Eric Schmidt scores US vs. China

Note: Transcripts are automatically generated from episode audio, and are not fully corrected for spelling, grammar, and formatting.

ERIC SCHMIDT: The China-U.S. competition is the defining competition of your generation, my generation, of our age. AI is the single most important thing that’s going to happen for the rest of our lives. Yes, we’ve got problems. I think it’s important that we continue focusing very hard on this. I want it to be America.

BOB SAFIAN: That’s Eric Schmidt, longtime leader at Google and advisor to the U.S. government, talking to me live at Masters of Scale Summit in San Francisco just a few days ago. Eric Schmidt has singular insights into the critical AI race between the U.S. and China, with high-level contacts both in Silicon Valley and across Chinese government and industry. This episode is part of a special series of high-impact episodes from the Summit stage that we’ll be sharing, so let’s get to it. I’m Bob Safian, and this is Rapid Response.

We’re going to explore one of the highest-stakes topics of leadership and responsibility: the AI race between the U.S. and China. Our guest has been in rooms most of us will never access, with presidents and generals and tech leaders. He was the CEO of Google for a decade. He helped build the U.S. government’s approach to AI and national security and has been vocal about what’s at stake in the tech competition between the U.S. and China. So please welcome investor, advocate, and current CEO of Relativity Space, Eric Schmidt. There he is.

SCHMIDT: Yes. Hi, Bob. Nice to see you.

SAFIAN: Thank you, Eric. Always great. Thanks for doing this.

SCHMIDT: It’s good to be back.

Copy LinkWhy China has nearly caught up in AI

SAFIAN: I want to say that I get whipsawed sometimes by the saber-rattling and jingoism that may be happening around China’s advancement in AI, and I’m glad you can help demystify this for us. The U.S. had a clear first-mover advantage in AI. There are questions about whether that’s going to persist or not. You’ve said the AI gap between the U.S. and China has narrowed to roughly six months. What’s changed most recently?

SCHMIDT: The biggest change in the last year that’s notable for everyone here is that the cost of software has gone from very expensive to almost nothing. When something falls precipitously in price, you get an awful lot more of it. So when I think about the next few years, all I think about is an enormous amount of software: what will it do, who will do it, how will it work together, how will agents work, and so forth. China thinks the same way, and my own view is that the China-U.S. competition is the defining competition of your generation, my generation, of our age. The reason is that both are trying very hard to develop reasoning and intelligence systems that are superhuman. That doesn’t mean like us. It means they have superhuman or superintelligence capabilities that are already beyond human capabilities. When you look at the Waymos running outside, they all have better vision than human drivers, especially because they don’t drink. You sort of get the idea, right?

In the race, I was part of the group that put restrictions on China some years ago on chips, and I assumed they would work for about 10 years. A year or two ago, it looked like China was a year or two behind. Now it looks like they’re six months behind, and there are many studies that show this. How have they done this? After all, didn’t we restrict their access to chips? We did, but they’re so smart they figured out a way to do the kind of algorithms we develop, and use the knowledge we develop, with essentially one-tenth of the energy and, in some factors, less performance. Now, there’s a huge debate in the industry as to whether that gap will get smaller or bigger. But at the moment, it’s fair to say that it’s a two-horse race, and that, to me, is the defining one.

The race for intelligence is literally the race of a millennium. Because once these intelligences exist, our children, grandchildren, great-grandchildren, and so forth will all be affected in really fundamental ways by the things that are occurring right here, right now.

SAFIAN: Some people say that pointing to China’s AI advances is an excuse and a rationale to worry about regulating later. Is there any truth to that? Do you worry about that at all?

SCHMIDT: People make up all sorts of stuff. I’ve been heavily involved with the China-U.S. discussions. Dr. Kissinger and I set them up before he died, in his last visit to see President Xi, which I did with him. The Chinese understand what the race is about. They understand the dangers of AI, and they’re not stopping. So one way to answer your question is to ask: How would a slowdown work?

How many frontier labs are there in America? Five, six, seven? Some number like that. How many are there in China? Five, six, seven. So you have a 12-body problem. On the Chinese side, presumably the government can order them to do what it wants. In America, you’d have to have some kind of voluntary agreement. So now it’s a two-body problem. How are we going to convince the Chinese to slow down, given they already think they’re behind? Not happening. We have to come up with other ways to ensure that this technology is safe while advancing it ahead of them.

Copy LinkWhat the U.S. can learn from China

SAFIAN: I mean, China has, in some ways, a structural advantage because the government can dictate AI as a national strategy and everyone can align behind it. America’s advantage, you’ve argued, is its diversity and decentralization. These are very different approaches. Are there things that China’s AI community has done that American leaders could learn from?

SCHMIDT: I’ll give you a simple framing. China has infinite energy, most of it renewable, something we said we were going to do in our country 10 years ago and failed to do. They don’t have interconnection queue waits that go on seven years. They don’t have the same kind of debates. They’re building the data centers, and they’re building them massively. Furthermore, even with a much more restricted financial market, the Chinese have been able to fund the start-ups. DeepSeek is now seeking funding. It’s illegal for us to invest in DeepSeek. I know because I tried. So trust me, it’s illegal under U.S. law.

DeepSeek just did a huge raise. The funding looks like it’s going to show up. Now, in China, you never know if it’s really the government or really the private sector. My friends in China say it’s the private sector, that the economics are going to work. There’s a complex question involving open source, which we can talk about.

SAFIAN: Yeah.

SCHMIDT: But fundamentally, they’re on their way. They just don’t have the chips. In the United States, we’ve got everything we need. We’ve got the smart people. We’ve got all the controversy. We’ve got incredibly deep funding markets. Where in the world could you get $50 billion on a text message? It just doesn’t exist, except, thank God, that the American financial industry is financing our hobbies here. We’ll see.

What we don’t have is the energy, and we don’t have the data centers. The correct answer is to build your own power systems. One of the great things about America — there’s an old quote from de Tocqueville about America, from 250 years ago: “America is a great country because of the optimism of its people, the abundance of the land, and the absence of a king.” At least we can score the three. We’ve got a lot of land. So, lots of empty land. Put these data centers in the empty land. It’s good for the local economy and so forth. And build the power plants yourself. That’s what we have to do.

Current estimates are that we need somewhere between 60 and 80 gigawatts of power to get through 2030 or 2031. We’ll see if we can get that financed. One gigawatt is about $50 billion in cost. By the way, one gigawatt is about one nuclear power plant. The last time I checked, we weren’t building them at a very high rate.

Copy LinkHow open models reshape the AI race

SAFIAN: You mentioned open source. China’s leadership in open source models, or open-weight models, has fueled the reach of its AI globally. For the business leaders in this room, does it make sense for them to build on Chinese open source models, or not?

SCHMIDT: Let’s define terms a little bit. There’s something called open source. I was part of the beginning of the open source movement back when I was at Berkeley, and I’m very proud of that. It was a long time ago, and my code is still open source and still available if you want to look at very old code. Then there’s something called open weights, where you spend all this money on the training and then take the numbers — of which there are many, trillions and trillions of numbers — and make those available. China’s strategy is largely open source and open weights, and America’s strategy is largely closed source, closed weights.

On the American side, it makes perfect sense. These things cost $100 million to $500 million for one model, right?

SAFIAN: Right. Why just give that away if you’re going to spend that money?

SCHMIDT: Well, this is what the lawyers say.

SAFIAN: Yeah.

SCHMIDT: I spent 50 years arguing with lawyers that you should give things away because it creates a platform. The open source movement basically says: create a platform that people build on top of. The closed source movement, exemplified by Microsoft and basically Bill and Steve way back when, focused on the end user: do a good job, or not, and so forth. That was the test. Open source has largely won in America as the foundation for what we work on, but the large models are largely closed for good reasons. I’m not criticizing them.

There are a couple of examples. One is called Reflection AI, which just released a 500 billion-parameter model. Nemotron, which is the NVIDIA product, has a similar-size model. They’re both very powerful. Google has something called Gemma 4, which is also a very powerful smaller model. You can download these things. So it’s not completely China versus the U.S. China chose open source, open weight because it was the most reasonable strategy for them. They were behind, they needed to get users quickly, and they’re willing to take the thieving risk, if you will. And it looks like it’s working because at least they’re getting funded.

So the contest of open source, open weights versus closed source, closed weights is the defining question. If you talk to the people here in San Francisco who are in open-source or closed-source, closed-weight companies, they’ll all explain to you that their strategy is better because they’ll stay ahead. If you talk to the Chinese people, they’ll say their strategy is better because it’s going to stay ahead.

So we have a real competition here. There are some issues, and I want to highlight one that I believe is, today, unsolvable, and we’ve spent a lot of time working on it. If you think about it, the closed-source weights have the property that they’re controlled by a company. The company is controlled by the laws of that country. The CEO is subject to the law. In China, they’ll just arrest you, whatever. So there’s real control over what those labs are doing.

If you release an open-weights model, an open-source model such as the Chinese models, you put what are called guardrails in it. There’s a technique called obliteration, which allows you to basically remove those restraints, and then you get the raw model. This is a problem because those raw models can be used for cyberattacks. And it looks like — remember what I said at the beginning about the cost of software coming to zero — the rate at which software capabilities are expanding in the next few months is so extraordinary that the technical people here in San Francisco and elsewhere are really worried. This is the source of the “Hey, can we slow down?” kind of optimism, which is very difficult to do.

So I don’t fully understand how we’re going to navigate through this open-source, open-weight stuff when it gets into the hands of terrorists.

You can be sure China does not want this stuff to be misused against itself. After all, they’re a one-party state, and if they’re going to attack the U.S., they won’t do it through that. They’ll do it through their own active cyberwar. The same thing applies to the U.S. versus China, although I’m not suggesting that we should do that. This is a hard problem, and many people believe there will be some kind of cyber event in the next 12 months that will illustrate this point.

SAFIAN: And if, as a business leader, I decide, “Oh, I’m going to try to use that Chinese model because it’s cheaper for me,” is that unpatriotic? Is that shortsighted?

SCHMIDT: Well, I think American CEOs are patriotic to a point.

SAFIAN: Uh-huh.

SCHMIDT: They’re also very focused on money.

SAFIAN: Yes, they are.

SCHMIDT: Welcome to capitalism. There are many examples where people are using small servers with open-source weights to solve some problem because the token cost is lower. The token costs are about three to five times cheaper if you run your own, and so forth. They’re not as functional. So one model for what’s going to happen, which worries me — everything worries me, right? — is that the frontier models will dominate the things that we care about here, these incredible achievements in math and chemistry and so forth. We have Armageddon for mathematicians. All of a sudden, their whole world, which has not changed in 50 years since Godel’s hypothesis, basically, was overthrown in one month, right? They have an existential crisis. That’s going to occur because of the frontier models. For most of us, we don’t need that kind of math to solve the problem of financial accounting. It just isn’t that complicated, trust me. So the idea is that many people will use these cheaper open-weight models for little things. Is that OK? It’s likely to be legal in the U.S. Is it a good idea? It depends on what your view of Chinese threats and espionage threats and so on is. In China, of course, they’ll use the open-weight models.

SAFIAN: Yeah. Coming up next, Eric gives us a point-by-point scorecard on who’s ahead in the AI race, plus what he calls the no-surprise rule. We’ll be right back.

[AD BREAK]

Before the break, Eric Schmidt laid out what’s at stake in the AI competition between the U.S. and China. Now he gives us a scorecard on key elements in the race, talks about what he calls the no-surprise rule, and explains why AI development is and isn’t a zero-sum game. Let’s jump back in.

Copy LinkA scorecard on the AI race between the U.S. & China

So this is complicated. I’m going to try to simplify it, or have you simplify it for us. I’m not sure this is going to work, but Carrie, can we pull up the scorecard? We created a scorecard. We’re going to try to play a game.

SCHMIDT: This is a test that I was not aware of.

SAFIAN: I’m going to give you some factors, and you tell us who’s got the edge: China or the U.S.

SCHMIDT: Okay.

SAFIAN: So model performance?

SCHMIDT: The U.S.

SAFIAN: The U.S. Okay.

SCHMIDT: The U.S., by all the benchmarks, is ahead. One of the ways to understand it is that every lab has the one that they’ve released and then a next one. The indications from the labs are that the next ones are incredibly performant, that we’re still in the scaling law, and there’s amazing things coming. U.S.

SAFIAN: All right. Next one: model cost.

SCHMIDT: China. Because, as I mentioned, the token cost is at least three times cheaper. In many ways, that’s actually because they were forced to be cheaper. We didn’t give them the powerful chips, and so they found algorithmic improvements, supervised fine-tuning, and other kinds of things that are very, very impressive. The interesting thing about their inventions is, the moment the Chinese invent something in open source, we copy it in America. Thank you, China.

SAFIAN: All right. So energy and compute. I assume that’s China, because you’d mentioned that earlier.

SCHMIDT: I’m supposed to answer.

SAFIAN: Oh, sorry. Energy: tell me what the answer is.

SCHMIDT: Okay. You are correct.

SAFIAN: I’m trying to pay attention.

SCHMIDT: No, no.

SAFIAN: Yeah.

SCHMIDT: I mean, China had, I think the number was 560 gigawatts, produced last year. It’s an extraordinary number. This is the benefit of industrial policy in an authoritarian country. If they need energy, they’ve built an enormous amount of it, largely renewable.

SAFIAN: All right, let’s try something like leadership and talent.

SCHMIDT: I’d give that one to the U.S.

SAFIAN: Okay.

SCHMIDT: We are doing some really stupid things in our country. Not that that’s a surprise to this audience. Taking the smartest people and arresting them and kicking them out of the country is a bad strategy. Denying them high-stakes visas to come to the country. My favorite is that we take these graduate students, typically from China but others too, we give them a Ph.D., and then we kick them out. That’s really smart, right? So my favorite example there is the guy who is leading the quantum program for China was such a person. He wanted to stay in America, and we kicked him out. And by the way, do you know who our quantum adversary is? China — and this guy. So QED: stupid.

SAFIAN: All right. I’ve got four more for you. Let’s see: military and defense AI.

SCHMIDT: The Chinese model is opaque to me. I work on military matters. I’ve done a lot of work in the Pentagon and in Ukraine, helping the U.S. in the Iran war and things like that. There are studies that indicate the Chinese models have been used by the Russians, for example, but not in an extensive way. You have to assume that China will, should there be a war — which is to be avoided, trust me — immediately use the dominance they have in consumer drones and other kinds of things, and they’ll move more quickly than the West. That’s a challenge for the West.

SAFIAN: But who gets the point?

SCHMIDT: I’d say it’s either 0-0 or slightly U.S.

SAFIAN: Slightly U.S. Carry, you get to decide whether you want to give that point to the U.S. or not. Safety controls. Oh, look at that.

SCHMIDT: Now that’s a—

SAFIAN: Clearly an American.

SCHMIDT: That is a patriot.

SAFIAN: Safety controls.

SCHMIDT: The U.S. has a better safety control system. There have been a series of meetings, which I’ve been part of, where the various labs have agreed to exchange safety practices and so forth. It’s kind of obvious. I think that, as a general rule, we should expect the government will regulate AI. Not that I’m welcoming it, but AI is so pervasive in society that there are going to be bad outcomes, and those bad outcomes will cause regulation. Safety is the most obvious one — human alignment and so forth. So I’d give that one to the U.S.

SAFIAN: To the U.S., okay. Speed of deployment?

SCHMIDT: I’d say China.

SAFIAN: China.

SCHMIDT: And I’ll tell you why. We are the ones inventing the frontier. They’re the ones doing deployment. I was just there a few months ago, and I’m going there again in the next month. The rate at which AI is being deployed into everything — digital, mechanical, and so forth — is astounding, and they love to show you their demos. There’s no equivalent product in America. For those of you who are building hardware, I encourage you to go to China to look and understand, because they’re beating us in this area. That’s clearly China.

SAFIAN: All right. And global influence.

SCHMIDT: If you measure it in terms of impact over 10 years, it will be China.

SAFIAN: It’ll be China.

SCHMIDT: And the reason, I’m sorry to say, is this: What’s the difference between a free product and an expensive product? More users for the free product, right? The vast majority of countries do not have the economics, do not have the wealth, to really support the kind of costs and token costs and other things that we impose, which I understand. I’m not arguing. It’s pretty clear to me that eventually the majority of the world is going to be using Chinese models. That has huge geopolitical implications. The average geopolitical person I talk to has not figured this out yet. It’s obvious if you think about it: free, cheap, good enough; expensive, powerful, reserved for elites, people who are wealthy, important companies, great research institutions.

SAFIAN: So by our scorecard, at least in this rough way, they’re—

SCHMIDT: But she awarded one of them to the U.S.

Copy LinkWhy the AI race is not a zero-sum contest

SAFIAN: Are we missing anything? Is there any scenario where the U.S. and China both win like that, or is it a zero-sum game?

SCHMIDT: I don’t think AI is a zero-sum game. China is not our enemy. I’ve worked in markets where two companies, although they are competitors, fundamentally create a whole new market. The best possible scenario, which of course is not possible with the current administration, is that you’d have some rational conversation between the U.S. elites and the Chinese elites saying, “How are we going to manage this? Let’s agree to some things.”

For example, a no-surprise rule. In my Pentagon work, I learned that when you launch a missile, you notify all the other countries, and they train their satellites on your missile launch for two reasons. One, they make sure that you did what you said, and second, they get to train their missile-launch watcher. It’s an example of an accommodation to keep everyone safe. You don’t want random missile launches and not know where they’re going. It makes perfect sense. These decisions were made 50 years ago. So there’s a list that I think is quite rational for collaboration in order to achieve human flourishing.

Let me finish by saying that AI is the single most important thing that’s going to happen for the rest of our lives. AI is the underpinning of scientific progress, which is what I’m working on. That scientific progress creates wealth. Wealth creates opportunity, well-being, and benefits for society. The future for our children and grandchildren is very, very bright because of what we’re doing in AI.

Yes, we’ve got problems. I think it’s important that we continue focusing very hard on this. The innovations are coming faster than even I can keep up, and I’ve been doing this a long time. And second, I want it to be America.

Copy LinkWhat leaders should do with AI now

SAFIAN: For the folks in this room, is there some message they should take away? Is there something they should do differently in order to make it be America?

SCHMIDT: Start by using it. I love it when people come up to me and talk to me about AI, and I ask them, “Have you actually been using it?” And they say, “No, no, no.” Okay. Maybe they’re smarter than I am. You’ve got to use this stuff.

What really did me in was when I started having it do long-chain thought, deep reasoning, and I went, “Oh my God, it has arrived.” Now, I don’t know that I understood exactly what it was, but I knew it had arrived.

It changed the way I thought about my life, the way I work, the way I invest, and the way I run companies. It’s very obvious, for example, that the finance function is relatively easy for an AI system. I’ll use an example, since there are lots of businesspeople here: Why don’t you have the AI produce the financial reports, and you and the CFO can have a conversation? Very interesting.

What about manufacturing inventory? What about security? What about cash forecasts? Stuff like that. You sit there and go, “Oh, that’s boring.” No, it’s transformative, because once your mindset is there, you’ll have another idea and another idea and another idea. That’s why this is such a precious moment to be alive.

SAFIAN: Well, Eric, this has been fascinating.

SCHMIDT: Bob, you’re the best. You know that.

SAFIAN: Thank you. Thank you so much.

SCHMIDT: Thank you very much.