Invest Like the Best with Patrick O'Shaughnessy
· Dan Wang
I thought it was really strange that Xi Jinping is trying to achieve something he calls “the Great Rejuvenation of the Chinese people,” and there are a lot of young people saying, “No thanks, we’ll give that a pass.” There are also a lot of millionaires—by one count, 14,000 millionaires departed from China in 2023—to move to Singapore, Japan, the U.K., and the U.S., in part because the Communist Party has smashed many of their businesses. There are also many people who are not so creative or not necessarily very wealthy who decided to hop on a plane to Ecuador, where they don’t need a visa, and walk across the Darién Gap into the United States. At its peak in 2024, there were several months in which U.S. CBP was apprehending something like 30 to 40,000 Chinese nationals at the Texas border. That, to me, is really striking—that so many people are willing to do this dangerous trek essentially to escape from China. I think that China will not become anything like a cultural superpower, in part because the engineers are so censorious. They’re so thin-skinned, and they censor everything they can’t understand—which is a lot. If you take a look at many aspects of cultural production—books, novels, films—in China, that has become much more constricted over the past 12 years of Xi’s rule. I also think China will not become a big financial superpower, in part because the control tendencies of the engineers lead them to impose a lot of capital controls, which makes it much more difficult for foreigners to want to hold RMB. Last I checked, RMB still made up only about 4% of global trade volumes. I want to make this narrow case that the engineering state is really good at building stuff. They have the manufacturing workforce, the entrepreneurial enthusiasm, and the government support to keep getting better and better at advanced manufacturing industries. But even if China pisses off a lot of its own people, its neighbors, and the broader set of nations, the very fact that it can make manufacturing products better and better is a pretty significant threat to the United States. There’s a chance, I think, that they will do really well in artificial intelligence as well. But even if they just get much better at making vehicles, I don’t want to see the U.S. further deindustrialized. Right now, there are 12 million manufacturing workers in the U.S.
· Dan Wang, Patrick O'Shaughnessy
I don’t want to get into a scenario where a decade from now we have only 6 million—that’s going to hurt the economy and introduce greater political dysfunction. Even if the engineering state does very well in the short run—for James C. Scott reasons—and even if in the longer run it does not do so well, in part because people want to leave and in part because of demographic problems, the future is made up of a series of short runs. If China is really able to do super well in the next five to fifteen years, it might be enough to subdue America in pretty substantial ways later on. [00:00:00] S0: That’s an incredible answer, Dan. I’m so glad I asked. One of the things that jumps out is the sensitivity of the Chinese model to the person in charge and the people aligned with him. If we had had this conversation when Deng Xiaoping was running China—or was so important in China—maybe the trajectory we would have projected would be very different than it’s ended up being. I’m curious what you think about the three, four, or five people that have been the most important since Mao, and how they’ve changed the trajectory of China. How likely is it that the next person is more like Deng Xiaoping, with more liberal or pluralist leanings, or not? It’s such an interesting aspect of this whole trajectory.
· Dan Wang
[00:00:00] S1: Yeah, I would bet not. Because first, I agree with you that the Chinese system is really under the sway of a single leader—and that is almost by design. I mean, authority in China is meant to vest in the Politburo Standing Committee, which is made up of seven to nine people. But it is also possible, as we’ve seen in the case of Xi Jinping, to really dominate the entire system. And what Xi says more or less goes. The part that I would hesitate to fully imagine when Xi Jinping goes away—that the system will be different—is that even if we have a Deng Xiaoping-like figure, China might still have a lot of dysfunctions. I would say that Deng Xiaoping introduced a lot of the dysfunctions that carry over to Xi Jinping. So I think Deng Xiaoping, broadly speaking, ought to be best known for being the architect of reform and opening, which liberalized the economy. And that is a tremendous good—perhaps the most tremendous economic good ever in the history of humanity—because about 1.4 billion people were lifted out of poverty. A few hundred million of those are living very comfortable lives, and a few million are living lives on par with some of the best in the United States. That is an extraordinary achievement. On the other hand, Deng Xiaoping was also very autocratic himself. In the words of one historian who was quite senior in the Party, he was “half a Mao.” He was someone who was really, really ruthless toward his political enemies. He was never quite able to figure out succession planning. He fired two of his hand-picked successors and almost fired his third one, Jiang Zemin, who ended up running the country. He was also really famous for ordering tanks into Beijing. For a while throughout the 1990s, people labeled him, as well as a few other Communist Party members, “the butchers of Beijing,” because they ordered in the tanks during Tiananmen. Deng Xiaoping did not permit substantial political reform. He promoted tremendous economic reform, but he did not figure out the problem of succession. We can maybe situate Xi Jinping as a direct analog of Deng Xiaoping because he is also grappling with questions of succession—what comes after him. He has been pretty ruthless toward people he has purged. This is where I’m skeptical that there are deep pluralist genes in China.
· Dan Wang, Patrick O'Shaughnessy
I think that China, throughout its imperial history, has substantially lacked a liberal tradition—a liberal tradition meaning one that is interested in restraining the power of the emperor and promoting the power of the individual, the family, or the corporation. So I think the future of China will look politically much more similar to today than many expect. [00:00:00] S0: I've been a little surprised since the original DeepSeek moment—which, for at least a week, really shook the U.S. in terms of the progress that had been made in AI in China—that I haven't heard more about it since. So I'm curious: your impression of AI in China, the country's and the Party's orientation toward it, the state of the technology relative to the leading U.S. research labs, et cetera. What does AI in China feel like to you today?
· Dan Wang
[00:00:00] S1: Yeah, I think AI, for me, what is most important is going to be in the next five years. Today, I think that the U.S. still has a decisive lead on most aspects of artificial intelligence, mostly because it has the compute resources here. It has the leading Nvidia chips, and China does not. But if we are going to have much more AI in our lives, I think the competition is much more of a toss-up again, looking forward to the next five years—which is a lifetime by the standards of a lot of AI, maybe a shorter time span for the rest of us. And here is where I think China actually does have a lot of advantages in AI relative to the U.S. If we have much broader AI in our lives, we’re going to be using AI for all of our queries. We’re also going to demand much more power than we currently have, and the U.S. has not been doing very well on adding new power to the grid. As I said, it is only going to be adding something like 50 gigawatts of solar this year, relative to 500 in China. No new nuclear plants are under construction. Maybe that will change because the Trump administration is more friendly toward nuclear, but the U.S. has simply not added much by way of power. China, I think, as a civilization—as the engineering-state civilization—is just very interested in making sure that no heavy industry ever goes hungry for power. The scale of electrical production in China is vastly outpacing the U.S. The question of talent is always very fluid. Right now, how many researchers really matter for artificial intelligence? I don't know if the figure is more like 10,000 or more like 1,000. Within the 1,000, it seems like China has a lot of really good talent, which was represented by DeepSeek earlier this year, showing that it’s not only Silicon Valley that can produce really good reasoning models. At least from what we can tell in some of the publicly listed hires for Meta’s superintelligence, a substantial number of these really elite engineers—I think there are 11 who are publicly disclosed—something like seven of them had gone to Chinese universities, and I believe they’re all Chinese nationals.
· Dan Wang
There’s some sense of nervousness among some folks in Silicon Valley that these elite engineers might pack up their bags and decide to repatriate back to China for a variety of reasons—either because ByteDance is offering a much better comparable pay package, and they can live a much more comfortable life in Beijing or Shanghai; or maybe they just want the better noodles, which California doesn’t really have. There’s also some sense in which they feel anti-Asian hostility in the U.S. They fear the rhetoric of the Trump administration. Some of my Asian American friends here are nervous that the Trump administration is going to turn super populist and drive out all of the Indian engineers as well as the Chinese engineers. So within this talent pool, if a substantial number of them decide to repatriate, then it’s a little tough to say exactly who’s ahead on talent. But it’s possible that the Chinese will have some sort of lead here. Then there’s also a question of if AI gets much better—what will the Americans use it for? What will the Chinese use it for? America is much more of a services-driven economy. Maybe we have all of the data to automate our healthcare sector as well as our consulting sector, and the Chinese are going to have the data to automate their manufacturing sector. To be very glib here, I’ll say that maybe America gets a much better McKinsey, and China gets a much better Foxconn. They’re going to use this much better AI Foxconn to produce a lot more drones, munitions, and ships as well. The U.S. simply doesn’t have the training data or the process knowledge in place to get much better at manufacturing. So maybe right now we haven’t seen another big DeepSeek moment, but I think the future is very much up for grabs. China right now has really good talent and really good reasoning models. It is now compute-constrained, but it seems like Donald Trump wants to make a deal and impose an export tax of 15% on Nvidia, while giving a lot more compute to China. I think the future is going to be a really tight race between who can get to AGI and who can deploy AGI in a much better way.