立法保护小猫小狗，化学加物理阉割性暴力犯罪者，给所有的小学生和中学生放放假，让所有的打工人有双休，让男性也能怀孕生子，让所有的手机统一一个充电头和耳机口，让全世界统一一种货币，甚至，给三体人发个信息？

只要这是你真诚的需求和渴望，都欢迎你的投稿！我们一起，be a proactive activist（做个主动的活动家）:我想要主动发起，我想要邀你参与！用你的发起活动，来让你的同类，和整个宇宙听见你的声音，收到你的订单!

欢迎来稿分享，自己最想对这个世界发起的一项活动或者一项倡议，以及想要这么做的原因。把它当做一个电梯演讲（Elevator Pitch），造物主正在和你共乘电梯，请用你的发起来打动一下她！

我们举办过“爹和爹味，吐槽大会”，以及“吐槽八卦大会”，每一次都给我们自己和收听的朋友带来了许多的欢愉。因为吐槽是一种释放，幽默且具攻击性的吐槽是一次破坏加重建。人被压缩紧皱的心脏和灵魂，在吐槽的过程中重新舒展吸水变得饱满。这一次，我们向西西弗斯的石头发起吐槽：工作和上班真的是吐槽素材的富矿！以及各种光鲜亮丽的行业和工作，真实在其中工作的人每天都在屎上雕花。在各种大厂大机构，官僚主义的笑话能写新时代汗牛充栋的《儒林外史》。

欢迎各位打工人和前打工人来稿分享你的吐槽，和你亲身经历或者观察到的，荒谬的打工生活及行业现象.当牛马开始吐槽身上的鞍鞯jiān绳索，那绳索也会发抖！Workers of the world, Unite and Roast!世界的无产者，联合起来吐槽吧！

人到底为什么不安于一隅，非要出门游荡呢？三毛给出的答案是为了天空飞翔的小鸟，山间轻流的小溪，宽阔的草原和梦中的橄榄树。在“月亮和六便士”中左右踟蹰的人是为了寻找一处它们自己也不知道坐落在何处的家乡。在从未寓目的景物里，从不相识的人群中定居下来，找到属于自己的宁静。

我出门游荡是为了什么呢？除了如上一切和美食美景的美好，我还为了辽阔的可能，为了那种走在大街上和山野中，感受到可能性纷至沓来的希望，像李娟想在乌鲁木齐的街道上张开双臂走一样，我想在这复杂缤纷的世界张开双臂走一走，去从生理上相信地球是一个圆，持续往西就能到达东方，去看看峡湾地貌和彩虹山长什么样。更关键的是，我想看看世界各地区人们不同的生活方式，想去发现人只要不侵害它人利益，自己怎么活都可以。以及“从来如此”并非世界都如此，且绝大多数“从来如此”不仅不对，且大错特错，是造成我们痛苦的主要枷锁。而发现世界和发现自我总是并行而来，我也想借由发现世界发现一个更缤纷有力包罗万象的自我。

Wander for the wonderful，你出门游荡是为了什么呢？欢迎你的来稿分享，让我们用你独特的眼睛和视角，再去看看这个世界的一切辽阔和美好。

让我们问出一些问题，那些真正的，重要的，不应该被回避的问题？那些理所当然却经不起推敲的问题？那些从来如此却不该如此的问题，问问世界，也问问自己，问问过去，现在和未来。

也问一些别人觉得不重要，但是我们好想知道答案的问题 。

问一问why and why not？

为什么太阳如此炙热？为什么树木可以生长？为什么奢侈品和消费陷阱会存在？我为什么还没有成为自己想成为的人？国家和政府为何存在？为什么贫富差距如此之大？为什么我们会过得如此辛苦没有活力？为什么这个世界时不时就有战争？为什么有的国家富有而有的国家贫穷？学习数学到底有什么用，可不可以不学，怎么才能学好？宇宙中到底有外星人吗？

让我们不停地追问一下：有没有，是不是，为什么，以及为什么不呢？

希望，活力，和生命力，都在追问中来。

欢迎来稿分享你对这个世界的疑问和追问，你为什么对这些问题感兴趣，以及你对某些问题找到的答案。

这个世界绝大部分节日是给活人的，有一些节日是给半死不活的人的，只有非常稀有的节日，是给已经离开这个世界的人的。我们向神祈祷，供奉祖先，对鬼恐惧，避之不及。逝去的人在我们的脑中变成了不同的形态，我们很少与之对话，总是要么祈求要么远离。在这个死气沉沉的时代，在这个尸斑传染的线上空间，我们想创造一个节日让大家和真正的死人说说话。

这是一个赛博亡灵节，我们一起举办一场赛博法事，不仅超度亡魂，也超度一下死感很重的自身。你想和哪个已经逝去的人说说话，朋友，家人，先贤哲人，某个遥远的历史人物，某个喜欢的创作者，请打开手机录音，把你想说的话录下，我们把这些录音集结在一起录制，用创作的火光把它烧给提早离开的人。

以上所有：截稿日期 2025年4月1号

【投稿方式】时长5分钟左右为宜，手机录音即可，录制时可将手机垫高，与嘴平齐，收音更清晰。音频请发送至 afterschool2021@126.com .（提示注意避免距离手机太近容易喷麦，或背景声音嘈杂收音不清晰，以及请保护好个人信息，避免透露个人ID，如有昵称可以用昵称投稿）

【温馨提示】请投稿发送前谨慎思考，如需变声处理也可提前处理好。如非人身安全或隐私威胁等重大原因，一般在已经录制上线后无法予以撤稿，敬请理解。

【本期播客Timeline】

00:40 放学以后的新年问候：这个冬天，我们一起Gap一下！

06:21 放学以后2025年征稿：一些让人对2025充满期待的话题！

24:04 当莫不谷提议Gap，霸王花和金钟罩听到后的真实反应是？

33:29 为什么决定播客Gap？袒露直接原因和根本原因

51:43 系统性的问题需要系统解决，靠热情，热情会被摧毁，靠良心，良心会被辜负

63:06 霸王花：为什么我最终决定服药治疗抑郁？吃药会有好转吗？

68:39 run&rebel后的生活会更好吗？

71:04 为什么要给自己的情绪开闸放洪的机会？

72:23 无论是人生选择还是人际关系，我们都有软着陆的可能

80:22 面对回避型人格，只有远离这一种可能吗？

86:22 金钟罩：播客Gap三个月期间准备要做的事情

94:24 莫不谷：每个人都有自我倾吐的欲望，要找到渠道去抒发

97:55 霸王花新增一个Flag，每周一剪辑更新放学以后YouTube！

102:07 不必担心Gap没有内容：推荐可以长期陪伴大家的两期精彩播客！

103:00 游荡者（www.youdangzhe.com）开发进展以及预告Workshop计划！

107:00 祝福大家，在这个冬季好好休息，储存能量，待到春天，我们再见！

【为全球华人游荡者提供解决方案的平台】

游荡者（www.youdangzhe.com），注册完成后可免费阅读由莫不谷和霸王花撰写的三篇文章（Run的800种可能、语言攻略和全球签证攻略），目前游荡者平台已更新上线文章分区功能（游荡区、学习区、欢愉区和闲聊搭子区），欢迎大家注册完成后开启内容创作并在游荡者游荡愉快！找到同类！交易自由！手机端用户可把新网址添加桌面，便于日常使用。在使用新网址期间如果有任何注册、支付、退款等需求，欢迎给我们客服邮箱wanderservice2024@outlook.com发送邮件。

放学以后Newsletter《新的一年会好吗？答案在这期播客和这些祝福中》

放学以后微信公众号《新的一年会好吗？答案在这期播客和这些祝福中》

【播客&文章&书籍&影视】

播客：

放学以后播客：活在XX的垃圾时间，我们如何度过历史的乱纪元？

方法1：解锁放学以后Newsletter月度会员https://open.substack.com/pub/afterschool2021/p/ed2?r=pilpv&utm_campaign=post&utm_medium=web；

方法2：解锁放学以后Spotify月度会员https://creators.spotify.com/pod/show/afterschool2021/episodes/ep-e2sscug。

方法3：爱发电单期解锁：https://afdian.com/item/90682ea4c68611ef8e645254001e7c00；特别提示，可通过浏览器打开网址将网页添加手机屏幕使用，不要花钱下载App Store里的爱发电软件！

放学以后播客：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？

放学以后播客：游荡世界的心愿清单，那无穷的远方也和你我有关

放学以后播客：假如可以不工作gap一年，我们要去做什么？

放学以后播客：我们挣钱做什么？10个人的心愿清单和金钱彼岸

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》（提示安卓和苹果用户可把爱发电网页版主页添加至手机桌面来使用，目前爱发电未上线苹果商店，谨防诈骗）

书籍：

《枪炮、病菌与钢铁》《共产主义宣言》《太白金星有点烦》《月亮与六便士》《我的阿勒泰》《我的痛苦有名字吗》《三体》

影视及音乐

电影《年会不能停》《好东西》

歌曲《橄榄树》

原创MV《东亚忍者之歌 Song of East Asian Ninja》莫路狂花出道曲！

《⁠东亚女游荡之歌 Songs of East Asian Female Wanderers⁠》

【延伸信息】

永不失联Newsletter订阅链接：https://afterschool2021.substack.com/（需科 学/上 网）

为全球华人游荡者提供解决方案的平台：游荡者（www.youdangzhe.com）

联系邮箱：afterschool2021@126.com （投稿来信及合作洽谈）

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

片头曲：《寄生兽》Bliss

片尾曲：《Miracle》Caravan Palace

播客封面：莫不谷用Canva制作

放学以后表情包：微信表情包搜索“放学以后”，感谢萝卜特创作。

小红书：游荡者的日常

播客收听平台：

【国内】爱发电、网易云、苹果播客（请科学/上网）、喜马拉雅、汽水儿、荔枝、小宇宙、QQ音乐；

【海外】Spotify、Apple podcast、Google podcast、Snipd、Overcast、Castbox、Amazon Music、Pocket Casts、Stitcher、Radio Public、Wordpress

Xiaohongshu favorites

Contributor Yiwen Lu reports:

We have seen tons of news stories this week about Red Note (小红书). Welcome to our favorite social media, TikTok refugees! ChinaTalk has always been scavenging Red Note for gems, such as how to purchase banned NVIDIA chips.

Without repeating what you have seen elsewhere, we compiled a list of the funniest interactions across our feeds.

The comment section under this post: Chinese — give us your American jokes. Please make fun of us so we can laugh.

The homework exchange goes both ways.

I’m obsessed with Billy from Florida, who learned about 甄嬛传 Empresses in the Palace and apparently classical Chinese as well. Here’s Billy’s Jing Hong dance, Zhen Huan’s signature dance in the show.

Jordan Schneider: fun while it lasted! I’d give it a 6/10 relative to 2021 Clubhouse moment, where you got to bear witness to incredible cross-strait discussions and minority voices from Xinjiang. See our podcast from that time on Spotify or Apple Podcasts.

The Rednote saga is another example of the latent demand in China and the US for real people to people communication, not whatever this KPI encompasses…

…and how the Chinese government is scared of it. From The Information:

Read more of Yiwen’s XHS reflections here:

Uncool Kids

Chinese internet as an identity

You don’t need more words from me to describe how popular the Chinese app Red Note, or Xiaohongshu, has become overnight…

Read more

4 days ago · 9 likes · Yiwen Lu

Jimmy Carter gave the US the chance to protect Taiwan while managing relations with China

Pete Millwood is Lecturer in East Asian History at the University of Melbourne. He is the author of Improbable Diplomats: How Ping-Pong Players, Musicians, and Scientists Remade US-China Relations, recently released in paperback.

Establishing diplomatic relations with China was, alongside the Camp David Accords, late President Jimmy Carter’s most substantial foreign policy achievement. As the United States pursues peer competition with China and rues what is remembered as the one-sided benefits of engagement in decades past, we might conclude that Carter should have driven a harder bargain in negotiations with Beijing before diplomatic recognition, or even that Carter was wrong to normalize relations at all. In fact, though, Carter won a critical concession from China that has helped to protect Taiwan even as the United States has an official relationship with the world’s most important post-Cold War actor outside of America.

Before 1978, Chinese leaders had told their American counterparts that diplomatic recognition would only come after the US government totally severed its relationship with Taiwan. Private trade and societal contacts could continue, but all official interactions must cease — including weapons sales. President Nixon and Henry Kissinger had gone a long way to meeting the Chinese position as early as 1971 and privately had concluded they had little chance of persuading China to accept anything more than the United States expressing its hope that China “reunify” with Taiwan peacefully.

After Carter defeated Republican Gerald Ford in the 1976 presidential election, his incoming administration asked to read the secret records of Nixon, Kissinger, and Ford’s negotiations with Mao and Zhou Enlai. Perhaps conscious of what they’d given away in early, excited visits to China, Nixon and Kissinger obfuscated. They only provided the records when Carter’s government threatened to sue. Carter was shocked by what he read. “We should not kiss-ass them the way Nixon and Kissinger did”, he concluded.

Subscribe now

In the first summer of his term, Carter sent his Secretary of State, Cyrus Vance, to China. In part because Carter knew he’d soon have to shepherd the Panama Canal treaties through Congress, the president told Vance to put forward the maximum position on Taiwan suggested by Kissinger at the height of the Watergate crisis in 1974, asking Chinese leader Deng Xiaoping to agree to continued US governmental ties to Taiwan after diplomatic normalization, perhaps similar to the semi-official ties that the US had with the People’s Republic at the time. Carter added a further request: to allow the United States to continue to sell arms to Taiwan after recognition.

On the face of it, Vance’s suggestion was flatly turned down by Beijing. Today, when Vance’s trip is recalled at all, it is remembered as a failure. But, back in Washington, Carter’s National Security Advisor Zbigniew Brzezinski had noted what he called a “loud silence”: Deng had said nothing about arms sales. This was the first suggestion that Carter might be able to square the circle of establishing diplomatic relations with China and preserving a meaningful say over the future of Taiwan.

Like Nixon and Kissinger before him, Carter saw real benefits to a normalized relationship with China, the term used to refer to full diplomatic relations. Brzezinski was born into a Catholic family in Warsaw and his father had been a Polish diplomat posted to Moscow during Stalin’s Great Purge. Carter’s National Security Advisor was a virulent Cold Warrior and helped persuade Carter to move away from the US–Soviet détente of the Nixon era and toward renewed confrontation with Moscow. This would become transparently clear following the 1979 Soviet invasion of Afghanistan, with Carter funneling weapons to the Afghan mujahideen and boycotting the 1980 Moscow Olympics. But Brzezinski’s arguments for challenging the Soviets found an earlier expression in Carter’s push for closer ties with Beijing.

One of the primary means through which closer ties were achieved was the transfer of scientific expertise and physical technology to China — including technology that had potential dual-use applications in both civilian and military sectors. As early as 1975, Deng Xiaoping had complained to Kissinger and President Ford about US export controls on the most advanced American technology. Deng said that, if US leaders wanted to move the relationship forward, they should reconsider blocks on China buying top computers. China wanted to use the computers for oil exploration, but the US government knew they could also be repurposed for conducting rocket tests and even nuclear weapons calculations. Kissinger quickly responded to Deng’s complaint, opening a new channel for circumventing the US government’s own restrictions on transfers of sensitive technology to communist states. After Ford’s election loss to Carter less than a year later, Carter continued and then deepened this policy of selling more powerful technology to China than would be transferred to other communist states including the Soviet Union — even before diplomatic normalization with Beijing.

Accelerating Chinese technological modernization helped strengthen China’s economy and military ahead of a possible future military clash with the Soviets. Seen in the context of China’s subsequent rise to economic but also military might, this triangular Cold War logic might seem short-sighted. In fact, Carter’s administration saw that more distant future coming. They knew that China saw the US as “a long-term adversary” and internally Carter’s top China advisers pledged not to “play Santa Claus”.

The United States sought reciprocity for this technology transfer in the form of normalization talks. Deng was the Chinese leader most strident in pushing for access to US expertise. Alongside his requests for physical technology, he also made Carter’s top science advisor, Frank Press, call Carter in the middle of the night to agree to receive 5,000 Chinese students as soon as possible. Carter barked back that Deng could send 100,000. Within five years, Deng had.

But Deng was, as Brzezinski had noticed after Vance’s 1977 China trip, the one leader willing and able to make compromises on normalization terms. In the final tense months of negotiations toward recognition, Deng continued to offer loud silences on whether arms sales to Taiwan could continue. The US side agreed to halt sales in 1979, the year after recognition. Other Chinese interlocutors, like Ambassador Chai Zemin who headed China’s Liaison Office in Washington, took this to mean a permanent cessation of sales. But Brzezinski had visited China in May 1978 and believed he’d got Deng to tacitly consent that weapons would continue to be sold again from 1980. After all other normalization terms had been agreed upon in December 1978, US Ambassador Leonard Woodcock met with Deng to clarify that the United States would resume arms sales. Deng said that China “cannot agree” to future sales — but he also consented to normalize regardless. In a press conference following the announcement, US officials said clearly that sales would continue.

And so they have. Between 1980 and 1987, yearly US deliveries of weapons to Taiwan did not exceed $400 million. But by the mid-1990s they were worth more than a billion dollars per annum on average. Major multi-billion-dollar deals in the Obama years were followed by a further spike under Trump.

China vehemently opposes these transfers, but there has been little that Beijing has been able to do to stop the sales, even when the bilateral relationship was close in the 1980s. Deng tried to reopen the issue of arms sales after recognition. Negotiations led to the 1982 communique between the governments issued by Carter’s successor, Ronald Reagan. Reagan offered vague assurances that weapons sales to Taiwan would decrease over time — if tensions between China and the island did — but Reagan steadfastly refused to give an end date for sales.

Only now can we fully see what Deng gave away and what Carter won. Normalization was a success for Deng personally and for China. But Deng also allowed the United States a tangible means to exercise a say in Taiwan’s future, while also maintaining a relationship with China.

Share

Some in the Nixon and Ford administrations were unconvinced that Taiwan’s continued viability was a fundamental US interest. Yes, Taiwan was a US ally — the alliance only ended after US recognition of the PRC in 1978 — but the Chiang regime had been useful primarily as a shared enemy of Communist China. Thus, the island appeared unimportant when Kissinger seemed to have made a friend of Beijing. The Chiang regime was also, in the 1970s, still brutally authoritarian, with its secret police murdering its opponents — even, in one case, outside their Californian home. Some officials might have been puzzled why Carter had worked so hard to maintain the US relationship with the island.

But then Chiang Kai-shek’s son, Chiang Ching-kuo, realized that the only hope for his regime after de-recognition was to win broad American support all over again, including for the nature of the Taiwanese government. Chiang went from Taiwan’s most feared policeman to the president who oversaw — however begrudgingly — the island’s democratization. Beijing seethed, but the 1996 Taiwan Strait Crisis showed it could not prevent Taiwanese self-rule. Carter probably did not expect Taiwan to become a democracy, even as the president who headlined human rights in his foreign policy, but his negotiations in 1978 helped to provide the space for this transformation.

US diplomats today are facing the same challenge Carter did back then: how to manage US-China relations while also preserving Taiwan’s political autonomy. Carter’s steadfast negotiating position combined with a willingness to leverage American strengths at the opportune moment laid the groundwork for a stable, normalized US-China relationship. In doing so, Carter gave Americans today the chance to find the same successful compromise he did.

Uber Eats’ Kinmen Expansion

Scooter-based food delivery is one of Taiwan’s many charms — and the industry has recently been in the spotlight thanks to Uber Eats’ attempt to acquire Food Panda, a regional competitor with better branding.

Taiwan’s government blocked the acquisition based on antitrust concerns, a move celebrated by Taiwan’s food delivery couriers’ union.

In response, Uber Eats is expanding service to residents of Kinmen, a Taiwanese island that lies 10km off the coast of China — and celebrating with pun-filled promotions. For example, coupon code 金選美食 “Kinmen-chosen gourmet food” sounds like 精選美食 “cream-of-the-crop gourmet food.”

Apart from the loss-leading discounts, the expansion might run into another problem — Kinmen is inhabited by Formosan rock macaques, a species of monkey known to attack delivery drivers and steal food. Hopefully, the union will mandate wildlife-related safety guidance for new drivers.

A college student in Kaohsiung brandishes a BB gun at monkeys attempting to steal her food:

Biden’s AI Infrastructure Order: The Right Vision with the Wrong Standards

Thomas Hochman is the Director of Infrastructure at the Foundation for American Innovation. You can read more of his writing at Green Tape.

Earlier this week, the Biden administration released its long-awaited Executive Order on Advancing United States Leadership in Artificial Intelligence Infrastructure. The order outlines an ambitious plan to build frontier AI data centers on federal land, establishing one of the most significant federal interventions in AI development to date.

The EO directs the Department of Defense and Department of Energy to each identify at least three federal sites suitable for frontier AI data centers by February of this year. These centers must be operational by the end of 2027 and be matched with clean energy generation sufficient to meet their electricity needs. Companies building these centers will be responsible for constructing both the facilities and the energy infrastructure needed to power them.

The order’s clean energy requirements will be a huge technical and economic challenge to overcome, particularly given the tight two-year timeline. There is an overwhelming consensus that powering the gigawatt-scale data centers of the near future will require a significant amount of natural gas — nuclear plants often take upwards of a decade to bring online, utility-scale battery storage for solar is still scaling up, and promising technologies like multi-gigawatt geothermal remain years away from widespread deployment.

The EO does allow for fossil fuel (i.e. gas) generation… if and only if it’s paired with carbon capture technology.

Taking this path won’t be simple in practice — the order states that any fossil fuel generation must achieve annual carbon dioxide capture rates of 90% or higher, a standard that has never been achieved at scale.

While 90% capture is technically possible, there’s no precedent for maintaining such high capture rates at megawatt scale for extended periods. Folks in the industry have told me that, for megawatt-scale, post-combustion natural gas, no entity has ever achieved a full month of 90% capture rates, let alone several years. While recent incentives like the expanded 45Q tax credit will help promote carbon capture innovation, it seems unlikely that massive breakthroughs will occur so rapidly.

But all of these restrictions could soon become moot — the incoming Trump administration will likely eliminate the order’s clean energy requirements while maintaining its other provisions, allowing these frontier facilities a realistic path to move forward. If the goal is actually to advance U.S. leadership in AI infrastructure as the executive order’s title suggests, this would be a very good thing.

The executive order proclaims that all permits and approvals required for construction should be issued by the end of 2025 and also directs the Department of Defense to conduct a programmatic environmental review of AI data center effects before the site solicitation process closes in June. Both of these timelines are extraordinarily ambitious, and the order doesn’t provide much authority to meet them beyond some hand-waving at “agency coordination” and “permit prioritization.”

As I’ve written, permitting a Manhattan Project-style AI push will require leveraging every statutory exemption and workaround we can possibly find. To this end, the Department of Defense has exemption authorities at its disposal that could become critical vectors of progress. The DoD can invoke Section 7(j) of the Endangered Species Act to avoid consultation for projects deemed necessary for national security. NEPA’s “alternative arrangements” provision allows agencies to bypass standard NEPA procedures in emergencies, including in national security contexts, and 40 C.F.R. § 1507.3(d) allows agencies to make the details of their NEPA documents confidential, effectively making it impossible to sue.

Share

Additional tools become available in the case of a whole-of-government push. These include presidential exemptions for both the Clean Air Act and Clean Water Act requirements when projects serve the “paramount interest of the United States.”

If you put all of these together, you can imagine a scenario where the government can effectively transcend every major permitting requirement for AI energy development. While using these authorities for AI infrastructure would represent a significant expansion of their traditional scope, the EO seems to lay the groundwork for this possibility.

As currently written, Biden’s EO isn’t much to look at. But if and when Trump loosens its clean energy requirements, the order could provide the foundation for a national AI infrastructure strategy that keeps pace with the underlying technology.

They called him “Sleepy Joe.” They said he couldn’t get anything done. And then, in the last week of the administration, Biden made us do two emergency podcasts in one week!

To discuss yesterday’s export controls, ChinaTalk interviewed Greg Allen from CSIS.

We get into…

• New foundry requirements that attempt to shut down the TSMC-to-Huawei pipeline,

• How Commerce’s redefinition of “DRAM” closes a major loophole outlined in our last podcast with Greg,

• Whether this rule will bankrupt chip design startups,

• Whether Biden-era regulations will have staying power across administrations,

• The qualifications of Jeffrey I. Kessler, Trump’s pick for head of BIS,

• What this week’s export control package says about the IC’s timeline for AGI.

Shell Companies, Shattered

Jordan Schneider: Greg, how are you holding up with this export control bonanza?

Greg Allen: I am running on fumes, but if the public thought we were too tired to record a podcast, they were sorely mistaken.

Jordan Schneider: What came down the pipe today, Mr. Allen?

Greg Allen: This new rule was really the missing piece in the big update that just came out. For the uninitiated, the AI diffusion rule came out on Monday and is designed to try and stop large-scale AI chip smuggling to China. But there was another problem that got less coverage — it turns out that Huawei, which has amazing chip design capabilities but doesn’t have access to amazing chip manufacturing capabilities inside China, was still doing okay because they had access to amazing chip manufacturing capabilities in Taiwan. Namely TSMC — the exact same company that makes chips for Nvidia. As it turns out, TSMC was making massive, massive numbers of AI chips for Huawei, in violation of US export control rules.

The Bureau of Industry and Security had already sent an “is informed” letter to TSMC which essentially told them to shut all of that down immediately. This rule is the follow-up to that “is informed” letter. Everybody’s been expecting this, but the solution they landed on is pretty remarkable. Things are kind of never going to be the same for TSMC, at least not at the FinFET node or better. For companies that are already established customers of TSMC — known, legitimate designers of AI chips — it’s not a big deal. But the Huawei’s strategy of creating another shell company to buy hundreds or thousands or millions of AI chips fabbed in Taiwan for use in China is over. For China, that’s all going to be shut down. That’s one big piece of this rule.

The second big piece of this rule was designed to address some limitations that Jordan, Dylan Patel, and I talked about in the podcast on the December 2 rule, which restricted DRAM and high bandwidth memory manufacturing in China. Now, they have changed the definition of DRAM. A lot of regulations that previously didn’t apply to facilities at companies like CXMT, now apply.

Those are the two big muscle movements of this rule. Even if you thought there wasn’t more they could possibly do, this was legitimate unfinished business. They absolutely needed to do this as soon as possible.

Jordan Schneider: During our last podcast together, Greg, we gave the December 2nd rule a C+/B- grade. With the addition of Monday’s diffusion rule and this foundry rule, how would you grade the entire export control package?

Greg Allen: The focus of that rule was largely on DRAM and HBM. We had complained that CXMT would still be able to buy large categories of equipment, even if they couldn’t buy everything. That was one of the big failure modes of the rule, which was part of the reason we gave it a C-.

Apparently, somebody in the government listened to that podcast, because we’ve now received a new definition of DRAM. This is very esoteric but if folks can bear with me for a second, it will make sense.

The original October 2022 rules set the standard for DRAM manufacturing that was prohibited on an end-use basis at 18-nanometer half pitch. This means if you’re a DRAM manufacturer in China and you approach an American equipment company saying, “Please sell me your high-end equipment to make 16-nanometer DRAM,” the answer would be no. However, if you say you’re only going to make 20-nanometer DRAM, the answer would be yes.

The issue here is that the US government had a decent definition in October 2022 with this 18-nanometer half pitch, but in October 2023 they issued a rule clarification related to some bit density calculation from IRDS. This is esoteric, but here’s the punchline — nobody in the semiconductor industry uses the IRDS definition except CXMT. This allowed CXMT to claim they were making 19-nanometer chips, permitting them to purchase fancy equipment. Under a microscope, these chips look like Samsung’s 16-nanometer chips.

All these purchases, which violated the intent of the rule and had been prohibited until this rule definition update in October 2023, were allowed to proceed. Now they have revised the definition to align with industry standards — what everyone else in the world uses as the definition, except for CXMT and other Chinese companies seeking convenient loopholes. This loophole closure is great, though it should have happened about a year and a half earlier.

Subscribe now

This represents a significant change. All CXMT facilities connected by wafer bridges to their advanced chip node manufacturing facilities, all claiming to make “20-nanometer chips,” will now be cut off from American equipment. According to news reports, an update to Dutch and Japanese export controls may be coming soon. Hopefully, this will include additional categories of technology restricted on a countrywide basis to China, specifically those useful for making HBM. Blocking China’s access to high-bandwidth memory manufacturing is crucial for winning the AI race.

Jordan Schneider: At this point, I feel like they’ve earned a B+ grade.

Greg Allen: Absolutely. This is indeed a huge improvement over December 2, particularly regarding the memory aspect.

To refresh everyone’s memory on why this matters, when you open any AI chip like an NVIDIA H100, you’ll find the big AI accelerator at the center — that’s the part NVIDIA designs. It’s surrounded by high-bandwidth memory chips. Creating AI chips requires both the AI accelerator, manufactured by a logic chip manufacturer, and the HBM, produced by a memory chip manufacturer.

Huawei acquired many AI accelerator logic chips while TSMC was manufacturing for them. The hope is they are now bottlenecked by access to HBM. Cutting them off from HBM sales from companies like Micron, SK Hynix, and Samsung is crucial. Additionally, blocking Chinese companies like CXMT and HMC from domestic HBM production completes the strategy. This rule effectively addresses a pressing national security need and significantly improves upon the December 2 rule.

Jordan Schneider: Can you explain the foundry restrictions?

Greg Allen: The problem they’re trying to solve involves Huawei, wearing a metaphorical wig, mustache, and hat, pretending to be another company called Sophgo, which managed to buy hundreds of thousands of millions of chips.

The original solution proposed in the October 2022 rules included not just the regulations themselves but also “red flag guidance.” Chip companies’ compliance lawyers, upon seeing certain “red flag” indicators, should automatically assume something suspicious is occurring and subject that activity to additional due diligence and scrutiny. For example, when receiving a GDSII file (the software package sent to TSMC with chip design specifications) for FinFET transistors, certain elements would trigger red flags requiring further investigation.

TSMC wasn’t particularly effective at implementing this red flag guidance, which allowed Sophgo to succeed by masquerading as a non-Huawei entity. Addressing the shell company problem — not just in this specific instance but in all potential variations — requires extreme measures.

These AI chips are classified under the ECCN number 3A090. For anyone wanting to make non-planar chips at 16 to 14 nanometers or below at a fab, strict restrictions now apply. Regulators were facing two problems here.

First, there is the technical problem of identifying super-advanced AI chips without burdening manufacturers with compliance requirements for less strategic components like jet ski fuel intake pump controllers.

The ideal solution would allow TSMC to calculate a chip’s processing power (measured in FLOPS) from the GDSII file alone. However, this remains technically impossible until the chip is manufactured and integrated with HBM for testing. The red flags were designed to give TSMC a workable technical evaluation.

The red flag guidance has been revised significantly. Rather than 50 billion transistors on the die, the new threshold is 30 billion transistors. This limit will increase gradually to align with overall chip performance improvements. The 30 billion transistor standard means more chips will fall under scrutiny. While calculating FLOPS remains impossible at the design stage, counting transistors is feasible. TSMC must now generate a technically derived estimate of transistor count.

Previously, Sophgo would simply declare their designs had fewer than 50 billion transistors, and TSMC would accept this claim without verification. The new technical requirements prevent this practice.

The second aspect of the regulations addresses methods for identifying shell companies themselves.

Jordan Schneider: Can you explain the white list and the black list?

Greg Allen: Beyond the technical question of identifying chips of concern, regulators must determine which customers require scrutiny. Their solution involves creating an inverse of the entity list. While the entity list identifies prohibited buyers, this rule establishes an “approved designer list” — effectively a white list. Companies are already included on this list at the rule’s launch, including NVIDIA, AMD, and non-American companies like Sony and Mitsubishi. These companies’ experience buying from TSMC or any global foundry will remain largely unchanged, with no new restrictions affecting their purchasing rates.

The challenge arises with startups that declare they aren’t Huawei in disguise. Without proof, TSMC needs a process to obtain permission to sell to them. These companies must undergo the process of being added to the authorized designer list, involving extensive notification and due diligence requirements — considerably more than previously required. It’s somewhat draconian, but TSMC must face consequences after producing numerous AI chips for Huawei.

This creates a potential issue for U.S. chip design startups seeking TSMC access, as the authorized designer list process takes multiple months. Speaking with policy designers, there are plans to expedite this process, particularly for American companies. The authorized designer list is designed to transition in a year to more of an express lane model once they’ve added desired companies and established proper due diligence processes. These measures are extraordinary but necessary to prevent Huawei from accessing TSMC again.

Share

Jordan Schneider: It’s worth noting that companies making chips this advanced aren’t small startups. They require R&D expenditures in the eight figures just to begin manufacturing on the advanced node for major players. And also, this protects them from being blown out of the water by Huawei…

Greg Allen: The distinction between startups making coffee maker chips versus those making NVIDIA-class chips is significant. These ventures are extremely expensive — NVIDIA invests heavily in its design teams for such development. While we use the term “startup,” we’re referring to extremely well-capitalized companies.

Jordan Schneider: The high-six-to-low-seven-figure one-time compliance cost for getting on the list is substantial but won’t necessarily determine a company’s viability.

Greg Allen: The main challenge with this rule isn’t the financial cost of due diligence but rather the weeks and months required for processing. This timing concern explains why they pre-populated the white list with major global designers and designed the rule to evolve into an express lane system after a year of maturation.

Enforcement is a Thankless Job

Jordan Schneider: BIS now has a lot more work than they did three days ago. Do you think they can pull all this off?

Greg Allen: The U.S. government faces a terrible trap. Congress dislikes hearing about BIS allowing xyz companies to sell to Huawei, since Huawei is on the entity list — this was very controversial in American politics in 2020, 2021, and beyond. Congress expressed anger at BIS for not achieving desired export control outcomes, responding by refusing to raise BIS’s budget, or even cutting it in some cases.

As a father now, this would be like if I told my child, “I’m so disappointed in how weak you are, so you won’t get dinner tonight — I’ll starve you until you get stronger.” But you need food to get stronger, just as you need money to accomplish difficult tasks in large government bureaucracies. We’re stuck in this death spiral where we keep demanding more from BIS while withholding necessary resources.

Think about what’s happened to the budgets of Russian smugglers since Russia invaded Ukraine, or Chinese smugglers’ budgets since the October 7, 2022 rules. Their budgets have increased exponentially, while BIS’s budget remains flat or declining in inflation-adjusted terms.

Undersecretary of Commerce Alan Estevez spoke at CSIS yesterday, noting that before the first Trump administration, BIS rules were typically 30 pages — now they’re routinely 260 pages. They released one Monday, another Tuesday, this one today, with possibly another later this week, though thankfully not AI-related.

The expectations placed on this agency — several hundred humans overseeing trillions of dollars of global economic activity — are unreasonable.

For the price of one helicopter, a properly-funded BIS could generate returns for national security unmatched anywhere in government.

Congress may be upset with BIS leadership, but new leaders are coming. They need resources to execute this strategy — anyone refusing to provide that support shouldn’t claim to be tough on China.

Jordan Schneider: These recent rules demonstrate the organization’s increasing sophistication and government learning. Compared to our October 2022 discussions, the mistakes are becoming less frequent. They’re learning to play to their strengths rather than attempting strategies beyond their budget capabilities. Moving toward countrywide controls, requiring TSMC to handle verification, and implementing the diffusion rule by placing checkpoints strategically for easier enforcement — these demonstrate BIS’s technical expertise and realistic assessment of their capabilities.

Greg Allen: That’s very well articulated. However, none of that replaces the need for more funding or an upgraded IT system to replace their 20-year-old infrastructure. These professionals need proper resources, and I sincerely hope they receive them.

Jordan Schneider: None of this substitutes for simply appearing on ChinaTalk when the rules are released, sparing us from guessing their meaning. The practice of holding webinars two weeks after rule releases is ridiculous.

Greg Allen: Agreed, absolutely. If I could add one point — if I were Donald Trump or Mike Waltz, the incoming National Security Advisor, one of my first week’s priorities would involve gathering the Secretary of Commerce, the Undersecretary of the Bureau of Industry and Security, the CIA director, the NSA director, and the ODNI head around a table for introductions. I’d then address the CIA and NSA directors specifically, asking them to outline their plans for supporting their colleagues’ success.

Jordan Schneider: Jeffrey Kessler is being floated to lead BIS — looking at the spectrum of Trump appointees, we had Pete Hegseth for Secretary of Defense on one end. Then there’s Jeffrey Kessler, magna cum laude in philosophy and classics, learned some Chinese, prior Commerce experience, spent the past 15 years practicing China-related law. I’ve reviewed his speeches — he’s a serious professional. This is manna from heaven! It could have been so much worse!

Kessler on Trump’s trade policy, 2019. Source.

Greg Allen: Here’s someone who examines the CIA’s Cold War toolbox with genuine interest. That doesn’t mean the extreme scenarios like orchestrating a coup in Venezuela — the Cold War playbook worth revisiting here is the use of intelligence to enforce export controls. I’ve studied declassified CIA reports from the 1970s and ‘80s about preventing Soviet access to semiconductor manufacturing equipment. Their work was impressive. The intelligence community succeeded then — they should return to these roots. Export controls matter, and the IC can contribute to their success.

Jordan Schneider: One thing worth reflecting on is how “America First” the diffusion rule appears to be. It literally puts America first in line for GPUs, and the rest of the world is supposed to adapt. I wonder if this rule shifted after the election results to make it more appealing for a Trump administration to continue.

Greg Allen: As part of the outgoing Biden administration’s plan to deprive us of sleep, the diffusion rule was part of a one-two punch. Monday’s diffusion rule admittedly makes building big data centers outside America more challenging. Tuesday brought the new executive order on AI energy infrastructure, focused on streamlining construction in America. These rules align perfectly with Donald Trump’s campaign messaging about AI’s future, infrastructure, and energy.

I believe there’s about a 0% chance those rules will survive unmodified. But when Trump’s team comes in to tweak what they want to tweak, I think they’ll find a lot that they like about these regulations.

A hearty farewell to the Biden export control team

Jordan Schneider: The Biden team deserves credit for their forward-thinking approach over the past few years. Implementing the October 2022 controls before ChatGPT’s release — though imperfect, showed that Washington’s national security officials grasped AI’s importance before the stock market. The diffusion regulations particularly demonstrate foresight, essentially preparing for the world of 2026 and 2027.

Greg Allen: My interpretation of the diffusion rule makes more sense in the context of o3 and the ARC-AGI benchmark. Consider that Dario Amodei, Sam Altman, and Demis Hassabis all believe AGI is potentially just years away, with significant national security implications depending on whether the United States or China achieves it first. The diffusion rule looks like an emergency measure for when AGI appears imminent and the stakes couldn’t be higher.

While their strategic foresight impresses me, I’d offer a different compliment relevant to the Trump administration, one I hear from industry. Regarding the energy rule, the Biden administration not only crafted an excellent strategy — pleasing to hyperscalers reading the energy executive order — but also mastered the intricate details of implementing these interlocking mechanisms effectively and legally.

This contrasts with the first Trump administration’s first year, when many initiatives were overturned by courts until Lighthizer arrived. Lighthizer’s strength lies not just in vision but in legal implementation expertise. Beyond good strategy, which everyone recommends, my advice to the incoming Trump team is that details are critically important when executing AI policy moves.

Jordan Schneider: We’ll see. I want to limit myself to one emergency pod per month for my own personal sanity. But the over/under on the first three months of the Trump administration is probably something like six emergency podcasts.

Greg Allen: You’re going to break your own record, if I had to guess.

Jordan Schneider: I can’t take it. Greg, we gotta stop. We need an interregnum.

Greg Allen: We have children to raise! People, please have some mercy.

Mood Music

The Biden administration is cracking down on compute smuggling with an export control encore! How will this new regulation impact global data center construction? What does it mean to be a Universal Verified End User? Will SMIC swoop in and fill the compute vacuum?

To find out, we brought on Lennart Heim from RAND, Jimmy Goodrich consultant for RAND and fellow at CSIS and UCSD, Chip War author Chris Miller, and Dylan Patel of SemiAnalysis.

We discuss…

• The rule’s three-tier system for categorizing importing countries

• The impact of GPU smuggling and the new verification measures designed to prevent it,

• How the controls will impact data center projects in the Middle East,

• Whether the regulations will financially burden cloud companies and sovereign AI projects,

• The political economy of export controls and what we should expect from the new Trump administration.

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

Mechanics of Compute Control

Jordan Schneider: To start, let’s take a step back and remind the audience why we’re doing export controls on all this AI stuff in the first place.

Jimmy Goodrich: We’re in a massive new era of the AI economy. If you haven’t been living under a rock for the last five to 10 years, the most valuable companies in the world and the most amazing new markets are being driven by AI. Where does that all sit? It sits in increasingly large, massive data centers with tens, hundreds, and possibly one day millions of semiconductor chips — the AI accelerators. The US currently is home to most of these large systems — xAI, OpenAI, Anthropic, AWS, all these companies that are training their models and inferencing them — most of that today is happening in the United States. The Biden administration is looking forward and asking how we keep that leadership here. At the same time, they’re addressing the real issue of China possibly diverting some of these chips. Is it in our national security interest to build massive sovereign AI facilities out in the middle of the plains of Kazakhstan? That’s a super meta question being answered in these rules.

Jordan Schneider: Let’s stay on that for a second. What was the state of play beforehand, and why was the American national security establishment uncomfortable with AI data centers being diffused in the pattern they would have been had these rules not come out?

Dylan Patel: China was still able to access GPUs, whether through renting GPUs from various cloud companies — ByteDance is a top customer at Oracle, Google, and several other clouds — or through data centers being built outside of countries where the US has significant control. This includes many Middle Eastern countries and countries like Brazil, but notably Malaysia. Malaysia has been adding around three gigawatts of capacity just over the last few years. Three gigawatts of capacity is a humongous amount — Meta’s capacity at the beginning of 2024 globally for data centers was roughly three gigawatts. Malaysia as a single country is adding an entire Meta footprint in just a few years. While some of that is Microsoft and other companies like Oracle, which were probably for ByteDance and other clients, a lot of it was ByteDance directly and many other Chinese companies. It represents a significant red flag.

Jimmy Goodrich: Consider this example — YTL Power is building out data centers just across the border in Singapore. They made a big announcement with NVIDIA, but they also made a huge announcement with GDS, which is still a Chinese company. It’s literally across the street from the data center hosting this big NVIDIA cluster, and then a Chinese company is operating the data center next door with the same company. This raises many red flags.

Lennart Heim: Looking at the broader context, Jordan has covered export controls extensively. Since 2002, we’ve had export controls on PRC, China, Russia, North Korea, and arms-embargoed countries. In 2023, more countries were added to country groups D:1 and D:4, notably UAE, Saudi Arabia, and others.

Since then, we’ve seen numerous reports about smuggling and data centers being built elsewhere. Many forget that unlike missiles or other weapons, you don’t need to have a computer in your basement to use AI chips — you can dial in remotely. A company can build anywhere and access resources remotely. When facilities are built in Malaysia or elsewhere, the entities we want to prevent from accessing the underlying computing power can still access it. This rule doesn’t cover this issue.

Chris Miller: There are two ways to read this. One is that this is a major escalation in controls, creating many additional sets of rules that need to be followed. Another interpretation sees this as a natural or inevitable extension of the export controls already in place. If you’ve got restrictions on who can buy GPUs that describe certain countries as off-limits, certain countries as okay, and others as requiring a license — is there a sense in which this is a normal progression of the rules already in place?

Lennart Heim: Let’s examine how this is an escalation. Previously, we covered advanced AI chips — certain AI chips like the NVIDIA A100s and H100s — and now it additionally covers AI model weights, which we can discuss later. The expansion covers more items, but for the model weights, there are broad exemptions.

The whole world now has a classification, which wasn’t previously the case. Previously, we had basically tier-three countries, where no chips were going, and tier-two countries, where chips went under conditions or licensing. Microsoft trying to expand into the UAE is the most prominent case — G42 trying to build data centers there. Before, you needed a license for every chip you sent over. If Microsoft builds a data center there, that’s one license. If G42 tries to import the Cerebras super wafer chip, that’s another export license.

You can now read this as potentially reducing the total number of licenses needed. If Microsoft wanted to build in the UAE or elsewhere, every single shipment needed an export license. Now they can get the “universal verified end user” — one license approved, and they can deploy around the world. The designers are aware of the downsides. They don’t have enough money, there are problems with licensing. They want to take a closer look at the big shipments, the big data centers, and another look at smaller shipments where broad exemptions exist.

Data is needed on how many chips are flowing around the world. Are countries actually buying AI chips, or is it mostly Microsoft, Amazon, and other hyperscalers deploying globally? The book “Cloud Empire” discusses how these US hyperscalers are the ones building around the world. With this new system, their life is easier.

Jimmy Goodrich: This is a natural evolution. We went from the A100 level, tweaked it in 2023, added the HBM late last year. Meanwhile, the US government is trying to figure out what to do with known diversion — both virtual access to data centers by companies like ByteDance accessing huge clusters outside of China, and physical diversion. BIS is quite limited in its enforcement capability. Trying to check every data center without knowing the baseline of where all the data centers and chips are located is like finding a needle in a haystack.

The government hinted they were thinking about this in the October 2023 update when they asked industry for other ideas to address possible diversion. They mentioned on-chip controls and on-chip governance. The industry reaction was pretty lukewarm. If you’re not going to do on-chip governance and you need to address the PRC diversion issue — and ensure that dictators in the Kazakhstani deserts aren’t building AGI supercomputers without your knowledge — you need full-time visibility of where these are going. The only way to do that is through a scheme like this. Questions remain about whether this is doable and whether the Commerce Department can implement something of this magnitude.

Jordan Schneider: It’s time to hand the reins over to Lennart to tell us what’s actually in this rule. How did the Biden administration try to shape the future of AI diffusion?

Lennart Heim: First, it’s important to understand there are three groups of countries. Previously in 2023, we had two groups. Group three consisted of PRC adversaries and arms-embargoed countries — no chips for them, which continues unchanged. Group two is now most of the world; previously this was only certain countries in the Middle East and Central Asia, like Saudi Arabia, UAE, Vietnam, and others. Group one is the US and 18 friendly nations, mostly NATO allies, Five Eyes, and Ireland.

This describes the staged approach. No chips for tier three (adversaries), tier two gets chips under conditions, and tier one has unrestricted access. We now also have export controls on model weights, defined as those using more than 10^26 floating point operations. Training such a model costs at least more than $100 million in compute, probably a billion just for hardware. Publicly available model weights are explicitly excluded. If OpenAI and DeepMind, who generally don’t release their models, were to create such models, there would be export controls on them. No model currently exists in the public domain crossing this threshold. This is about future risk — you might call it AGI — and trying to regulate future models and their deployment, with broad exemptions in place.

Chris Miller: Could we dig into tier two things? The rules for tier three exports to China are pretty clear, and tier one exports to the US, Japan, or UK are also clear. Walk us through the rules around tier two.

Lennart Heim: To make it more complicated, it always matters where we export to, but it also matters who exports — is it an American company or a non-American company? Let’s talk about American tier-one companies. When America and the 18 allied nations wanted to build in tier two — for example, Microsoft wanting to build in the UAE where they’ve had previous plans — they needed a license for each export shipment. Now they have three options.

The first option applies when staying below 1700 H100 equivalents (they measure it in computing power or total processing performance in the rule). This is roughly equivalent to 1700 Nvidia H100s, which is one of the leading chips right now. They have an exception and can basically export them with a presumption of approval. They should still notify BIS because they want to keep track of the numbers.

The second option is getting a normal license to export up to a country cap. This country cap applies to everyone who wants to export. If AWS was first, they already used some part of this country cap, which sits at roughly 50,000 H100s. That’s a lot of chips. To put it in perspective, clusters with more than 20,000 chips were limited until recently — they’re building more and more now. Elon Musk just recently announced this 100,000 H100 cluster. That represents a significant investment in chips for export.

For those who are really serious about building big things, there’s the third option, which is to undergo one of these verified end user programs. In September, BIS announced a Data Center Verified End User program (DCVU). Verified end user programs have existed for a long time in export control, giving trusted actors a streamlined way to get access.

Making it more complicated, they’ve now bifurcated this DCVU. There’s one called Universal Verified End User for companies headquartered in tier-one countries like France and the US, and there’s a National VEU for companies in tier two. The main difference is the universal one allows you to get one license, one authorization to deploy everywhere. Using the Microsoft example — they get one universal VEU and can deploy everywhere in tier two.

However, if you’re G42 in the UAE, you cannot do this. You can only get a national verified end user agreement specific to each country. Deploying in Kenya? New authorization. UAE? New authorization. Different caps apply to these different categories. Generally, the best way to deploy around the world is through universal verified end user status. This tells the tier-one companies and leading cloud providers, “Here’s your simplified way to deploy around the world.” Everyone else has a harder time deploying globally because they don’t get universal agreement, only country-specific authorizations.

Chris Miller: How do I prove to the US government that I’m verified if I want to get a VEU?

Lennart Heim: That’s exactly the key part of this diffusion rule. Before discussing export conditions, we should talk about what concerns us. We’re worried about chips being diverted to different countries through smuggling. One condition requires verifying that chips remain where they’re supposed to be. There’s a biannual chip accounting requirement where you notify BIS that all chips are accounted for. BIS even has the right to conduct on-site inspections.

There’s an interesting technical mechanism to verify chip location through pinging. If you have a computer somewhere and ping different computers worldwide, knowing how long it takes for this ping to travel allows you to geolocate it. This shows BIS telling hyperscalers and hardware companies, “You built this beautiful AI tech — can you build technical mechanisms to help with our problems?” Making sure chips stay put is the first priority.

Another concern involves data centers. Sometimes you want to serve local government and business needs, but more prominently, you want to help leading AI companies deploy globally. ChatGPT needs local deployment to minimize latency and fulfill data requirements — that’s a common motivation for hyperscalers to expand worldwide. When you deploy model weights on a cluster in Saudi Arabia, many people worry about securing this knowledge, as model weights are the most valuable element.

To become a verified end user, you must meet cybersecurity requirements and physical security requirements to keep the processing and model weights secure. Stealing model weights is similar to stealing GPUs — if someone has stolen the model weights that required significant compute power, they don’t need the GPUs anymore. The focus is preventing smuggling, keeping assets in place, and most notably, securing model weights. These explicit security requirements for protecting model weights are unprecedented.

Jimmy Goodrich: Lennart, it’s worth mentioning that we’re only talking about the restricted category 3A090 chips, essentially Nvidia A100 level and above. Looking at older medium-tier Intel Xeons, AMD processors, lower-performance GPUs, V100s — those don’t count. It’s basically A100 level and above.

Jordan Schneider: Let’s stay on that for a second because the Nvidia press release we’re going to discuss today mentioned that gaming chips would be controlled by that.

Lennart Heim: There’s an A/B categorization that’s important to understand. Tier two only has controls on category A — that’s only data center chips. Any gamer in Saudi Arabia, UAE, India is fine.

Jimmy Goodrich: It doesn’t apply to autonomous driving chips either. Your Jetsons are excluded from that as well.

Lennart Heim: Indeed. In the PRC, there is a notification requirement if your gaming chip is above a certain threshold of processing performance. Nvidia needs to notify BIS about these kinds of exports so they have visibility into these activities.

When we talk about data centers being built, we’re talking about AI. When I mention share of computing power, it means AI computing power. These data centers use GPUs with chips costing about $30,000 each. This isn’t about data centers used for watching Netflix or using anything in a Microsoft suite or cloud. This is about data center GPUs being used for AI workloads. All other operations can continue as usual.

Jimmy Goodrich: A good way to think about it is you’ve got your standard data center for web storage, but we’re talking about accelerated AI compute, which is an entirely different category. You have much more liquid cooling, more cooling fans and towers. You literally engineer the data center differently to accommodate these AI-accelerated computing clusters. You can see it visibly when driving around places like Loudoun County — this data center doesn’t have the 50 cooling fans surrounding it or the massive compressed liquid cooling tanks. There’s a categorical difference between AI data centers and your traditional web storage or cold storage data center.

Chris Miller: How does this play out over time? Right now we’ve got the level of chips that are restricted, but chips get better at a pretty rapid rate. Should we expect regular updates so that as chips age out they get opened up, or will this largely remain frozen with real restrictions imposed on what types of data centers can be built in third countries?

Lennart Heim: Setting aside the policy question of whether to move it up or not, let’s look at the reality. Computing chips are improving exponentially. More chips will run into these restrictions. More importantly, we claim it’s only AI data center chips, but more chips are becoming AI chips. More workloads are becoming AI applications. Everyone has seen it — whatever software you open nowadays has a new AI feature.

This highlights a key challenge. More chips are becoming AI chips, therefore they have increased performance. They might eventually hit these thresholds. For example, at some point an Apple Ultra chip might hit these thresholds. Then you must ask the US Government whether we should control these chips and how useful they are for the things we’re worried about.

It gets more complicated. Can you hook up a bunch of Minis and build a supercluster? It’s clearly not as good as the alternative, but practically speaking, computing power is computing power. When you have something exponentially improving and draw a line in the sand, it will eventually get crossed. This needs continuous improvements and threat model assessment. Maybe it turns out fine for AI in the next three years — then clearly this line should go up. But if developments accelerate next year, there might be reasons to control these things. Export control is a blunt tool. You will hit more targets, and the more things become AI-related, the more things you will affect. That’s a fundamental challenge of this framework.

Chris Miller: Dylan, what do you think about the financial implications for the companies? One of the debates has been whether this is impactful to companies or if they will sell the same number of GPUs, just put them in different countries. Dylan might have a view on that.

Dylan Patel: This regulation definitely impacts the financials of many companies in the space, primarily NVIDIA. You can argue the GPUs will just get rerouted, but that’s not always the case. A significant number of GPUs would not have data center capacity in the West — there is a data center shortage. This explains why people are converting anything they can into a data center. That opportunity is now gone. You can’t reshore everything. China was increasing demand for GPUs and affecting the elasticity of demand required for supply creation and sales to them. NVIDIA is impacted — it isn’t just wholesale rerouted because of data center capacity and potential demand issues. On the flip side, companies that can get a universally verified list will benefit massively.

Jimmy Goodrich: Companies will find a way to get these licenses for the real demand out there for huge AI data centers, particularly the universal VUs and large hyperscalers. They have the resources, huge teams, lawyers — they’re all FedRAMP certified, as the rules require. It will be much easier for them.

The big question concerns the NVEUs — these sovereign AI cloud startups or small GPU-accelerated cloud projects in different parts of the world. Will they have the wherewithal to do this? Well-resourced organizations like G42 will absolutely take a shot. But will the Nepalese cloud supplier thinking about this actually make it through? If they don’t, should they have been building an AI data center in the first place?

These questions might force a rethink among some sovereign players who were considering spending $5-10 billion of their government money on data centers instead of healthcare, education, or transportation for their citizens. What is the actual viability of these sovereign AI projects in different parts of the world? Some are totally legitimate with real demand. Others are somewhat questionable.

One key indicator to watch is what companies are saying to their shareholders. Watch carefully to see if any affected companies put out statements notifying shareholders of possible negative impacts. We haven’t seen anything yet, which doesn’t mean we won’t. It’s a huge regulation that will take time to work through and understand exactly how to revise financial projections. Previously, we’ve seen companies claim something would be catastrophic to their business, have a massive negative effect, and then clarify to shareholders that it wouldn’t have any effect at all.

Lennart Heim: Anyone who’s worked in export business is used to this. What’s different this time is that it’s public, being done via press statements. The Oracle press statement is definitely worth reading to get an idea of the mood out there. As Jimmy was saying, the question is whether they’re complaining because of real financial impacts, or because the administration is changing and now they can protest at full throttle without constraints. To some degree, it makes sense to take a shot at potentially killing it in the next administration. It’s hard to balance these considerations — is there a real impact we should discuss, or is this just taking advantage of an opportunity? We haven’t seen such public backlash against export controls before.

Jimmy Goodrich: It impacts every company differently depending on their market position. Companies designing and selling controlled GPUs are most impacted. Companies integrating them into hardware and selling them as part of a solution, like Oracle, are basically as impacted as the chip makers. It might actually be more favorable for the tier one hyperscalers because you get one license to operate around the world. For sovereign NVEU operators in tier two countries, many won’t make it through NVEU authorization, but many will. It will be interesting to see where the dust settles.

Jordan Schneider: It’s interesting to consider the political economy of Google, Amazon, and Microsoft not supporting this, even though it’s a regulatory burden that could eliminate many competitors. Brunei, if they want cloud services, will have to go to AWS rather than build sovereign AI brought by Nvidia DGX cloud.

Jimmy Goodrich: Looking at business history, when has any industry welcomed a new regulatory framework that oversees most of their business? It’s natural for them to prefer selling to whomever they choose at their own pace. The Biden administration is saying they need oversight to check customers and ensure they meet security standards preventing Chinese access.

Consider historical parallels like the debate over seatbelts in cars — car companies initially opposed that idea. Or the Foreign Corrupt Practices Act, where industry claimed they couldn’t operate with restrictions because they needed to make payments to extract oil in places like Zimbabwe. The law passed, people complied, and moved on. It’s natural for businesses to resist new regulatory oversight of their business model. This shouldn’t surprise anyone. The real impact will become clear when we examine the financial statements over time.

Chris Miller: One other question on that front — Dylan mentioned the build-out happening in Malaysia as a good example. There have been discussions that some is for Malaysian domestic demand, while some might be diverted towards China, maybe a lot of it. Jimmy, what’s your view on the share of GPUs currently going to tier two countries? What portion is for tier two country demand versus likely diversion? Is it 20% diversion, 80% diversion? Hard to know exactly, I’m sure.

Jimmy Goodrich: No one really knows because those numbers are only visible to the suppliers of chips and server-level systems integrators. In 2023 and 2024, we saw some interesting spikes in sales to Southeast Asia. As Dylan pointed out, there’s real data center growth happening there. Indonesia is building data centers, with many Chinese companies building locally. With China’s slowing economy, some listed Chinese data center companies like GDS and 21Vianet are growing faster outside China than inside.

The Wall Street Journal and The Information have reported that ByteDance, although technically headquartered in Singapore, has been looking at accessing these data centers in Malaysia. In some cases, you’ve got a big AI-accelerated data center right next door to a GDS data center owned by the same company. Huawei and others have had large manufacturing operations in Malaysia, with reports of diversion occurring there.

Regarding real end use, we don’t really know. Reports indicate that Nvidia physically inspected data centers to verify chip installation, but they only looked once. After inspection, everything was shipped out — the entire data center was built as a scam for inspection. They packed it all up and moved it out, like a Looney Tunes animation secretly swiping everything out the back door. These new rules are saying more oversight is needed. What was in place before, mostly industry self-initiating, despite good efforts, isn’t enough.

Lennart Heim: None of this surprises anyone who has followed the semiconductor industry and export controls — there are wider schemes to accomplish goals. The advantage of computing power is you want to move the unit of governance from AI chips, which are hard to govern and can be smuggled, to computing power itself. Who’s using the computing power? You move in this domain by saying that even when selling to a trusted entity, we need this continuous, ongoing relationship with you as a verified end user. Then you check that your customers aren’t diverting computing power. We could have all the chips outside the PRC, but they could still use them. You want to ensure whoever operates this is checked for their usage, which is exactly what cloud businesses do. Practically speaking, it provides better governance.

Jimmy Goodrich: Another important element ensures the US doesn’t look away as data center capacity is built overseas, particularly by deep-pocketed sovereign nations. We’ve seen this with semiconductors through the early 2000s. Chris, you wrote about this in Chip War — Taiwan, South Korea, China, Singapore doling out billions in cash incentives for semiconductor factories. The US share of semiconductor manufacturing dropped from nearly 40% in 1990 to less than 20% , requiring billions to recover.

This appears to be a preemptive measure to keep advanced data center capacity in the United States. Countries like Saudi Arabia and UAE have over a trillion dollars in cash, eyeing these companies and data centers. The fear is waking up in 10 years wondering how our AI ended up in a desert, controlled by nations friendly with Russia and China. We should have had policy earlier to prevent that. Lennart, you should probably discuss the ratio thing, which is crucial for tier one because it addresses that issue.

Jordan Schneider: Let’s stay on that for one second, Jimmy. It’s illustrative that the Biden administration paired this with an executive order trying to create easier ways to get carveouts for building new power plants. Trump also seems influenced on this — his influencer’s initials might be E and M, but you never know. Power generation is the big bottleneck domestically in building out more data center capacity.

It’s a two-step game — if you don’t want the infrastructure built abroad, you need to make sure it’s possible to build in all those tier-one countries. A domestic regulatory agenda needs to be paired with all this. During World War Three, these buildings would be a weird asset — protecting them is another issue we’ll address when necessary.

Lennart, this reminds me of the Washington Convention of 1923 — the US gets five ships, the UK gets five ships, Japan gets three ships, Germany gets one and a half ships. Now we’re doing this for AI compute. What’s the ratio deal within this regulation?

Lennart Heim: The word “ratios” could be better chosen. It could be called “share.” Jimmy was alluding to how 20% of semiconductors are produced in the US — that’s not great and should be different. This framework requires that if you’re a universal verified end user who can deploy around the world, like Microsoft, 75% of all your AI computing power must be built in tier one. US companies must keep 50% of their whole share in the US.

That means Microsoft must keep half of its computing power in the US. The recent AI executive order on permitting infrastructure, building data centers, and getting more energy is critical. If you tell companies to build here, you must simplify the process. We’ve had issues getting transmission lines and everything else ready in the US for the data center buildout needed in the future.

Companies must keep 50% in the US, 75% in tier one, leaving 25% for the rest of the world. This is per company, not the total share of all chips worldwide. You can have 25% in tier two because nothing can go to tier three. Within tier two, you’re only allowed to put 7% in a single country. If Microsoft wants to build a big cluster in Saudi Arabia, the cluster can only be 7% of their whole computing power.

The numbers emphasize America first — America and our partners first. The data centers must be kept here for security. The role of AI in World War Three and the longevity of these data centers remains uncertain, but currently, keeping them here ensures security.

Jimmy Goodrich: This is definitely forward-leaning because the United States rarely considers how to stay ahead with allies at the cusp of an industrial revolution. Many other industries, particularly physical assets, have dispersed around the world. Look at the procession of “kiss the ring” voyages over the last two years — when Saudi Arabia or UAE hold a tech conference, every tech leader rushes there praising sovereign AI. This isn’t because they have a huge population of internet users wanting ChatGPT access or because of strategic location — it’s because they have enormous cash reserves.

If you’re a big tech company offered substantial money, you’ll build there regardless. The US Government is now saying that while cash is good for companies, national security implications must be considered. The fastest supercomputer in the United States modeling nuclear weapons, just installed at Lawrence Livermore National Lab, has 44,000 GPUs. This rule allows any country to build a supercomputer of that size. Every tier-two country remains tactically able to build really powerful supercomputers — some of the most powerful in the United States aren’t even that large.

Lennart Heim: The numbers are critical here, as is the status quo. The rule includes total processing performance numbers — they’re big numbers nobody will grasp. We mostly discuss H100 equivalents. When setting these ratios requiring 75% in tier one, the US government surely crunched these numbers knowing more than 75% is currently in tier one and more than 50% is in the US — probably significantly more.

Looking at the current reality of AI compute worldwide, it’s quite concentrated with the big hyperscalers, primarily in the US, then in other tier-one nations, followed by tier-two nations. The goal is maintaining this structure because, as Jimmy mentioned, some entities have deep pockets. The message is, these pockets are deep, but they aren’t bottomless.

Jordan Schneider: Keeping the data centers here reveals a central contradiction in Biden’s initial approach. The Biden line toward China was always “small yard, high fence” — they weren’t trying to contain China’s economic rise, but viewed this as a national security question. They weren’t comfortable with China having technology to model missiles. These regulations only matter if AI becomes more important, not just in a national security context, but in economic growth.

You’re using these to constrain technology that impacts both military outcomes and has potential for industrial revolution-level effects on a nation’s economic prospects. Two questions arise here.

1. To what extent will having data centers in the US and allied countries privilege those nations in gaining economic benefits from AI? If you’re in a country with lag time or you’re stuck going through AWS or Microsoft, are you losing out on vertical integration benefits from artificial intelligence?

2. Will the Trump administration stick to this line? It doesn’t fully align with logic, since you’re clearly constraining other countries’ future economic potential by limiting AI access. Once Trump talks about being in an AI race and intending to win it, that creates second-order effects. This doesn’t align with “small yard, high fence” — sitting down with Wang Yi for 12 hours explaining these export controls are just for this small thing that only affects missile systems.

Chris Miller: Jordan, isn’t the fact that API access remains broadly open the counterargument?

Jordan Schneider: Maybe it doesn’t even matter. It’s just an off switch for the future, unless there’s some significant lag where using compute in a tier-one country from far away will slow you down.

Lennart Heim: Latency might be an issue, though not for current frontier models where computational workload is high. You constrain people from accessing GPUs directly, where they’d have more flexibility, versus only getting computing power as infrastructure-as-a-service. Leading AI developers don’t keep GPUs in their rooms or basements — they dial in remotely via cloud computing.

Most people access it through the cloud, with all its benefits and downsides. AI is a dual-use technology — more chips are becoming AI chips. Export controls are blunt instruments. It’s extremely difficult to carve out one specific concern. If you’re worried about AGI, by definition it’s general — you can try to take over the world or grow your economy. You can still access computing power unless you want to train a really big model. Similar to the banking system, it gives on-demand access with the possibility of cutting off access — this is weaponized interdependence.

Jimmy Goodrich: This data center revolution will drive massive upstream infrastructure growth — energy, water, concrete, steel — creating new industries in the United States. That’s a huge net positive for the US and its allies. However, research shows you benefit most from diffusing AI throughout your economy. If you’re Thailand, you’ll get more value from $5 billion spent incentivizing AI use versus developing sovereign AI.

Quickly integrating AWS, OpenAI, or Meta into banking, utilities, public transportation, and healthcare will determine which economies succeed — you don’t need locally hosted AI. The challenge comes with sensitive personal data that countries want to keep within their borders. Solutions will emerge through NVEU or UVEU systems. These rules don’t prevent building in tier-two countries — they just require additional steps.

Will Adversaries Fill the Void?

Jordan Schneider: Okay, Jimmy. Let’s talk about China.

Jimmy Goodrich: The mega question is whether China can fill the void. No one wants to see China fill a void anywhere. China having the digital Silk Road in Africa and the Middle East poses a major challenge for the United States. Travel the world and you’ll see Huawei or ZTE phones in citizens’ pockets throughout the developing world, or Huawei cloud offering email web services, or Chinese companies offering Smart City and surveillance solutions.

However, that’s entirely different from AI-accelerated solutions. Due to the technology’s complexity — producing sub-7 nanometer semiconductors for advanced AI accelerators, 3D stacking layers of the most advanced DRAM for HBM — plus all the export controls, China is struggling to meet its own demand with indigenous AI GPU production. Reports indicate China’s operating one, maybe two factories to produce advanced logic chips around 7 nanometer, which is two or three generations behind where TSMC, Samsung, and eventually Intel will be. On HBM, China is just starting up several efforts and remains years behind.

The big question is, can China swoop into the UAE, Brazil, and India with a ready-to-go, full-stack ecosystem? Today, the answer is no. Even in delivering to Baidu and Tencent within China, they’re struggling to meet fulfillment orders. Numerous reports highlight how lower yields and production issues at SMIC and other facilities create challenges for China.

Subscribe now

They will improve — anyone who’s written off China from an innovation perspective usually gets it wrong. However, this isn’t just about innovation — it’s about scale. China will eventually produce five, maybe even four-nanometer chips, but they need to produce millions of units. They need millions of HBM units, millions of packaging substrates for integration, plus excellent software for easy plug-and-play with customers worldwide.

Even if China could export large quantities of their AI chips, would a sovereign AI company in Thailand have the skill set to completely re-engineer its software stack to run on something like the Huawei Mindspore? Probably not. That skill likely only exists in the United States and China.

What’s likely to happen is developing economies unwilling to navigate licensing hoops will buy H20s — good chips usable for inference. Many sovereign AI projects focus on inference, not training, because they can rent that capability to other companies and players.

I’m skeptical of claims that China will capture market share from US players. Currently, China can’t meet its own demand. This could change — we must watch closely. That’s why all roads lead back to export controls on semiconductor manufacturing equipment and leading-edge semiconductor fabrication. If those don’t work, none of this policy works. It’s like a house of cards.

Jordan Schneider: Lennart, Jimmy mentioned inference. The last show we did about a month ago discussed the increasing importance of inference versus training when producing and getting the most out of models. How does that technological development interact with this rule, which was years in the making, probably before scaling on inference-time computing became widely recognized?

Lennart Heim: This rule leverages existing export control by using the same export control number. The H20 is not being controlled — everyone around the world can buy an H20 as much as they want. This is good for Nvidia’s revenue stream, particularly in China. The line between training and deploying becomes more blurry over time. When you deploy for longer and let the model think for longer, it leads to better capabilities.

For careful readers of this rule, there’s a reporting requirement if your chip has a higher memory bandwidth than a certain amount, which the H20 meets. My understanding is that if you export it, you need to give BIS a heads-up. They want some idea of how many H20s are out there, which seems sensible. If not many people are buying it, it might be fine, but if many people are buying it, intervention might be necessary. These export interventions need to expand to cover the H20 for more exhaustive coverage of new AI developments.

Jimmy Goodrich: This point is super important because if you read carefully the criticism from the private sector, it claims this will tie the hands of American companies while China comes through. The reality is that today, no one has been able to find a single data center outside of China that’s at scale — meaning 10-20,000 H100 equivalent — fully loaded with semiconductors built indigenously within the PRC. Those don’t exist today.

There’s a huge barrier that a local sovereign player would have to overcome to port their whole system from CUDA software to Huawei MindSpore. This is actually a major, underappreciated advantage that U.S. companies have — everyone’s already coding and developing AI frameworks on the existing firmware that these companies, AMD included, have developed. That itself is a huge disincentive to use Chinese systems.

China could play it smart by telling their Chinese companies to use a larger quantity of H20s, then start exporting a smaller quantity to some keystone projects to claim the US AI diffusion rule isn’t working. It’ll be important to look at the overall numbers to see the actual scale. If China can produce millions of comparable GPUs loaded with comparable HBM, we would need to take a hard look at this rule and the levels being set. However, the evidence shows that China isn’t there yet.

Lennart Heim: The most likely way this could fail is by assuming that because it’s fine now, it will be fine in the future. That’s far from granted. It needs to be monitored, and we need to examine the evidence and data. As Jimmy was saying, we tried to find data centers and didn’t find any. I found only two systems trained with Ascend 910s out of approximately 700 systems in the database, with hardware information for 260. People aren’t really using it. If this changes — and this is a job for the intelligence community and others — it needs to be monitored and adapted. This is the most likely way things can go wrong if we’re too stringent and they just fill the void.

Share

Jordan Schneider: One last point about what you don’t get from having your own data center. Kevin Xu wrote a fantastic article about DeepSeek’s three advantages. One thing he pointed to as part of their special sauce is that because they came from a quant hedge fund, they’ve been running their own servers for 15 years. One of his theories about how they were able to be so efficient with their training is their expertise all the way up and down the stack when it comes to training models.

There could be a future where having to outsource much of what you do to Google or AWS means missing out on technological innovations. Companies in other countries around the world might not be able to explore different ways of training and deploying models. These are the trade-offs we’re weighing when considering all the different risks here.

Lennart Heim: It would be wrong to tell ourselves we can have the best of both worlds simultaneously at no cost. This clearly isn’t the case when you intervene. The rule reflects where people fall regarding national security risk and how much should be done here. That’s what we got.

Outtro Music

What can American diplomacy do to head off an invasion of Taiwan?

To find out, ChinaTalk interviewed R. Nicholas Burns, Biden’s Ambassador to China, whose diplomatic career spans 35 years and 8 countries.

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

We discuss…

• Lessons from Kissinger’s career,

• How China’s negotiating tactics are so different from those of the Soviet Union,

• Great power responsibilities, and whether Chinese leaders truly appreciate the reputational costs of helping the Russians and the Houthis,

• How talking can lower the risk of WWIII.

We have a podcast! Listen to this interview on Spotify, iTunes, or your favorite podcast app.

The Art of Diplomacy

Jordan Schneider: Ambassador Burns, welcome to ChinaTalk.

R. Nicholas Burns: Thank you very much. Long time coming, Jordan. I’m a big fan of yours. I’m glad we could schedule this before I leave China next week.

Jordan Schneider: When you co-authored a book on Kissinger’s negotiating style, you wrote about the diplomat who, among many other things, probably had the most fun negotiating in U.S. history. As he described his Chinese interlocutors, he said…

“Mao, Zhou, and later Deng were all extraordinary personalities. Mao was the visionary, ruthless, pitiless, occasionally murderous revolutionary; Zhou, the elegant, charming, brilliant administrator; and Deng, the reformer of elemental convictions…”

Do you ever get bummed out that the Chinese officials you were destined to deal with are so much more boring?

R. Nicholas Burns: I don’t think they’re boring! In fact, we have animated discussions. When I was a professor at Harvard, I co-authored a book on Henry Kissinger called Kissinger the Negotiator with two other professors. The main author was Professor Jim Sebenius of Harvard Business School. We spent hours with Secretary Kissinger in New York and in Cambridge at Harvard. He told us that while hundreds of books had been written on him, no one had ever written a book about his negotiating theory and style.

Subsequently, when I was nominated to be ambassador, I spoke to him multiple times before I came to China and then multiple times while I was in China. This extraordinary thing happened when he was 100 years of age — he came here to Beijing in the summer of 2023 for five full days. I met him at the airport, and he was ready to go. I really benefited from his historical perspective on the arc of the U.S.-China relationship and what it’s like to deal and negotiate with the Chinese.

Many of the principles he developed based on his conversations with Mao, Zhou Enlai, and Deng Xiaoping remain true today because the Party is still the primary agent of Chinese power. In my conversations with the Chinese leadership — I’m having a series of outgoing meetings with ministers here and other senior officials — these are challenging discussions. As a career diplomat, I have a healthy respect for how well-trained these Chinese diplomats are. They’re really worthy adversaries, and they’re thoughtful in many different ways. I’m having big-picture discussions this week about where this relationship is heading. They’re not boring.

Jordan Schneider: After taking Nixon to China, Kissinger reflected on how the U.S.-Soviet relationship changed:

“Prior to my secret trip to China, Moscow had been stalling for over a year on arrangements for a summit between Brezhnev and Nixon . . . then, within a month of my visit to Beijing, the Kremlin reversed itself and invited Nixon to Moscow.

“Suddenly, the Moscow summit was not elusive. . . . Other negotiations deadlocked for months began magically to unfreeze[.]”

Today we’re in a period in U.S.-China relations where Beijing is not super interested in negotiating anything necessarily to the level that the U.S. and Soviet Union achieved over the course of détente. What are your reflections on that?

R. Nicholas Burns: Many people have raised this with me in conversations — is there an opportunity that we in the United States could look at the current situation in our fraught relations with both the Russian Federation and China and do a reverse Kissinger? The situation we’re in today in 2025 is completely different from what Kissinger and President Nixon faced in ’69, ’70, ’71 as they began to think about the opening to China.

You’ll remember, Jordan, that Mao began to distance himself considerably from the Soviets in the very late ’50s. By the early ’60s, he had kicked the Soviet advisors out. They’d gone on to develop their own nuclear weapons program by 1964. What Kissinger and Nixon did was historic, but they had an opening that is certainly not present today.

The fundamental question for the Chinese is this: they continue to say — and many of them believe — they want to be agents of world order, they want to be responsible, they want to respect the international system.

But they’ve got a choice to make because they’ve aligned themselves with agents of world disorder.

They’ve aligned themselves with Russia, with its brutal invasion of Ukraine, and the Wagner Group sowing mayhem in West Africa. Iran, now weakened thankfully, but for 40 years has been one of the biggest problems in the Middle East and a country threatening to become a nuclear weapons state. North Korea is one of the biggest agents of disorder in the world today.

There’s a loose alignment between China and those three countries, and they can’t have it both ways. At some point, they’re going to have to choose. Obviously, we would hope that the Chinese would want to be a more responsible country than they have been on some of these major issues of global order and the future of the international system.

One good example that leaps out to me, because I’ve lived it here since late February of 2022, is that the Chinese say they’re neutral in Russia’s assault on Ukraine, but they haven’t been neutral. Hundreds of Chinese companies have been giving very important dual-use technologies to the Russian war machine. The Chinese deny that’s happening, but it is happening, and we’ve sanctioned those companies.

Another example is the cyber aggression by China against the United States, which Director Wray and other American officials have been talking about publicly and how objectionable this is. These are the questions that we Americans need to put before the Chinese leadership. I’m doing that. I’ve been doing that for several months, as have Jake Sullivan, Secretary Blinken, and others. That’s what the Chinese need to really think about as they think about their own position in the world.

Jordan Schneider: In the book, you discuss how Kissinger, when at an impasse where the other side didn’t necessarily see your point of view, would start playing 4D chess, moving things around on other parts of the global chessboard. The dream is to get to a direct bilateral dialogue of the sort that Dobrynin had with Kissinger, where he says, “I can say with certainty that had it not been for my channel with him” — where they’re doing all this Marilyn Monroe-style sneaking into the White House — “many key agreements on complicated and controversial issues would have never been reached and dangerous tension would not have been eased over Berlin, Cuba, and the Middle East.”

The question to you is, if the stage is not set for that real dialogue where both sides are interested in really grasping, really trying to push forward solutions on the biggest, most difficult questions, what can dialogue do?

R. Nicholas Burns: Dialogue does two things, and we do believe in it.

1. First, it’s the daily practice of diplomacy. There are a thousand issues, metaphorically, in the U.S.-China relationship. We deal with an extraordinary range of problems with China. Conversations from my embassy to the Chinese Foreign Ministry and other parts of the government here happen every day on all those issues. They take place in Washington with the embassy from the PRC as well. That’s one order of business you’ve got to have.
   
   We learned in this relationship — and I certainly learned — what happens when you don’t have daily connectivity. After Speaker Pelosi’s visit to Taiwan in August 2022, the Chinese very unwisely and objectionably shut down eight channels of communication. We went for several months until November of 2022 with not a great ability to contact them. I was having meetings with the Foreign Ministry, but we didn’t have our cabinet engaged.
   
   In early 2023, when that strange balloon drifted across the national territory of the United States in a strange Orwellian way, the Chinese blamed it all on us when the President rightfully ordered the balloon shot down. We went through another three to four months of no appreciable contact. That’s a dangerous situation when the two strongest countries in the world, two strongest military powers in the world, where our militaries are juxtaposed very closely to each other in the East and South China Seas, are unable to have conversations if there’s an accident or a misunderstanding.
   
   What we’ve tried to do over the last 18 months in re-engaging with the Chinese is to establish reasonable, sustained conversation about diplomacy. It’s identifying problems — sometimes you can’t find a solution, but you can manage the problem and you can disengage or sanction them, whatever the answer is. You just can’t afford in the modern age to have a situation where we’re not talking on a daily basis.

2. This second part is very meaningful — can you have thoughtful conversations that are not tit-for-tat and trying to one-up each other or score points? Can you try to investigate what’s behind that policy that you’re running? Why are you doing it? What are the limits of it? What are the boundaries of it?

We’ve begun to have those conversations. I’ve been with Jake Sullivan in Malta and in Bangkok and here in Beijing over the course of the last year and a half in two-day meetings each of those times with Wang Yi, who is the Foreign Minister of China but also has a more senior position — he’s Director of the Foreign Affairs Commission. In 12-15 hour conversations in each of these settings, Jake and Wang Yi have been talking in a much more introspective way. Secretary Blinken has had those types of conversations with Wang Yi as well. It’s harder with the military here because they’re more closed, they’re more opaque. But it’s important to have those conversations as well. It’s part of diplomacy between two superpowers.

Jordan Schneider: My bellwether is Wang Yi coming on ChinaTalk. When that happens, then we’ll know we’re really ready to get down to business.

Michèle Flournoy said in an interview I did with her a few years back:

“Risk reduction measures and crisis communications are things we tried even when there was a lot of dialogue back in the Obama administration. We tried to push them on hotlines or incidents at sea agreements. These were mechanisms we had with the Soviets. The Chinese have never been willing to talk about that. They just say, ‘You seem to really want this, so give us something else like stop talking about Taiwan.’ And of course that’s never going to happen.”

I’m curious for you to reflect more broadly on the relationship between talking and the probability of great power conflict. People refer back to World War I where great powers sleepwalked into war. But there’s also World War II, where the U.S. and Japan were talking the day of Pearl Harbor, where Hitler and Stalin were talking the day before Operation Barbarossa, where Hitler and France were talking before Czechoslovakia. What are your thoughts on that?

R. Nicholas Burns: This is a central question. We are competing with China. The answer to the question, and it’s structural, is that we’re global rivals. It’s going to continue. One of the tests of this relationship is, can we compete and yet do so in a way that doesn’t elevate the probability of a conflict between us? Our job is to diminish the probability of a conflict.

I remember the EP-3 incident in the spring of 2001 — I was not involved in China affairs then. We had an air collision between Chinese and American military aircraft, which led to the death of the Chinese pilot and led to the impoundment of our plane and our crew. One of the things that we’ve worried about, and I certainly have worried about in my nearly three years here, is can we handle that kind of crisis in the U.S.-China relationship? Can we have the kind of higher level connectivity so that if 24-year-olds collide in ships in the Spratlys or the Paracels or the Senkakus, senior people can intervene and diffuse the crisis?

We’ve worked on it for a long time. The PLA wouldn’t talk to the senior levels of the military. During the balloon crisis, I was a primary port of contact with Vice Foreign Minister Xie Feng, who’s now the Ambassador of the PRC in Washington. What I was saying to him was we need to get our senior military leaders talking about this incident, and they refused. President Biden pushed this at the San Francisco Summit and again at the APEC Summit in ’23 and ’24 with President Xi Jinping. The result is the Chinese have agreed and we’ve begun to have those higher-level military-to-military contacts. Admiral Sam Paparo, a very gifted leader who’s the head of Indo-PACOM Command in Honolulu, has had two meetings with the Southern Theater Commander of the PLA just this past autumn.

That’s just the beginning. We’ve got to have those kinds of contacts so that in the eventuality of a profound misunderstanding or an accident, we can intervene to keep the peace. Strategically we’ve got to compete with China, but we also have to live with China. We’re not trying to head this into a brick wall, we’re trying to avoid the brick walls. That gets to the heart of your question: do you have enough connectivity at the senior levels to do that?

We’ve begun to do it with Jake and Tony Blinken, Secretary Blinken with Wang Yi, and certainly with Admiral Paparo. We’d like to be able to get our Secretary of Defense, our Chairman of the Joint Chiefs — they have talked to their Chinese counterparts — but obviously elevate the conversation to them. As the American Ambassador here, one of my fundamental jobs is to establish these relationships of my own with Chinese leaders so that I can — and my successor obviously can — play a role in this. The central question is: how do you compete vigorously and at the same time keep the peace? That is what is at stake here for the United States and China.

Jordan Schneider: The key question is — even if they were really excited to talk about military matters X, Y, and Z, there’s the World War II analogy, where you can be doing one thing with one hand and be planning very dastardly deeds with another.

R. Nicholas Burns: Listen, you have to have your eyes wide open at all times in a great power relationship.

I’ve been asked a lot, “Are you trying to gain the trust of China?” or, “Do you trust China?”

My answer to is always the same — it’s not a question of trust. It’s a question of judging the Chinese by what they do, not just what they promise to do or what they say publicly or privately. Judge them, and call them on their actions, whether they’re positive or negative.

What we have been doing here on a practical basis is basically running a relationship that is, in my mind, about 80% competitive. I spend about that amount of time on the competitive edge of this relationship, whether it’s our military differences in the Indo-Pacific, our technology differences, or human rights differences. A generation ago, when I was Under Secretary of State for Condoleezza Rice, our ambassador, Sandy Randt — and he was a really good ambassador — probably was 75-80% engagement. I’m now 20% engagement. You have to judge these countries by what they do, and obviously we have plenty of disagreements strategically and tactically with what the Chinese are doing.

Jordan Schneider: You said on a podcast that, growing up and even into the ’80s, you never imagined the USSR would disappear. Then all of a sudden it did. You also noted that you didn’t really like Reagan a lot in your 20s but, in retrospect, came to respect his stance on the USSR. There’s the Gallagher-Pottinger thesis today, and I want to give them the full quote because people say “regime change,” but it’s a little more nuanced than that: “The U.S. shouldn’t manage competition with China. It should win it. What would winning look like? China’s communist rulers would give up trying to prevail in a hot or cold conflict with the United States and its friends, and the Chinese people, from ruling elites to everyday citizens, would find inspiration to explore new models of development and governance that don’t rely on repression at home and compulsive hostility abroad.” What’s your reflection on the Reagan-inspired Gallagher-Pottinger playbook for U.S.-China relations today and into tomorrow?

R. Nicholas Burns: I joined the State Department as an intern 45 years ago in 1980, and then full-time as a diplomat in ’82, so I served in President Reagan’s administration. President Reagan was an absolutely key figure in preparing for the end of the Cold War, which, by the way, we didn’t expect. I was actually at the White House between 1990 and ’95 at the NSC, working for H.W. Bush and then Bill Clinton. Right up until a month or two before the fall of the Soviet Union — really, it was only until October, November of 1991 that we thought the Soviet Union would collapse. It was that late, and that was a judgment of our entire government.

This is a very different situation. While it’s tempting to compare that old Cold War — and I was part of it in the early stage of my career — to this time, China is infinitely more powerful in its own self, in its own state, its economy, its technological base, its science and technology expertise, its incredible universities and research institutions — far stronger than the Soviet Union. That has to be said first.

I have a lot of respect for Mike Gallagher and I’ve talked to him a lot about these issues, as well as Matt Pottinger. They’re really smart guys, and they’re trying to be thoughtful about this relationship. I would pose a couple questions because I don’t necessarily agree with what they’re saying.

The first question is — how are the Chinese going to react to a strategy where the Americans say we’re going to win this competition? That implies that China is going to have to change its system of government that’s been in place since October 1, 1949. In this party-central state here in China, you’ve got to calculate how the other state is going to react in order to pursue a policy like this.

The second question is — how are the allies going to react to this? If we have a stated policy that we’re going to win now, people in the international system will think that means the kind of victory we had on December 25, 1991, when the Soviet Union imploded and disappeared. What does that pose to Japan, to the Philippines, to Australia, to India out here in the Indo-Pacific, to the NATO allies and EU partners? Because they’re a big part right now of our ability to try to limit the options that the Chinese have around the world.

What will countries in the Global South think of that strategy? Does it pay off for the United States? Does it win us more strategic weight in our relationship with the allies? Can you bring countries along in a strategy like that? I have real doubts that you could do that.

The final question is — what does it do to this really uneasy balance of trying to get certain things done where our interests are aligned? What does it do on issues like climate change, on fentanyl — where we are beginning to see the Chinese help us, but they need to do a lot more — and what happens on the military-to-military side? Would they shut down all communications with us?

You have to think through the upside and downside. I have great respect for those two guys, and we need this type of fermentation and dialogue in our own system in the United States. I just don’t agree with what they put forward — respectfully — in the Foreign Affairs article.

Jordan Schneider: Let me challenge one of your contentions. You just said China has had one system of governance since 1949, but we’ve had many different eras. Yes, the party has existed, but Mao almost overthrew it himself. We had a Deng era, we had a Hu era, and we have a Xi era now. The same thing happened with the Soviet Union — the way the party worked and the way it looked at its relationship with the world changed dramatically over its 75-year history. The one thing I can be sure about is China is going to change a lot over the next 30 years. How, if at all, should the US be thinking about what Chinese governance looks like in a world after Xi?

R. Nicholas Burns: Part of being competitive — and we’re waging a competitive relationship — is to object to much of that system, to object to the denial of basic human rights in Xinjiang and in Hong Kong and Tibet. I’ve gone to church here in multiple cities. We object to the lack of religious freedom in this country. That’s part of what the competition is. It’s normative. It’s about the nature of what governments should and should not do in giving rights or denying rights to their own people.

In saying that, one of the problems with a strategy of winning is, would the Chinese system actually implode? It doesn’t mean we want to preserve the system because we’re competing against major aspects of the system. But you do have to ask the question, what would be the consequence of trying to win, and would that strengthen us in some ways? What would we lose by our inability to work with them and work with others? That’s the central question that’s got to be asked.

Jordan Schneider: If China ever returns to a late ’70s, ’80s ferment, is there anything that the US can and should do to help bring about or prepare for that moment?

R. Nicholas Burns: Jordan, at one point in my career, a long time ago, I was State Department spokesperson for President Clinton, and I learned a big lesson there. Don’t ever answer a hypothetical question. That’s a supremely good but hypothetical question. I’m still the American ambassador to China for another eight days, so maybe ask me six months from now, but not today.

Jordan Schneider: Fair enough. Let’s go back to Kissinger. It’s fair to say that he is one of the most skilled diplomats. He had as many reps as anyone is ever going to get at the stakes that he did. One of the big takeaways I had from your book is two of the things he was most proud of and excited about and thought he did the best job on — Rhodesia and Vietnam — ended up being maybe the two biggest disasters. You could put Bangladesh in another category maybe. What does that mean if the person who is the best you’ll ever get at this and thought he was doing a great job actually had two of the worst outcomes he was ever involved in?

R. Nicholas Burns: Well, in a long career — it’s extraordinary that I’ve just picked up and I’m starting to read the book he wrote, posthumously published, with Eric Schmidt on AI. He was in his 100th and 101st years on earth, writing about AI. Think of the longevity of that career which begins in Nazi Germany and goes through the Second World War and all the way through.

Subscribe now

No one’s going to get everything right. He was intrigued by the fact that we wanted to write a chapter in that book about his attempt to begin to abolish white minority rule in Rhodesia and then to get at the bigger question of the apartheid regime in South Africa. He made clear to us, this is 1975-76, that he and President Ford had decided they wanted to separate themselves from the white regimes in then-Rhodesia, later Zimbabwe and South Africa. He didn’t succeed.

What we did in doing some of the research of the book was to talk to many of the American diplomats and other diplomats from around the world with whom he worked. Most of them would say he didn’t succeed as Secretary of State, but he laid the groundwork for the Lancaster House conference in 1978, just a year after he left power, that overthrew white minority rule and established the modern state of Zimbabwe. He was very interested that we were going to study that because he said it had been forgotten as part of his work. He leaned into it and spent a lot of time with us going through his strategic assumptions and what he tried to do tactically.

The other thing, Jordan, that I remember very clearly from our conversations with him — and I also talked to him about this as I was preparing to come out here — is that there’s been a lot of criticism historically of Secretary Kissinger that he didn’t elevate human rights or the nature of the regime he was negotiating with, in this case, the Soviet Union or Mao’s China. He made a point — and I think Niall Ferguson, his biographer, has been making the same point — that he thought the prevention of nuclear war, driving down the probability of nuclear war, and creating some kind of balance in great power relationships was a highly moral objective.

That was an interesting product of all the conversations that we had with him. As we talk about US-China relationships, we’ve taken the position that we cannot afford not to talk about human rights. I just issued a statement a couple of days ago about a young man named Ekpar Asat, who is a Uyghur young man who’s been incarcerated for the last eight years unfairly because he was trying to promote better relations between Han Chinese and Uyghurs. We’ve decided we’ve got to stand up for the massive human rights deficiencies of the People’s Republic of China here. We’ve made a different calculation decades later, but that’s how he responded to the criticism that he was not sufficiently concerned about human rights.

Jordan Schneider: I really like the Rhodesia chapter. You guys made a great illustration of just how much of a tour de force it was from him at the negotiation table to doing the 4D chess of lining Nyerere up and the South Africans up. You convinced me that it helped get Zimbabwe created. But it’s a cruel irony that the creation of Zimbabwe ended up being one of the biggest humanitarian catastrophes of the past 30 years.

R. Nicholas Burns: The idea was that ending white racist regimes was the right thing to do. Obviously, Kissinger didn’t foresee what would happen with Robert Mugabe in the ensuing decades.

Red Lines and Great Power Responsibilities

Jordan Schneider: Regarding Salt Typhoon — should I be more annoyed with the Chinese or with the US cybersecurity establishment for allowing it to happen in the first place?

R. Nicholas Burns: You should be annoyed with the Chinese. One of the points we’ve been making to the Chinese leadership in recent weeks is, you’ve gone way too far. These are objectionable assaults on American infrastructure, on telecommunications and other aspects of our infrastructure, and on the rights of Americans. There’s going to be a price to pay. We’ve already begun to sanction some of the companies and the hacker groups involved in this. It’s become a major issue in this relationship and not just with us — because the Chinese have been going after these extraordinarily aggressive cyber assaults on many American allies in Asia and in Europe.

We have made it clear to the Chinese, as have some of the allies, that this has got to stop. There have to be boundaries here. What I’ve been thinking about as I wrap up my time here is how aggressive the Chinese have been on multiple fronts — on cyber in the South China Sea, these repeated attempts to intimidate the Philippines, our treaty ally, their attempts to intimidate Japanese administrative control of the Senkaku Islands. We’ve been having conversations with the Japanese about this, and we both objected to the Chinese.

Look at what they’ve done recently over the last year on Taiwan, the simulated blockade earlier in the autumn. Then think about this massive undertaking where all these Chinese companies — and the Chinese deny this, but it’s true — are giving very critical support to the Russian defense industrial base and allowing Russia to continue this war in Ukraine. There are other examples, such as what the Chinese are not doing to use their influence with Iran to stop the Houthi rebels in the Red Sea.

On issue after issue, the Chinese are over-the-top assertive against American, Japanese, and European interests. There’s a price to be paid for this.

I’ve tried to tell people here, “You have united both of our political parties in Washington, and both houses of Congress rightfully are pushing back against you because of the number of assaults against really important, in some cases vital, American interests.”

They’ve just got to understand how big a hole they’ve dug for themselves as they prepare to work with another administration.

Jordan Schneider: Let’s stay on Chinese international aggrandizement. There’s one argument that Putin — before he invaded Ukraine — invaded Georgia, invaded a smaller piece of Ukraine, he sort of took over Syria. You have a Xi track record — what happened on the Indian border is not great, what’s happening right now in the South China Sea, also not great. But this is almost a difference in kind between actually rolling tanks into countries. Is that fair? How do you reflect on what Chinese international aggression looks like when you’re thinking about even more scary eventualities like a blockade or invasion of Taiwan?

R. Nicholas Burns: The problem, Jordan, is that they’re challenging the very basis of country sovereignty. They don’t have a leg to stand on in their claims in the Spratlys and Paracels, these exorbitant claims for the islands and islets hundreds of miles beyond the territorial waters of China. They don’t have any support legally for this, but they’ve done it through force of arms to the Philippines and the other claimants in the Spratlys and Paracels.

They have been extraordinarily aggressive against the Japanese. The number of air sorties and naval sorties on a nearly daily basis — you do have this new agreement between India and China, and yet you see India in the Quad. The Indians are one of our strongest partners now in the Indo-Pacific because of this uncertainty about what China is up to on their long Himalayan border.

Until recently, until the Biden administration came into power, you saw a lot of voices in Europe about taking some kind of middle ground between China and the United States. When President Biden came into office in 2021, the EU was on the verge of a major investment treaty between China and the EU.

Look at the hole the Chinese have dug for themselves now. The European governments have risen up against China because China is assaulting the existential issue confronting Europe — the indivisibility of the continent. It’s the fact that there’s war now in one of the major states of Europe, Ukraine, that Putin is assaulting the sovereignty of Ukraine.

Look at all these mysterious incidents happening in the Baltic Sea with cables being cut and the suspicion that the Chinese were involved in that. They have really harmed their relationships with Europe through this uber-aggressive policy, especially the alignment with Russia, with North Korea, and with Iran.

This gets back to the question I asked at the beginning of this interview that’s been on my mind a lot. At some point, the Chinese are going to have to decide, are they really going to be for the next 10 years with these agents of disorder in the world, or do they want to position themselves differently so that they don’t dig a hole for themselves with the European Union, the United States, Japan, the Philippines, Australia, India? If you alienate all those countries, that’s well more than 50-60% of global GDP.

Jordan Schneider: My theory of why China keeps doing this is that they don’t actually see as many costs as they could, and they don’t really think they’re going to trigger World War Three or global 50% tariffs.

How would you assess this? Biden has placed a lot of emphasis on signing new military agreements, and now we have AUKUS and the Quad — but it seems like China is not getting the message yet. Will they ever? Will that be okay as long as they don’t invade Taiwan?

R. Nicholas Burns: Frankly, in their heart of hearts, I sense that the Chinese are really worried that they’ve lost Europe, maybe temporarily for the life of the Ukraine crisis. We’ll have to see how that plays out over the next couple of years, but they’re worried about that. There’s a massive charm offensive underway by the Chinese Foreign Ministry to try to win back some credibility with the Europeans.

Share

The Europeans have been very clear — with the exception of countries like Hungary, an outlier — about the price that China has to pay for supporting Putin in the war in Ukraine. Similarly, out here in the Indo-Pacific you’ve seen a real toughening of the policies of Japan, South Korea, and the Philippines against what they see as Chinese overreach.

One of President Biden’s accomplishments of over the last four years is that the United States is in a stronger strategic position in the Indo-Pacific in this long-running competition with China. In large part, it’s because we’ve strengthened our alliances with those three countries and with Australia, we’ve created AUKUS, and the Quad now is meeting at the head of government level. I just had a Quad meeting yesterday here in Beijing with my Japanese, Australian, and Indian counterparts. We meet regularly and we are lined up together out here in Beijing. That wasn’t the case five or six years ago, but it’s the case now because of Chinese overreach.

There have been penalties here that the Chinese are very well aware of. One of my takeaways, as I conclude my time here in Beijing, is the allies of the United States in the Indo-Pacific are force multipliers for American power. Academically or politically, diplomatically, a lot of people just look at this competition between the U.S. and China and they try to size up the weight — economic, strategic, military — of the two countries and compare. That’s not a valid comparison on the U.S. side, because China has no allies to speak of. To measure strength, you’ve got to put the U.S. alongside Japan, the Philippines, Australia, and, in many ways, India.

For me as a diplomat, the most important thing we can do in this competition is to keep the allies close. I also say that as a former American ambassador to NATO — and I was there on 9/11. Let me name two countries that stood up on 9/11, and called me immediately in the aftermath of the strikes. The Canadian ambassador called me first and said, “Let’s invoke Article 5.” Then the Danish ambassador called me.

I am an alliance-centered American who believes that these allies are just precious strengths for America because they multiply our power. These are reliable countries that have been with us through thick and thin. As I walk away from this job, I’ve been working very closely with all those allied ambassadors here and we are better off for it. I really hope that the United States is going to remain true to those alliances because they’re in our self-interest.

Jordan Schneider: This gets to a question of what is the scarier world state. There are a lot of instances in history — if you look in the Nazi archives, in the Japanese archives, in the Weimar Germany archives — where you see leaders who see they’re hitting an inflection point. Even if they don’t have good chances, they’re only going to get worse. This thinking has started some really awful and tragic wars.

But there are also times where countries have seen the writing on the wall and realized that maybe there’s a different, more peaceful approach to engaging with their neighbors. What do you think about what shifting global power balances mean for China’s leadership today and maybe tomorrow? What could the U.S. and the rest of the world do to help push the thinking one way versus the other?

R. Nicholas Burns: It’s a very thoughtful question, Jordan. You need an academic seminar on this or we could design a whole course at a university around it. Obviously, the United States needs to keep the allies close. They’re critical for the United States in the 21st century. We are still the single most powerful country without any question across all the dimensions of national power. We want that to continue. We don’t want China to be that country. You want the United States to be that country.

What’s unusual about this century, what’s changed in this century is that you can’t stand alone. Isolationism won’t work in a world that has so many different power vectors and challenges that are transnational in nature. We’ve got to have these allies. That’s point one.

Point two, obviously, the United States needs to strengthen its game in our strategy in the global South. I was in Lima with President Biden. I stayed an extra day and spent some time with Peruvians who are very smart about China, by the way, and they’re very involved with China. It’s painful to reflect upon the fact that 20 years ago the United States was the leading trade partner of nearly every country in South America. Now China is the leading trade partner most South American countries.

We need to have a concerted, bipartisan long-term national strategy to be more engaged economically, politically, and diplomatically in Sub-Saharan Africa, South America, and Central Asia. There’s a real battle for that kind of power influence in the ASEAN countries here in Asia.

We also have to stay engaged as a society with China. One of the peculiar aspects of what happened here because of COVID and the Chinese policy of zero COVID, which made it impossible to come here in my first year or two, is that we’ve only had one American governor visit China in the last five years. It was Governor Gavin Newsom, who came here in October 2023. We’ve only had one congressional delegation over five years. It was an extraordinarily effective delegation because it was bipartisan. The majority leader, Chuck Schumer came out with Senator Mike Crapo of Idaho and four other members of Congress, three Republicans and three Democrats. They were very effective in their meeting with President Xi Jinping on fentanyl. They actually arrived the day of the Hamas attack against Israel, October 7, 2023.

My message when I become a private citizen and meet with members of Congress will be, we need members of Congress to come here. We need China hawks — and I consider myself in many ways a China hawk, by the way — to be here to size up the adversary and to see this country up close. It’s not a nice thing to do. You do it because you need to understand the adversary better in the adversary’s home, on the adversary’s home turf.

Those are some practical things that I think we should try to do on a bipartisan basis because we want to preserve a bipartisan, rough consensus on how to deal with the Chinese. Those are a couple of ideas on how we can strengthen our game.

Jordan Schneider: On how to get China to understand — if things go well in the next 10 years, you have an increasing global national power balance where the U.S. and allies are gaining on China and friends. What can the U.S. do to communicate that to China in a way where they don’t do what Imperial Japan did in 1941?

R. Nicholas Burns: The Chinese are already looking — I read the Chinese press every day with my staff — and trying to make the point as they look ahead to the next couple of months that the United States is an unreliable ally, that the United States is becoming increasingly insular and isolationist and China is not. This is just Chinese propaganda — that the Chinese play by the rules, that they’re invested in the international system, that they’re responsible, that you can count on us.

If we walk away from the Paris Climate Change Agreement, the Chinese are saying if the Americans walk away, they’re going to be irresponsible, but we Chinese will not walk away. We should not want to give the Chinese those successes. Remaining engaged in the international system, if you want to be effective in the global south, you have to be present in arenas that are important to the Global South. That gets to climate change, it gets to trade, it gets to long-term infrastructure lending that won’t bankrupt and lead to massive indebtedness on the part of these countries.

The Biden administration has begun to make real changes there. We just need to have that continued. I hope in the future Congress can play a big role, which is another reason why I really hope we can get senators and members of the House out here, take a look at this competition and figure out from the ground here how we can be more effective. Keeping allies with us is going to be absolutely critical. Being respectful of our NATO allies, being respectful of their basic sovereignty and borders — I never thought we’d have to say that — is very important.

Jordan Schneider: Well, China just is trying to take islands, but at least Trump is offering to buy them. Do you think that’s a meaningful difference?

R. Nicholas Burns: Jordan, can I say this? That’s a really important point you just made. We have had the strategic advantage. We’re winning the argument globally against Russia on whether or not Russia had the right to invade Ukraine — it didn’t. We are winning the argument globally that China does not have the right to use force across the Taiwan Strait, that it ought to commit, it must commit to a peaceful resolution. But if you don’t act like that yourself, if you contest other countries’ sovereignty and borders, allied countries, it gives a pass to Putin and Xi Jinping that we shouldn’t want to give them.

Jordan Schneider: Would you like to offer any advice for the incoming administration and your successor?

R. Nicholas Burns: What I’ve actually tried not to do is to give a lot of specific advice. I’ve been asked a ton of times to comment on this or that statement or what’s the new administration going to do. I’ve declined because obviously they have a right to figure out what they want to do, get their team together. I don’t want to complicate things, but we just talked about matters of high policy about how the United States should act in the world. I gave you my honest answer.

I’ll say this about the new administration: I don’t know Senator David Perdue, but when he was nominated by President-elect Trump, I said publicly that I congratulated him, wished him well, and that I would give him as much help as I can. I’ve reached out to him and I hope to meet him before he goes out to China. I really want him to succeed because if he succeeds, the country succeeds.

We have a huge mission out here. We have one of the largest American embassies in the world in Beijing and our four consulates in Shenyang, Wuhan, Shanghai, and Guangzhou — 48 U.S. government agencies. I think he’s very well qualified, based on his career and what he did in the United States Senate, to be the American ambassador. We out here are nonpartisan as civil service. We were nonpartisan throughout the election. It didn’t come into the lifeblood of this embassy at all. I was really proud of our team.

As I’ve thought about it and explained to people out here, we have enough problems dealing with the government of the People’s Republic of China. We want to stay out of the divides back home. That’s why I thought it was really important for me to say I’m going to give Ambassador Perdue, when he does become ambassador, all the help I can.

Jordan Schneider: Kissinger pioneered many things. One among them was Kissinger Associates. The path where former officials end up serving government relations jobs for countries and companies — has that been like a net positive or more neutral or negative development over the past 50 years, that this is something a lot of former officials end up doing?

R. Nicholas Burns: I think it depends on what the nature of the activity is. I retired actually the first time from the Foreign Service in 2008, and I was a professor at Harvard. But I also consulted for a D.C. consulting firm and I found it really interesting and enriching. People have to make their own choices as to what they do out of government.

Jordan Schneider: What are your biggest outstanding analytical questions about China?

R. Nicholas Burns: The biggest question is how China wants to lead as a global power. Let me give you two examples. Who do they want to work with most prominently? Can they work more effectively with Japan, the United States, Western Europe, the EU, and NATO? They’ve got real problems working with all of us now.

The door is open. We would like to see China be a more responsible country on these big international questions than it has been. The door is open to that kind of cooperation if they’re willing to walk through it, but they have not been willing because they’ve got this intense association with Russia. Russia is a country that’s assaulting the whole edifice of the global system that we built after the Second World War.

I’m afraid that China, through its Global Development Initiative, Global Security Initiative and Global Civilization Initiative, is giving us the impression that it wants to alter the global system and make it more friendly for authoritarian countries. That’s a big question that the Chinese leadership is going to have to reflect upon. Do they really want to stay with North Korea, Iran, and Russia in a loose association? Is that going to be good for Chinese interests?

The second question we’ve been reflecting on is that great powers have responsibilities to do really difficult things — to negotiate a ceasefire in Lebanon, as the United States succeeded in doing, to negotiate a ceasefire in Gaza. We’re not there yet, but we’ve got to have a ceasefire in Gaza and the hostages released. You expend a lot of capital when you do that. You can make yourself very unpopular, but there’s a greater good.

We’ve not seen that kind of attitude from the Chinese. They have an opportunity in the Red Sea because China is the major importer of Iranian oil and they have a very strong strategic relationship with the Iranian leadership. If any country could convince Iran to use its influence with the Houthi rebels, it’s China. They haven’t done it. We have said, and I’ve certainly said to them in repeated meetings, you really ought to use your influence here. The world needs you to do it so the commercial shipping traffic can resume along one of the major routes of the international system.

They’ve been on the sidelines of the crisis in Gaza. They talk every day about the Palestinians, but what have they really done to help the Palestinians? When you get to be a global power — and China is a global power by virtue of its economic and military strength, technological strength — you gotta act like it. We haven’t seen that from the Chinese.

Subscribe now

Jordan Schneider: What open question about U.S.-China relations are you curious about going forward? On what topic do you wish you could read a book that hasn’t yet been written?

R. Nicholas Burns: I believe that China is our strongest adversary now (competitor, adversary, you can choose the word) and they will be 10 years from now, maybe 20 years from now. We have got to be stronger as a country to take on that competition with China.

I worry that we don’t have enough American students here learning Mandarin in the current generation of 20-year-olds. For now, we have a core group of people in the country who understand China, speak Mandarin, and have had experience here. One of my great advantages as ambassador is that this building that I’m in, our embassy, is filled with people on their second, third, fourth China tour. They speak Mandarin, they understand this country. I worry that my successor 10 years from now, 15 years from now will not have that deep bench.

We have to think about this competition carefully. There’s the hawk in me in the sense we’ve got to compete and we can’t afford to become the second strongest power out here in the Indo-Pacific ten years from now. We’ve got to compete and succeed in the competition, but at the same time, we have to engage the Chinese.

It’s not 50-50 — it’s 80% competition, 20% engagement. The engagement part is important because it’s about other national interests that we have. Fentanyl, climate change, and getting American prisoners out of jail — we’ve helped four Americans get out of jail in the last several months because of quiet diplomacy with the Chinese.

If we don’t see both sides of that equation, then we don’t serve American national interests. Because the ultimate goal here is to compete, but also to live with China in the world, avoid war, and drive down the probability of conflict. That doesn’t always come out in the American debate. There’s so much focus on competition. The harder and more complicated question is, how do you do both at once with competition being the overwhelming priority?

Jordan Schneider: You mentioned the generational change in national security leadership. We had this situation where by the time we’re starting wars in the Middle East, all we have are Soviet analysts running the game. Ten, twenty years from now, we’ll probably have ambassadors to China and secretaries of state who speak Mandarin. I’m curious, reflecting about what having the principals also be regional specialists in the place that’s most impactful in the world, or where the U.S. is most focused — what that does and doesn’t change about how policy gets formulated, implemented?

R. Nicholas Burns: My career has been in the American Foreign Service, and I’m so proud of the Foreign Service. I’m actually a big proponent of not being a generalist. Here’s where I disagree, respectfully, with Secretary Kissinger — he didn’t want to have lots of people specializing on Latin America or the Soviet Union or China. He wanted people to have general skills and to have diverse careers.

We need people like that. I was such a person. I served in Africa, the Middle East, Europe, and on Soviet affairs from the White House. I was a generalist. But we need China specialists. I’ve started a program here in our embassy called “Path to Senior Leadership.” We read books together on the U.S.-China relationship. We think through careers — how do you build a career where you become an expert on the history, culture, politics, economics, language of China? I want our younger officers here in China to think about coming back two or three times, and I encourage them to do that.

We also need to have Arabists, obviously, and Latin Americanists. When I worked for Secretary Rice — and I have huge respect for her — we concluded that we needed about half of our officers to be specialists and half to be generalists. That’s probably about right.

A lot of signals are sent to our Foreign Service officers about careers — “Don’t focus on one country, don’t focus on one language, you’re going to get promoted more easily if you do two or three regions.” I don’t always trust that or agree with it. I’ve had a lot of conversations with our great younger diplomats here about becoming the George Kennan of your generation, or the Victoria Nuland or John Beyrle (probably the two great Russia specialists of my generation in the Foreign Service). You’re going to serve the national interest. You’re going to have a fulfilling career.

The Pentagon and the Treasury Department, we all need to think about how we bring along the next generation. I didn’t expect when I came out here as ambassador that I’d spend as much time as I have, but I’m glad I did, trying to work with our younger officers to think through and help train them for the longer-term goal of having profoundly important China specialists in the U.S. government.

We hear a lot of criticism of the federal government, and it’s not perfect, but I’ve served in it a long time and now I’ve come back into it for these past three years. We have tremendously talented people in the Foreign Service, in the Commercial Service, at the Commerce Department, in the Treasury Department, in DoD — and they are a national asset. We’ve got to keep those people in government, encourage them, and thank them more often than we do for their service.

Share

Jordan Schneider: Let’s talk about Martin Luther King Jr. and theories of change. You have the NAACP approach, which involves technocratic methods — filing lawsuits and building brick by brick. Then you had Dr. King saying that as long as you’re pure and spread truth, people will come around whether they like it or not — even if it doesn’t feel popular or tactical and even if JFK is saying you should wait two weeks.

Do you have any reflections on the tension between the prophetic and the technocratic when it comes to international relations?

R. Nicholas Burns: Thank you for bringing up Dr. King. He was somebody I really admired when I was a teenager. I was in sixth grade when he was assassinated, in April 1968. Our sixth grade teacher wheeled a television set into our classroom and said, “We’re going to watch the funeral of a great man.” That awakened me as a 12-year-old to what a heroic figure he is in American history.

Your question about politics and foreign policy — what kind of pace of change should we seek or the degree of change we seek? It depends on the moment, the era, the country, whether that change can be brought about easily or with difficulty. The United States has played the most important role in the world since the end of the war 80 years ago. We’re commemorating that here in China. We’re working to commemorate the end of the war because we were allied with China back then.

A lot of the battle for the future is going to be within the United States itself in terms of what we tell each other about our responsibilities as a great power to the rest of the world and to our own country. As a career diplomat, I deeply believe the United States has to be outward focused, true to its allies, maintain those alliances, see the allies as force multipliers, as adding to American power. They also add to the normative struggle that we have with communism — communism in China, communism as pseudo-communism or authoritarianism as practiced by Vladimir Putin. We differ with them and we object to the fact that they deny individual liberties and individual freedom, and we support that.

We can’t win these battles if we stay at home and forsake our alliances and don’t think strategically about American power overseas. Part of what’s going to happen in the next 10 or 20 years will depend on our own national conversation about these issues.

Jordan Schneider: I’m coming off paternity leave, and as a new father I’ve been reading a lot of Civil War history.

R. Nicholas Burns: Congratulations! I’m reading Jon Meacham’s book on Lincoln, And There Was Light. It’s a great book about Lincoln’s lifelong focus on the issue of slavery starting from the time he was a little kid, and it’s really been enlightening for me. It’s not a complete biography — it’s really about the question of race and slavery leading up to the Emancipation Proclamation.

Jordan Schneider: Lincoln famously, a month before releasing the Emancipation Proclamation, wrote a letter stating that if he were to free the slaves, it would only be because it was necessary to save the Union. A lot of historians over the past 150 years have given him a really hard time for that.

Some argue that what Lincoln was doing was making it easier for the Democrats in the North and the Copperheads of the world to stomach what he was about to do with the Emancipation Proclamation.

Relating that to today — that’s a very un-MLK-like thing to do, to disavow the thing that you most believe in, that you’re standing on from the strongest moral perspective. Thoughts on that? That’s a really interesting contrast. I’m curious how that relates to idealism and how America should walk in the world.

R. Nicholas Burns: I am neither a Civil War historian nor an expert on Lincoln, but consult Jon Meacham. Meacham says Lincoln had to deal with the Copperheads, the anti-war Democrats in the Northern states. He had to deal with the border states and keep everything together and defeat the Confederacy. What I get from Meacham is that Lincoln had to do that and bring people along slowly. His views also evolved towards the very end of his life, holding a radically different view. But that’s just based on my reading.

Jordan Schneider: This brings us back to the Kissinger question — what types of things should the U.S. be willing to hold its nose at for the sake of advancing a greater good? Where do you draw the line?

R. Nicholas Burns: I’ve really enjoyed this conversation and I enjoyed my time as a professor when we could have lots of conversations about this. We can have another conversation when I’m a private citizen.

But in that context, here’s how I would answer your question. I’ve lived in eight countries in the last 52 years, started when I was 17, living overseas and in Africa, in the Middle East, in Europe and in Asia, places where I’ve lived and other places where I’ve traveled.

What stands out in the minds of foreigners about the United States is we believe in human freedom. That’s how people see us. While we are not a perfect democracy, we are a democracy and we have a balance of power, and our constitutional freedoms and the Bill of Rights are the backbone of this country.

The second thing people really appreciate about the United States is that, however imperfect we have been in getting some things wrong, like the Iraq War back 20 years ago, we are devoted to being a responsible global power and we’re a good ally. When other countries get in trouble, we back them up. As I saw on 9/11 when I was ambassador to NATO, when we get in trouble, our best allies back us up, like Denmark and Canada.

We’ve got to protect those two strengths that we have in the world. It’s our moral strength that comes out of the Declaration of Independence and the Constitution and the way we’ve acted inside our country, and it’s the way we’ve acted responsibly, particularly since the end of the Second World War, in the great generation of Truman and Eisenhower and Kennan that led us to be permanently engaged because that was the American national interest.

If you look at our national conversation, some of that is up for grabs right now and is being questioned. I tend to be a defender of the faith. Stay with what made America great. That was both of those things: our democracy at home and our reliability as an ally and partner of like-minded democratic countries in the Indo-Pacific and Europe and elsewhere in the world. There’s a lot at stake here for us to get these two right. I know what side I’m on on both of these questions.

We ran a fun podcast earlier this week with Divyansh Kaushik talking about the tech bros vs MAGA fight where we got into implications for immigration and AI policy as well as education and the Asian immigrant experience in America. Check it out on iTunes, Spotify, or our favorite podcast app.

DeepSeek’s Three Edges

An excerpt from Kevin Xu’s excellent Interconnected Substack

Three idiosyncratic advantages that make DeepSeek a unique beast.

These are idiosyncrasies that few, if any, leading AI labs from either the US or China or elsewhere share. Thus, understanding them is important, so we don’t over-extrapolate or under-estimate what DeepSeek’s success means in the grand scheme of things.

No Business Model

DeepSeek is incubated out of a quant fund called High Flyer Capital. Its AI models have no business model. How this came about can be understood from its unique corporate history.

High Flyer Capital’s founder, Liang Wenfeng, studied AI as an undergraduate at Zhejiang University (a leading Chinese university) and was a serial and struggling entrepreneur right out of college. He finally found success in the quantitative trading world, despite having no experience in finance, but he’s always kept an eye on frontier AI advancement.

When ChatGPT took the world by storm in November 2022 and lit the way for the rest of the industry with the Transformer architecture coupled with powerful compute, Liang took note. DeepSeek, as an AI lab, was spun out of the hedge fund six months after ChatGPT’s launch. It is internally funded by the investment business, and its compute resources are reallocated from the algorithm trading side, which acquired 10,000 A100 Nvidia GPUs to improve its AI-driven trading strategy, long before US export control was put in place.

DeepSeek’s stated mission was to pursue pure research in search of AGI. This idealistic and somewhat naive mission – not so dissimilar to OpenAI’s original mission – turned off all the venture capitalists Liang initially approached. DeepSeek’s failure to raise outside funding became the reason for its first idiosyncratic advantage: no business model.

A lack of business model and lack of expectation to commercialize its models in a meaningful way gives DeepSeek’s engineers and researchers a luxurious setting to experiment, iterate, and explore. Despite having limited GPU resources due to export control and smaller budget compared to other tech giants, there is no internal coordination, bureaucracy, or politics to navigate to get compute resources. No one has to wrestle between using GPUs to run the next experimentation or serving the next customer to generate revenue.

Almost no other leading AI labs or startups in either the US or China has this advantage. OpenAI used to have this luxury, but it is now under immense revenue and profit pressure. Evidently, OpenAI’s “AGI clause” with its benefactor, Microsoft, includes a $100 billion profit milestone! Every other AI shop you’ve heard of – Anthropic, Mistral, xAI, Cohere, 01.ai, Moonshot – has revenue and commercialization expectations in one flavor or another. That inevitably leads to constant internal friction between the sales team that needs to sell compute capacity to make money, and the R&D team that needs to use compute capacity to make technical progress.

But not DeepSeek! Have a hunch for an architectural breakthrough? Do a training run and see what happens. Want to test out some data format optimization to reduce memory usage? Go test it out.

Runs Own Datacenter

One of DeepSeek’s idiosyncratic advantages is that the team runs its own data centers. Unlike OpenAI, which has to use Microsoft’s Azure, or Anthropic, which has to use Amazon’s AWS, or 01.ai, which has to use Alibaba’s cloud platform, DeepSeek racks its own servers.

To be clear, having a hyperscaler’s infrastructural backing has many advantages. Not needing to manage your own infrastructure and just assuming that the GPUs will be there frees up the R&D team to do what they are good at, which is not managing infrastructure. However, having to work with another team or company to obtain your compute resources also adds both technical and coordination costs, because every cloud works a little differently. Meanwhile, when you are resource constrained, or “GPU poor”, thus need to squeeze every drop of performance out of what you have, knowing exactly how your infra is built and operated can give you a leg up in knowing where and how to optimize.

Software-to-Hardware Optimization Expertise

If you combine the first two idiosyncratic advantages – no business model plus running your own datacenter – you get the third: a high level of software optimization expertise on limited hardware resources.

This expertise was on full display up and down the stack in the DeepSeek-V3 paper. By far the most interesting section (at least to a cloud infra nerd like me) is the "Infractructures" section, where the DeepSeek team explained in detail how it managed to reduce the cost of training at the framework, data format, and networking level.

Its training framework is built from scratch by DeepSeek engineers, called the HAI-LLM framework. To increase training efficiency, this framework included a new and improved parallel processing algorithm, DualPipe. At the heart of training any large AI models is parallel processing, where each accelerator chip calculates a partial answer to all the complex mathematical equations before aggregating all the parts into the final answer. Thus, the efficiency of your parallel processing determines how well you can maximize the compute power of your GPU cluster.

This framework also changed many of the input values’ data format to floating point eight or FP8. FP8 is a less precise data format than FP16 or FP32. Think number of decimal places as an analogy, FP32 has more decimals than FP8, thus more numbers to store in memory. This reduced precision means storing these numbers will take up less memory. The bet is that the precision reduction would not negatively impact the accuracy or capabilities of the resulting model. This method, called quantization, has been the envelope that many AI researchers are pushing to improve training efficiency; DeepSeek-V3 is the latest and perhaps the most effective example of quantization to FP8 achieving notable memory footprint.

The networking level optimization is probably my favorite part to read and nerd out about. There are two networking products in a Nvidia GPU cluster – NVLink, which connects each GPU chip to each other inside a node, and Infiniband, which connects each node to the other inside a data center. In the H-series, a node or server usually has eight chips connected together with NVLink. Since we know that DeepSeek used 2048 H800s, there are likely 256 nodes of 8-GPU servers, connected by Infiniband. With NVLink having higher bandwidth than Infiniband, it is not hard to imagine that in a complex training environment of hundreds of billions of parameters (DeepSeek-V3 has 671 billion total parameters), with partial answers being passed around between thousands of GPUs, the network can get pretty congested while the entire training process slows down. To reduce networking congestion and get the most out of the precious few H800s it possesses, DeepSeek designed its own load-balancing communications kernel to optimize the bandwidth differences between NVLink and Infiniband to maximize cross-node all-to-all communications between the GPUs, so each chip is always solving some sort of partial answer and not have to wait around for something to do.

I don’t pretend to understand every technical detail in the paper. And I don't want to oversell the DeepSeek-V3 as more than what it is – a very good model that has comparable performance to other frontier models with extremely good cost profile.

However, what DeepSeek has achieved may be hard to replicate elsewhere. Its team and setup – no business model, own datacenter, software-to-hardware expertise – resemble more of an academic research lab that has a sizable compute capacity, but no grant writing or journal publishing pressure with a sizable budget, than its peers in the fiercely competitive AI industry. These idiocracies are what I think really set DeepSeek apart.

How Much Did They Really Spend?

Nathan Lambert recently published an excellent breakdown of Deepseek V3’s technical innovations and probed more deeply into the $6m training costs claim. An excerpt below.

The cumulative question of how much total compute is used in experimentation for a model like this is much trickier. Common practice in language modeling laboratories is to use scaling laws to de-risk ideas for pretraining, so that you spend very little time training at the largest sizes that do not result in working models. This looks like 1000s of runs at a very small size, likely 1B-7B, to intermediate data amounts (anywhere from Chinchilla optimal to 1T tokens). Surely DeepSeek did this. The total compute used for the DeepSeek V3 model for pretraining experiments would likely be 2-4 times the reported number in the paper.

A true cost of ownership of the GPUs — to be clear, we don’t know if DeepSeek owns or rents the GPUs — would follow an analysis similar to the SemiAnalysis total cost of ownership model (paid feature on top of the newsletter) that incorporates costs in addition to the actual GPUs. For large GPU clusters of 10K+ A/H100s, line items such as electricity end up costing over $10M per year. The CapEx on the GPUs themselves, at least for H100s, is probably over $1B (based on a market price of $30K for a single H100).

These costs are not necessarily all borne directly by DeepSeek, i.e. they could be working with a cloud provider, but their cost on compute alone (before anything like electricity) is at least $100M’s per year.

For one example, consider comparing how the DeepSeek V3 paper has 139 technical authors. This is a very large technical team.With headcount costs that can also easily be over $10M per year, estimating the cost of a year of operations for DeepSeek AI would be closer to $500M (or even $1B+) than any of the $5.5M numbers tossed around for this model. The success here is that they’re relevant among American technology companies spending what is approaching or surpassing $10B per year on AI models.

At this rate it will be true that you can train a model at the performance of DeepSeek V3 for ~$5.5M in a few years. For now, the costs are far higher, as they involve a combination of extending open-source tools like the OLMo code and poaching expensive employees that can re-solve problems at the frontier of AI.

The paths are clear. DeepSeek shows that a lot of the modern AI pipeline is not magic — it’s consistent gains accumulated on careful engineering and decision making. In face of the dramatic capital expenditures from Big Tech, billion dollar fundraises from Anthropic and OpenAI, and continued export controls on AI chips, DeepSeek has made it far further than many experts predicted. The ability to make cutting edge AI is not restricted to a select cohort of the San Francisco in-group. The costs are currently high, but organizations like DeepSeek are cutting them down by the day.

Earlier last year, many would have thought that scaling and GPT-5 class models would operate in a cost that DeepSeek cannot afford. As Meta utilizes their Llama models more deeply in their products, from recommendation systems to Meta AI, they’d also be the expected winner in open-weight models. Today, these trends are refuted. Meta has to use their financial advantages to close the gap — this is a possibility, but not a given. I certainly expect a Llama 4 MoE model within the next few months and am even more excited to watch this story of open models unfold.

Hardware Alone Can’t Win the AI Race

Ritwik Gupta is a PhD candidate and AI researcher at UC Berkeley. In this piece, he introduces the overlooked role of software in export controls.

The Chinese large language model DeepSeek-V3 has recently made waves, achieving unprecedented efficiency and even outperforming OpenAI’s state-of-the-art models. This is an eyebrow-raising advancement given the USA’s multi-year export control project, which aims to restrict China’s access to advanced semiconductors and slow frontier AI advancement.

Trained on just 2,048 NVIDIA H800 GPUs over two months, DeepSeek-V3 utilized 2.6 million GPU hours, per the DeepSeek-V3 technical report, at a cost of approximately $5.6 million — a stark contrast to the hundreds of millions typically spent by major American tech companies. The NVIDIA H800 is permitted for export — it’s essentially a nerfed version of the powerful NVIDIA H100 GPU. DeepSeek’s success was largely driven by new takes on commonplace software techniques, such as Mixture-of-Experts, FP8 mixed-precision training, and distributed training, which allowed it to achieve frontier performance with limited hardware resources.

Mixture-of experts (MoE) combine multiple small models to make better predictions—this technique is utilized by ChatGPT, Mistral, and Qwen. DeepSeek introduced a new method to select which experts handle specific queries to improve MoE performance. Mixed precision training, first introduced by Baidu and NVIDIA, is now a standard technique in which the numerical precision of a model is variably reduced from 32 to 16-bits. DeepSeek-V3, interestingly, further reduces the precision of the model to 8-bits during training, a configuration not commonly seen previously. DeepSeek crafted their own model training software that optimized these techniques for their hardware—they minimized communication overhead and made effective use of CPUs wherever possible.

This remarkable achievement highlights a critical dynamic in the global AI landscape: the increasing ability to achieve high performance through software optimizations, even under constrained hardware conditions. A recent paper I coauthored argues that these trends effectively nullify American hardware-centric export controls — that is, playing “Whack-a-Chip” as new processors emerge is a losing strategy. We reverse-engineer from source code how Chinese firms, most notably Tencent, have already demonstrated the ability to train cutting-edge models on export-compliant GPUs by leveraging sophisticated software techniques.

We explore techniques including model ensembling, mixed-precision training, and quantization — all of which enable significant efficiency gains. By improving the utilization of less powerful GPUs, these advancements reduce dependency on state-of-the-art hardware while still allowing for significant AI advancements. DeepSeek-V3’s advanced capabilities appear to validate the paper’s thesis.

As software-driven efficiencies accelerate, resource-constrained entities will increasingly be able to compete with larger, well-funded organizations. But by focusing predominantly on hardware, U.S. policymakers have overlooked the transformative potential of software innovations, inadvertently enabling adversaries to maintain technological parity through creative workarounds.

Hardware-only export control strategies can be made more effective by hinging themselves on concrete benchmarks that account for changing software. The field of machine learning has progressed over the large decade largely in part due to benchmarks and standardized evaluations. The United States’ security apparatus should first concretely define the types of workloads it seeks to prevent adversaries from executing. Then, it should work with the newly established NIST AI Safety Institute to establish continuous benchmarks for such tasks that are updated as new hardware, software, and models are made available. A data-driven approach can provide more comprehensive assessments on how adversaries can achieve particular goals and inform how technologies should be controlled.

Simultaneously, the United States needs to explore alternate routes of technology control as competitors develop their own domestic semiconductor markets. Limiting the ability for American semiconductor companies to compete in the international market is self-defeating. The United States restricts the sale of commercial satellite imagery by capping the resolution at the level of detail already offered by international competitors — a similar strategy for semiconductors could prove to be more flexible.

Read more

Read more

Read more