The Great Reckoning —
US-China AI Race, $1.7T Bubble
& the Stag Hunt

"I was born during the Cultural Revolution... my mother sent a letter to Mao's wife because she had closed the ballet school that my mother had helped co-found and was sent to be re-educated, and this is why I was actually born on a Chinese re-education farm during the Cultural Revolution."

"I'm a US citizen for over 45 years... I moved here when I was very young."

"It's worth noting your brother is the first nuclear sub commander in the US Navy."

"He was a senior officer in a nuclear submarine, one of the big boomer ones that has the capability to destroy nations. In fact, three of his four children are now active officers in the US Navy."

"This is probably one of the most critical questions that we need to get resolved because having a race condition forces people to make irrational decisions. There is a perceived race condition and it's based on some assumptions that I think are actually misguided, there's a perception that there is a finish line, that the world is zero sum, that whoever gets to AI first can rule the world forever."

"The world is not zero sum, as you guys talk about every day, every episode. There is no clear finish line. As these open source models advance, the gap is moving from a year and a half to probably two or three months between open source and closed source models."

"It's not like the space race, the space race, you say 'we've landed on the moon, we've won,' there is an end. An arms race type model that we are in today is a constant spend and constant pursuit without clear value being returned."

"In your essay 'Misdiagnosing the US-China AI Race,' you lay out a staggering shift. Chinese open models climbed from 2% to 61% of OpenRouter traffic in the last 2 years. Alibaba's Qwen model alone has over 700 million downloads."

"It's up to a billion now, with 180,000 derivative models."

"The whole open source strategy, people think this is Chinese strategy to destroy American economics and pop this bubble. The reality is that it's an emerging strategy. I was talking to friends at DeepSeek after they came out. Before that nobody knew who they were, they were not on the radar, they were not funded by the government, nobody told them to open source. The CEO of the company is very open-source-minded and thought this is a great technology, I want to share it with everyone."

"When they first did that, they got their hands slapped, the government was like, 'This is such a great model, why are you open-sourcing it?' But because of the soft power value, the PR value from having a local champion, they became celebrated. And then essentially most companies in China followed this, it became the de facto emergent standard."

"The week after we stopped the Chinese from buying all the high-end chips from the US, I had calls with friends in the semiconductor industry, every one of them got calls from the government saying, 'Would you like some extra funding? Extra resources? How can we help you accelerate? Could we get you customers?'"

"A few of the CEOs of these Chinese GPU companies were saying 'nobody wanted to buy our stuff, we're two or three generations behind, we were less energy efficient, but now we can't make enough because every data center in China has to buy our stuff. We would have died if it wasn't for American policies.' We created the current competition: Moore Threads, Cambricon, Biren, would not be as successful or maybe would have gone bankrupt if it wasn't for American policies."

"The actual training right now is not even happening in China because they don't have the Blackwell generation chips in China. They're actually doing it in international data centers, training it and then bringing it back on a disc or something."

"What's the point of a chip embargo? I think it's more optics than anything right now."

"The distillation thing. I think it's more of a PR tool that certain companies are using. Right now there's only one company that is kind of against that. Anthropic. They're the ones lobbying the government saying 'we're being distillation attacked,' which is actually a word they kind of invented."

"Every lab, both domestic and international, distills from each other. I went back and did an estimate of what it would cost to do the number of queries based on average responses on Anthropic's highest models, it was like two or three million dollars across three different labs. For DeepSeek, I think it was only in the thousands of dollars."

"If you have a model that you spend a billion dollars on and somebody can distill and duplicate with a couple million dollars, then the whole economics of Frontier AI doesn't make sense."

"Every month, Meta spends somewhere between $100 and $200 million on Anthropic tokens. They're one of the biggest buyers of AI from Anthropic. If anybody was going to distill, they would have been distilling for the last year to two years, and they just finally got a model out that is somewhat competitive in the last week or two."

"There was a couple models ago where if you used Chinese to ask Anthropic Claude what model you are, and they said 'I'm Qwen.' So they're distilling from us. We just can't tell because they already downloaded our models."

"Right now, 45% of the US stock market value is in the AI sector. That is a very, very fragile place for us to be. At the height of the internet bubble, I think around 30% of the stock market was internet companies."

"The Buffett indicator says a market is healthy when your stock market is the same value as your GDP. At the height of the internet bubble we were around 120% of GDP. Right now we are at 240% of GDP. In fact, the AI sector alone is worth more than the GDP of America today. That to me is a sign that we are in a very, very fragile place and an economic correction is due."

"There is right now $1.6 trillion — $1.7 trillion, of off-the-book debt of the major hyperscalers today. During Enron days there was $200 million of off-the-book debt. Just to give some context of the scale of the kind of problems we are looking at."

"If you look at Amazon and OpenAI revenues when they talk about AI, probably more than half of it comes from two companies. That is not a very diversified revenue base."

"The idea of the Prisoner's Dilemma is that two guys are in prison and they have to defect on each other. On a single-turn Prisoner's Dilemma, the optimal thing is to snitch. In other words, if there was a US-China AI control policy of saying 'we're not going to release, we're going to put safety in place first,' but the country that says 'no, we just developed AGI, we're going to let it loose', that would be defecting."

"The world is not a single-turn game. The world is actually a multi-turn game. Even in Prisoner's Dilemma, a multi-turn game, the game theory optimum is tit-for-tat, you start with cooperation, and if they defect, you defect, and long-term you both go to cooperation."

"In a Stag Hunt game, it's actually a positive-sum game. Whether you both defect or both cooperate, you get two stable Nash equilibriums. The Stag Hunt was invented by Jean-Jacques Rousseau, two hunters go into a forest. You could decide: do I go for the rabbits or do I go for the stag? The stag, because it's bigger, needs two people to hunt together. If I hunt the stag by myself, I get nothing. If I go for rabbits by myself, I feed my family for a week. If we both go for the stag, we feed both families for a month."

"Right now, we're doing the worst thing, we're going for the stag and China's going for the hare. They're going for the 'good enough' AI that helps the economy today. And we're going for the giant AGI that's going to solve everything. When you do that alone, the worst situation is created: if America doesn't get there or if it gets there and cannot control it and it runs away, it gets zero, whereas China continues with their little gains and survives."

"The optimal solution for a Stag Hunt game is: first both slow down, both make good enough stuff, get to a point where the technology is helping grow the economy and we understand how to manage it, and then together go hunt the stag. We've put ourselves into this Prisoner's Dilemma game where we think 'just defect, just defect.' It's a self-imposed game."

"If we roll back to the Cold War, the US and USSR were competing on an arms race. Essentially the reason we won was not because we sent a missile and blew up the Soviet Union, it's because they bankrupted themselves building military arms, up to 15–20% of their GDP on arms that was not creating real value for their society. In some ways, this is what China is doing to us. They're spending one-tenth as much on data centers and getting to 97% as good."

"We are leveraging hundreds of billions of dollars. Last week Nvidia announced they have a $500 billion deal with BlackRock, Carlyle, Blackstone to essentially securitize chips and compute. This sounds a lot like the subprime issues."

"This is the first time I've heard anyone analogize the enormous amount of private credit that the West is allocating to compute to a reverse SDI Star Wars moment. Presumably what you're gesturing at is that could lead to the Chinese century and the collapse of Western dominance."

"We are actually right now acting like the USSR. What perpetuated the arms race was this missile gap, the idea that 'oh, they have more missiles, we have more.' In both cases, they were having the wrong numbers provided to leadership. We're doing the same thing with AI."

"They are spending a ton more on energy generation. They're building more new electric generation than the rest of the world combined, about 10× what the US is every year. They're trying to electrify their society because right now 40% of their oil is imported."

"They are building giant solar farms and wind farms on the west side and in the desert parts of China, then using their very high voltage power transmission, a thousand miles, you lose less than 1% of electricity, bringing the energy to the east where most of the population is living."

"Their energy cost is around two to three cents per kilowatt hour, which is in some cases 10 or maybe 15 times cheaper than many parts of the US. In the long run, that's actually where the constraint will be."

"When I was in China, every lab I talked to, I said 'Do you guys have enough compute?' 'No, this is our biggest issue. We don't have compute.' So from an export control perspective, are we slowing down China? Yeah, the export controls of chips is slowing down China."

"I went back and looked across security use cases, biological, chemical, cyber, from millions of parameters to trillions of parameters. There was no correlation between risk in the real world versus size of models. You had 10 to 50 million parameter models for chemicals that were creating chemical warfare weapons. You had 1 to 50 billion parameter models that were able to create viruses and genetically engineered harmful agents."

"Mdash from Microsoft came out with a CyberGym score of 95: Mythos was 83 or 84, and what Mdash did was took a hundred tiny models and just organized them together to use different skill sets."

"The danger is already there today, particularly on the bio-chemical side. Nobody's talking about it. Nobody's working on protecting the world from those agents, because a 10 million or 1 billion or 5 billion parameter model will run on your laptop in your basement, and you can design these genes or chemical weapons."

"Those are already out. A lot of those are open source. How do we make sure that the precursors are controlled? That the synthesis machines that generate these designs into real genetic material are controlled? These are the kind of things that should be higher up in the agenda, and it's not right now."

"The UK AISI, their AI Security Institute, just came out with a new report last month. What it showed was the cyber threat capabilities of the leading open models including Kimi was about half of the capabilities of Mythos and GPT 5.6. For their highest level of autonomous attack, none of the Chinese models were able to autonomously attack and control an external system."

"Out of the 30-something levels, the average American models were able to get to 20–25 in terms of how far they went in attacking a network. I think none of the Chinese were able to get to the higher levels, and a very small number got to the lower levels like four or five."

"AI is like a match. You can say, 'Here's Qwen 40B, look, it doesn't burn. Then you take Kimi K3 and you light it and you're like, oh, this is a match. Now it's burning. You say, I'm going to try to burn this microphone with it, look, it didn't burn, it's not dangerous, go ahead and let it out.' But it's a match that's burning. If I take that match and I use it to light a piece of paper, and I use that piece of paper to light a tree, and then I try and burn this microphone, it ignites, and then it's unstoppable."

"You're testing the wrong thing. You're testing it out of the box as opposed to its self-improving version, which I know for a fact it can do. I have 5,000 Kimis running tomorrow: 5,000. I guarantee you it can improve itself."

"There is 150-plus humanoid robot companies. I think when I was at WIC there was over 200 robotics companies related to humanoids demonstrating stuff. Which is crazy, because the total volume of humanoid robots last year was in the tens of thousands globally, and 80–90% came from two or three companies. There is no market right now for that many companies to exist."

"If you went to WSC a year ago, there was about 150 labs demonstrating large language models. Now there's really 10 that are relevant. In the next year or two, you'll see that 150 go down to a single-digit number of surviving humanoid robot companies."

"Humanoid robots are actually not really good form factors for doing much of anything right now. I was just at Unitree headquarters, almost all of our customers are research labs buying our stuff, and also some doing demos, kickboxing, but there's very little of these machines being used in commercial practice."

"The Unitree guys said right now they're moving to an only upper-torso model, those are actually selling better in the commercial space because having feet is a negative. You have to keep balancing it. When it falls, bad things happen. It's actually having less components and a big base with a big battery, it lasts longer."

"If you look at the current negative sentiment in the youth today, every student is having a tough time finding jobs and they're just very worried. I'm spending a lot of time in universities and so does Dave. You can see the anxiety within the youth because of their difficulties in finding internships."

"There's about 10 or 20% AI adopters on campus and a violently opposed 70–80%. This is a really big cultural gap. The AI-aware people are busy talking to their agents. The other side is just mad."

"Over the last 40 years, people have seen their lives get better and better in China, 800 million people have been risen out of poverty. A lot of that is attributed to technology adoption. People see this as the next step saying, 'Here's another technology that will make our lives better.'"

"China has a lot more control over the media. They're mostly good news all the time. They're essentially the 'always good news network.' You don't hear about plane crashes and murders on Chinese news. It's all about some new invention came out, some new building went up."

"A year and a half ago they came out with something called the AI Plus Plan. Their idea is: within the next 5 years, we want to have 70% of companies integrate AI into their business, manufacturing, education, medicine. Within the next 10 years, we want 90-plus percent. It's all about fusion and deployment into industry and society."

"This is different than the American AI action plan that came out last year. The American AI action plan is on the supply side, how can we create the best models, how can we dominate in the best chips, but it doesn't talk about what happens after. There was nothing in the AI Plus plan that says we have to get to AGI, we have to dominate. It says how do we get more industries to use this, adopt it in a smooth safe way so that it grows the economy. That was all they cared about."

"Almost every single leadership in the senior leadership of the central party are actually engineers, probably 80 or 90% of them. They're actually quite technical."

"In the US, most of our politicians are lawyers. In China, most of the politicians are engineers. I found that a fascinating distinction."

"With all the industries and innovation that's happened in that country over the last 30 or 40 years, it was never a top-down thing. There were maybe directional things, they would say, 'Hey, for the next 5 years, we should work on clean energy, automation of robotics, add AI to that.' When the central party initiates these plans, the provinces say, 'What companies do we have in our area that support this goal? Let's go find them, support them, give them stipends, free recruiting, benefits.' They'll have provincial champions and city champions. The 30-plus provinces all compete against each other."

"It is actually very distributed in terms of how these plans get initiated. Nobody's saying 'You need to use this technique to go do that.' They're actually a very highly competitive landscape between all of these labs."

"Post-World War II, we were 50% of the manufacturing capability of the entire world. Right now, around 35% of global manufacturing capabilities is in China, about 15% in the US. The forecast is it's going to go to 40 or 45% over the next 5 or 10 years."

"America right now is very good at financial services, creative services, consulting services, things that are informationally-driven and things that are actually highly susceptible to AI exposure for displacement. We are running this race to get to AGI which is the force that will actually displace us from global preeminence because we are commoditizing the very sectors that we are strong in."

"The US workforce right now is around 70% white collar workers. White collar workers, as you guys all know, are the first to be displaced by AGI when it arrives. China is around 40%. Africa is probably 10 to 20%. Hard manufacturing industry is something that even when we have AGI, we will still need those type of facilities."

"Re-industrialization is definitely needed. The long-term workforce redistribution is not going to be going back to manufacturing, we're not going to create 180 million workers going into manufacturing. With automation that's coming, that number will go less and less, just like farming. We used to be 80% farmers, now it's less than half a percent."

"What we will actually move to is service, but not accountants and lawyers. Service in the sense of teachers and nurses, elderly care, things that require human-to-human services. That is the labor pool that will be able to absorb the 60 or 70% of displaced future workers."

"If I was a mid-tier, low-tier McKinsey employee, I'd be very worried right now. I've talked to partners at consulting firms, accounting firms, lawyers, they are all saying, 'We actually don't need these junior guys anymore. We can do more work, faster, with a few senior guys and an AI system.'"

"China has a lot more control over the media. They're mostly good news all the time. It's essentially the 'always good news network.' You don't hear a lot about plane crashes and murders on Chinese news. It's all about some new invention came out, some new building went up."

"Chinese games and Chinese movies, you can't show blood. Everything has to go through essentially a censorship board. They have green blood instead of red blood in games."

"When we're watching the news every night, we're training our neural net, our 100 billion neurons, our 100 trillion synaptic connections. If there's fear-mongering on the news all the time, that's how you think about the world."

"The youth unemployment in China is probably around 20%. Youth unemployment in the US is around 9%. But it's actually worse than that, there's around 42% underemployment. For college grads, if you're working as a gig worker or as a barista, that counts as being employed. If you take the 20% plus the 42%, there's essentially half of college grads not getting proper jobs in China."

"This is why there's that 'lying flat' issue the last few years. The problem is that young people, because of the one-child policy, have been told how great they are their entire life. Their whole parents and grandparents are putting their hopes on this one generation. When they get out in the real world, it is hyper-competitive and they have to do grunt jobs and they don't want to do it. They say, 'I'm going to lie flat. I'm just going to stay at home and do nothing.'"

"There is an assumption right now in DC that the reason China wants to invade Taiwan is to get access to these fabs. The reality is that if anybody attacks the island, there is nothing there to be had in terms of workable fabs."

"I was having breakfast with the CTO for TSMC and a former senior official from the CIA. The TSMC guy goes, 'Hey, I heard that you guys are going to blow up our data centers if China attacks. Is that true?' And the CIA guy says, 'I can never confirm nor deny that. But what we do have is we have a thousand engineers of yours that we know we will fly out before anything happens.'"

"If China actually does attack Taiwan, they will not get these fabs, and they realize that. Fabs by themselves require materials from all over the world, maintenance, chemicals, supplies. Even if the US doesn't blow up the data centers, even if the Taiwanese don't sabotage their own systems, after a little while you run out of supplies."

"The reason China cares about Taiwan is not because of the fabs. It is absolutely because of political history. The 1949 movement of the nationalist party to Taiwan, that to them is an uncompleted civil war. The two countries actually right now is recognized by America including 190 other countries as being one country. This is like Hawaii and the US, or Puerto Rico and the US."

"Just like during the 2008 great financial crisis, China helped out the US a lot in terms of keeping financial stability. If China actually started to sell T-bills versus buying, it could have completely destabilized the American financial system. They kept buying, trillions of dollars, which helped keep interest rates down."

"We potentially might have a repeat of this situation if there's a correction in the market due to the overbuild and over-leverage of the AI sector. Maybe at that point, the Americans, or maybe Trump, will give a call to Xi and say, 'Hey, can you help us out again?' And maybe I'll just be more hands-off or more clear instead of the ambiguity issue."

"What I hear you saying in your war game is sometime in the next two years, there's a private credit bubble that the US is using to finance its data center buildout. The bubble pops and then the US asks China to help financially in return for a quid pro quo regarding Taiwan."

"Not in the sense of 'here's Taiwan,' but to say, 'Hey, as long as you agree to some kind of a peaceful thing over a mutually agreed term, we're going to stay out of it.' We've been very involved in the Taiwanese political sphere for a long time. We've been selling weapons to them for the last 40 or 50 years. At one point we had soldiers based in Taiwan. We still have military advisers based in Taiwan. This is like saying if the Chinese were selling weapons to Puerto Rico and helping fund them, what would America do?"

"The key right now is that we don't need to get to a solution on day one. The success factor of this discussion is not that we come out with a massive framework that solves everything. If we can come out of these discussions saying that we're going to have a second discussion, that's already success."

"In the past, there was a discussion in 2024, the dialogue. There was a lot of disappointment because China didn't come with all their technology people, and we didn't come up with a solution. You need to understand: just like the Chinese work in decades for their strategic plans, they also take a lot longer to prepare for these diplomacy discussions."

"For the May visit, there was really no discussion on any of this until the day before between US and China. To the Chinese, that was chaotic and very unprofessional. They're like, 'How can you guys be sending your president here and you haven't talked to us? What do you want to talk about?' The fact that we're now at least a month in advance having discussions, it's a start of more proper dialogue. When Kissinger was doing cross-border discussions, he would be there months in advance to talk to the Chinese before the Nixon visit."

"Pax Silica from the US, launched at the end of last year, was a US-led organization talking about how do we keep US dominance and leadership in AI, which allies are we going to pick to be on our side. It creates a block saying we want to be the winning block. Whereas what they announced was to say, 'Hey, we want to create a global organization, make AI a public good, make it shared, anybody that wants to join can join.' They're going to put out thousands of training centers and compute facilities to members."

"I actually talked to one of the people involved in organizing this and said, 'Wouldn't it be good if you invited the US to join?' They're like, 'Oh no, it would be amazing if the US would join, we would want them to join. But if they join, you can't call it WACO because that's a Chinese-led organization.' They said, 'If you guys are interested, we would be open to changing the name, we would be open to having a truly global organization.'"

"For America to be successful in this industry and take advantage of what we've created is to think about creating something akin to an AI Marshall Plan. The Marshall Plan post-World War II, we spent around $15–18 billion rebuilding much of Europe and some parts of Asia. That created allies, created giant markets for our goods because at that time we had 50% of manufacturing, and it created loyalty for eight decades."

"We need to be thinking more about that type of thing today, with AI data centers, with AI technology. That's what China is actually doing right now. Their WACO, the AI cooperation organization, essentially is their AI Marshall Plan. We should be either doing something like that or working with them to do it together."

"If we did that, we would have a big market to sell our chips when we stop building data centers here, which we probably will at some point when people stop being able to finance it. We're going to need to sell those Nvidia chips in other places. We're going to need to sell services."

"If you look at Amazon and OpenAI revenues when they talk about AI, probably more than half of it comes from two companies. That is not a very diversified revenue base. As more and more of the capabilities move to open source, move to edge computing, the dependency on cloud-based premium services will continue to erode."

"If you look at the recent disclosures of where Anthropic's revenues are actually starting to flatten out, earlier this year they were growing like 10× over just a few months, and now they've essentially flattened out at the kind of AR in the $70 billion range. Although even though their AR is $70 billion, their first two quarters was total less than $20 billion in revenue. But they're committed to hundreds of billions in capex in debt."

"They did the hard thing. Over the last three years, there's been about a 30% deflation in total real estate value in China. And they managed it in a way that it was not a crisis. We need to do a soft landing for these things so that it does not create a crisis."

"When the crisis happens, you need to be willing to take some near-term pain. And they did. They took major hits in their GDP slowdown, they were growing at 8, 9, 10% and now they're growing at 4 or 5% per year GDP, mostly because the real estate sector stopped growing. Global policy folks are talking about this China managed crisis as a hallmark case study on how to do it in the future."

"Home ownership right now is something like 70% in China, and it's probably less than 50% in America."

"At some point we will get to a super intelligence type of a scenario. If and when we do, the concept of nations will probably become a lot less important than they are today. The reason we created nations, it started with city-states, was because we wanted to protect resources. In a world where we achieve ASI and get the kind of abundance that Peter and all of you have been talking about for ages, the need to have separate nations with this type of competition really would not exist. If it still existed, then we would probably have destroyed ourselves."

"The whole evolution of this is going to take probably on the order of decades, maybe by the end of this century we will get to a point. We need to move at a pace that the world can adapt. Trying to move too quickly creates a lot of instability. If you look at the prior industrial revolutions, it was 80, 60, and 40 years respectively in playing out. We're talking about going from where we are today to the singularity in 5 years, that is not a speed that the world can digest."

"We're running the world on an architecture of 17th century nation-states and we're trying to run 21st century applications on that 17th century operating system, and it's simply not going to work. A huge chunk of the issues that we see in the world are that fundamental problem."