Inside the US–China decoupling: What’s really at stake (AI, rare earths, Taiwan, trade)

0:00Hello, everyone, and welcome to this conversation. I'm Azim Azhar, the founder of Exponential View. Please do introduce yourselves and tell us where you're from in the chat. I'm in London. I'm delighted to be with a legend of Substack, Jordan Schneider. Jordan, where are you? I am in New York City, and I'm so glad to be here. We have just partnered up in a new publishing project as well, which I've been finding pretty exciting, right, to have all of these new business partners. Why don't you say something about it for a minute? Sure. So if you care about AI, which if you are watching this right now, I think you make it into that set and you perhaps work at a firm or an organization that would be interested in buying the one AI content bundle to rule them all.

0:52you should head down to readsale.com, which sells group AI subscriptions for teams. We got an incredible lineup, including myself, Azeem, Nathan Lambert, Jasmine, Kevin Hsu, and I think that there's like eight in total of us. Substack doesn't let you bundle for whatever godforsaken reason. So we have taken it upon ourselves to save you the time and allow you to make yourself and your team smarter about what's coming down the pipe with artificial intelligence. It's a great initiative. I'm super happy it started. We will, of course, talk about AI today, but you are best known for China talk. And there is so much going on between the US and China.

1:39And I just want to make sense of it. And hopefully you'll be able to help us do that. It seemed to me that we're watching the US and China do more than traditional trade wars. And we've moved from a phase where the US was trying to contain China, perhaps into something more, a decoupling. And what I want to argue that that decoupling is actually more interesting and perhaps more dangerous than things we've seen in the past. It's not so much separation, but the emergence of these two parallel tech universes. but they remain deeply entangled. And the rub of all of this is the tools we use to manage it, export controls, tariffs, sanctions, well, they're 200 years old.

2:23We're dealing with a 21st century problem, a 21st century economy, and a 21st century technology. So are those traditional containment tools helping or harming, or was I overstating the case in the first place? oh man where to even start with that I'll do a little personal history I guess because I'm introducing myself to you and the in the audience as well so I moved to China in 2017 and I was like you know played video games like built PCs in like middle school but my I was a history major my profession wasn't in in working in our understanding technology it was a policy analyst. And the reason I reoriented my professional life around technology was because I hit the thing that you just mentioned, Azeem, this idea or this reality that living in China, you live in a similar but different technological ecosystem where the apps that you interact with on your phone are just different than the ones that the rest of the world uses.

3:27And that kind of trickles down into like literal like UI changes of how WeChat works versus how Facebook works at the time means you relate to your friends in a different way and just kind of understanding like how fundamental these products are to our lives nowadays. And it didn't sink into me until I lived in that separate ecosystem and everything that I kind of assumed was like baked in or took for granted actually surfaced to me as like actual product managers or decisions that policymakers made to create a sort of like different way of being almost in mainland China versus the rest of the world.

4:09So that's like step one. Zooming way out from where did this decoupling come from and where is it going? It is really messy, man, because there are not easy takeaways and there are not clean lines that I think either country really wants to draw. So from 1949 to 1973, there was basically no technological or commercial contact between the U.S. and China. The U.S. didn't even recognize China or mainland China until that time. You know, for for American diplomatic purposes, Chiang Kai-shek and the KMT in Taiwan were like the rightful owners of mainland China. So that's where my daughter has just flown off to this today, in fact, to a school trip to Taiwan for 10 days.

5:03Oh, amazing. I hope she has a great time. So, you know, 1973, we have the Sino-Soviet split where Mao decides that he's sick of Khrushchev's BS and sees an advantage in becoming better friends with America. And all of a sudden we have this, I would say, 50 year arc of actual pretty dramatic U.S. and Chinese economic, technological and even up until the 1990s military entanglement, where America had this idea that during the Cold War that supporting Chinese development would help gain an edge over the Soviet Union. And then afterwards, that morphed into, oh, this is an important emerging economy.

5:46There's a lot of economic opportunities sort of couched in this like liberalization thesis that the more we engage with China, the more kind of open, democratic, friendly it will feel to our small liberal sensibilities. That may have ended up being the case had 1989 turned out differently. But that's not the world we're in. And I would put our third periodization starting around 2013, 2014, where the Obama administration, followed by Trump one, really started to reframe China as more of an enemy or I would say a competitor. And in that, we've had a lot of different permutations of the thing we're seeing today from Trump one's trade war to covid to Biden putting on all these export controls and banning stuff like Chinese electric vehicles to now Trump to where we're now like, you know, six months into a saga of we had Liberation Day and now we're escalating on this and that front.

6:49But the fact of the matter is it's a big mess because these are the two largest economies in the world and they're not separate. And there's a lot of sort of pieces of both systems that see advantages in keeping connections. Right. And we've seen that a little bit with these new rare earth tariffs or controls, rather, that were announced a couple of weeks ago, where it's the heavy rare earths that you can find in missile control systems that have fallen under this new control. But the lighter rare earths that live in things that the Chinese export, like wind turbines, don't fall into that control.

7:28There's some pragmatism that's going on. But I guess the thing that I thought felt like it might look different is a sense that every attempt at control creates an unintended resilience, a greater capability to be independent. I think of the quote, it's a misquote of Fontaine, but it's in Kung Fu Panda. And Master Oogway says to Master Shifu, one often meets one's fate on the path one takes to avoid it. And if the path you want to take is to have more control in this case, what you may end up doing is making the other side much, much more resilient. And we've known from China's various five-year plans that they wanted greater independence in semiconductors.

8:16They wanted greater independence in a range of other products. They do lean more heavily on open source for their software infrastructure than we do in the West. If there is any coherence to this big mess that you described, does it actually just lead to a control illusion rather than a real sense of control? And I feel, I mean, one of the reasons I wanted to talk to you today was that it starts with those October the 9th rare earth rules. You know, what breaks first in that environment? It's a fascinating question. I'm going to start with a history analogy. You know, there was a big autarky fetishization in the late 20s and 1930s as well.

9:02And you had this idea of sort of Nazi Germany and Imperial Japan that, you know, the equivalent of chips or rare earths in that age was, you know, stuff like access to oil or other or other minerals. or there are all these like, you know, ball bearings, all these like critical inputs that if you didn't have control in a literal colony of yours, you didn't have factories in your country, then you were sort of subject to the whims of another country. And if you think the world is fine and we're going to keep trading, then like, that's great. That's like, you know, this is like Ricardian trade and like we're all richer for it.

9:43If you come at it from a perspective that they did in the in the 1920s and 30s with, you know, Mussolini and Stalin and Hirohito that like this is a vulnerability. We are bound to get screwed. So we need to act first and, you know, get what's ours because it's all going to, you know, come to shit sooner or later. Then you end up in these very dangerous dynamics. And look, I don't think that we are like the the fetishization of doing everything in a way which is entirely like on your land and controlled by you. And, you know, you can feed your country with like the the farms that are in your landmass.

10:30Like we're not entirely there yet. But it has been fascinating that I think China was very early to this. you know there is like a deep aspect of like PRC history where they feel like they've gotten screwed and so even before the October like far before the October 7th controls like when she comes to like you can find dung quote you can find Mao quotes about this you can find dung quotes about this one of the first speeches she gives is all talking about sort of self-reliance and not being vulnerable to having the the other world like other countries like bully us around be that sort of economically, culturally, what have you.

11:10Let me just dig into this question, which is, we have got these leaders. And, you know, when you talk about self-reliance and you bring up these examples from the 1920s, we can also think about Juche, right? The North Korean theory of self-reliance. But what really are the politics of the leaders? Because it does feel that China and Chinese communism looks very different today to how it looked in the era of the bicycle or just at the moment that Nixon started that rapprochement. I mean, there was something changed. So what are their politics and their influences around this particular question?

11:47Yeah. See, this is the problem with self-reliance is it is expensive. It is uneconomical to do things because if it wasn't, then your country would already have it, right? So after the disaster of Mao, lots of Chinese leaders made what I think is like the correct strategic decision that, yeah, we are going to need to get more dependent on exports, to get more dependent on sort of foreign technology. Because if we're not like, yeah, you can be self-reliant like North Korea and you can be really, really poor. That is basically the experience that you had in Mao era China. We're like, yeah, they weren't importing a ton of technology because they were too worried about starving every seven years because of like, you know, terrible.

12:31The sparrows are the locusts and the four plagues and these types of things. Exactly. What I think is kind of really interesting is over the past five years, you've seen this mindset start to come home to roost in the West, right? Like you have EU leaders talking about strategic autonomy. You have Trump talking about reindustrialization. and a lot of people I think over the past few years have made the argument oh like America cutting off chips to China that's only going to make them like industrialize you know build build chips faster like well we're about to run that experiment in the opposite direction because I think the rare earth stuff has gotten a lot of people very focused and concerned in the west and then we're going to have this very fascinating sort of like system competition where like is the fact that America still like, I guess kind of a little bit has friends around the world who are also worried about Chinese rare earth export controls?

13:33Like will that cut against the advantages that the Chinese system has of being able to keep their eye on the ball for a longer period of time and being able to sort of boss companies and countries around? Like that's a whole different ballgame. Well, I think there's something else that has changed in the international environment over the last 10 or 15 years. which has been the explicit weaponization of dependency. In this Ricardian trade world, we always have dependencies on each other. Famously, the Airbus passenger jet is manufactured in 100 countries and the wings crossed borders 20 times.

14:10And you do that because of the gains of trade. You do that because of the gains of specialization. But underpinning trade is that sense of dependency. and countries have generally been a little bit wary of doing that. I suspect the OPEC cartel is a good example where the dependency on oil was used as a manifest as a tool. But in the past five or six years, that weaponization has become increasingly manifest and it's becoming perhaps part of the geopolitical landscape. So I'm just curious about whether you see that as a phase change as well? Or is it a rerun of the tape that you so eloquently described in the 20s and 30s?

14:51Well, rerunning the tape of the 20s and 30s, that would be a real phase change, right? I mean, I think it's somewhere in between, right? Because I agree with the thesis. So during the Cold War, you did not have these dynamics between the U.S. and any kind of non-Warsaw Pact country, Right. You know, and then we had this really nice, you know, 25 years from 1990 to, you know, 2013, 2014 or so, where like as long as you weren't Iran or North Korea, you kind of got to play the same game. Everyone was kind of on the same side. And the benefits to the Ricardian version of international commerce outweighed any sort of latent concerns that weird history nerds had about the return to a world where you had great powers kind of competing much more directly than they did in the past.

15:42But do you think that what we see in this specifically between the US and China today has a grand plan to it or is it a bit incoherent? I mean, I look at things like 100 % tariffs, but carve outs for NVIDIA's H20 chips, you know, with the Chinese response to their domestic industry saying you've got to use domestic chips. It doesn't look like either side is fully proactive and escalating in a consistent way. And it certainly looks from the U.S. side that there's a little bit of shooting from the hip. Look, like we now have policy that's basically driven by three people. And that is new. I'm trying to work out which three you're talking about.

16:26I think we got like Trump, we got Lutnick, we got Besant. And then, you know, give J.D. like 25 percent. We'll see how many group chats he ends up being on. But like there was a mode of policymaking that I think America really recoiled from. FDR, he was president for three and a half terms, basically shot from the hip. And it was what he said goes and all of his buds just like listen to what he did. And there was an understanding at the start of the Truman administration that there needed to be institutionalization, that you needed sort of more people thinking about these things, more planning.

17:03The world's a really complicated place. We got to trade off all these different dynamics and interests and think about the military and the economic and the diplomatic and the social components so that you need bureaucracies to like move paper, make sure all these opinions are heard such that when you start making these very consequential decisions, it's not just like, you know, FDR vibing it out. And we have now returned to a president vibing it out with the world immensely more complicated than it even was in 1946, right? So yeah, it's totally incoherent. It is not just that Trump is a very unique historical figure, which is making it incoherent.

17:45I think you could also look at the Biden administration, which was a lot more process forward as having a sort of Janus-faced, like, confused version. You know, they literally call it, like, we're going to compete and collaborate. It's like talking about AI, like, we got to, like, use the benefits and mitigate the risks. It's like, okay, well, you know, you got to square that circle somehow. And if you stop trade with China tomorrow, the U.S. economy stops working. Like, GDP just goes away, right? So maybe you don't stop trade with China, but maybe it is that any kind of systems change is necessarily messy, feels incoherent at the tactical level, because what you need to do is generate momentum, generate force in order to change the system.

18:31The White House, the administration has been reasonably clear about the gestalt that they want, which is sort of make America great again, have fewer dependencies, have a greater, more transactional control over relationships with other nations rather than alliances that are written on paper. So perhaps tactically that will always need to have that sense of frivolity and friction rather than people in suits walking down a checklist. I wonder whether what's really going on here is system building, that we tend to talk about the components or the minerals. We tend to talk about the bandwidth of the memory in the chips or whether the battery can come out or whether a particular rare earth is allowed.

19:22But perhaps it's as much about creating these innovation and technology systems that are not dependent at a fundamental and strategic level on your great rival, right? You can compete in other areas, but you want to be soup to nuts in that area. And if you want to have systems change, It's going to be messy, especially given that there were all these interrelationships that you're working from. Yeah. I mean, look, Xi has this great line. I think he said it in a 2020 speech where he is like, we want to not be dependent on other countries and we want other countries to be dependent on us. I think, you know, when you see Lutnick talking about how he wants to get China addicted to age 20s, like it's the same dynamic playing out on both sides.

20:10And yeah, that is super messy. And that means that sort of there are two system which are really wrestling with each other. Like, I think that, look, I honestly don't feel like the Trump administration deserves to earn as much sort of like cognitive coherence, as you may have put it so eloquently, Azeem. But there is, I think, a new instinct to the melange of like things you care about of not wanting to be reliant on China because you can't trust them because they're scary. And oh, by the way, they can shut down Ford plants in a week if they feel like it. And vice versa, right? Sure. That understanding, I think, has been in the Chinese system for much longer.

20:56And there was an understanding in the 70s, 80s, 90s, 2000s, even early 2010s, that it just like we're just going to have to live with it because we're too poor to kind of break out of the interdependence. But now that China is richer and I think maybe also Xi is more ambitious, he is more excited to develop a higher degree of resilience as well as to weaponize the interdependence that China has vis-a-vis the rest of the world. There is this word that goes around, which is sovereignty. And it's appeared in Google Trends far more in the last two years than it did in the previous 50. And of course, there's a very legalistic, formal term meaning for the word sovereignty.

21:41But there's also vibe-based sovereignty, which kind of means we can sort of make our own decisions. And of course, lots of countries are talking about sovereign AI and sovereign energy and sovereign this and sovereign that. In truth, there are only two countries in the world that have the depth and the capacity, the dynamism to achieve that vibe-based sovereignty in any real way. But even they, it seems, are quite dependent on each other. Totally. Okay, so now we get to talk about chips and AI. You saw the EU this week talking about semiconductor sovereignty. Like, I'm sorry no like it's not happening are you really going to try to spend like what 50 billion dollars 75 billion dollars to build a competitor to Intel or TSMC or Samsung it is an interesting fact to me that that is even a a kind of like a worry that I think Europe you know maybe 10 years ago would not feel the fact that you even have plans about that today is interesting but like The reality is like, no, you're still going to get all of your advanced chips for decades to come from Asia or from the U.S.

22:49Absolutely. So let's talk about AI then, because, of course, AI is running on the chips. Let's give the received wisdom right now. So the received wisdom is that the U.S. leads in the frontier, that the U.S. foundation model labs are by and large chasing after the machine god that is AGI or ASI, and that the US is highly constrained by some of its physical infrastructure, so the power systems and so on. And on the other hand, China is doing really well with open source models. It has abundant power capacity and the open source models are getting better and better. But critically, the Chinese state civil enterprise fusion is all about deployment and that these are two entirely different approaches to AI.

23:38I think that's a received wisdom. Is that right? Is that wrong? Can I stay on the sort of like, can anyone be sovereign in AI for a second before we get to the tail of the tape stuff? Because I think it's really interesting, like at what level different countries want to be sovereign. Because as you said, like Azeem, there's only two countries in the world that can go from sort of chips to cloud, to labs, to, you know, to models, to like companies and have that whole AI stack be something that they can like roughly indigenize. I mean, we'll we'll round out the fact that like China still needs a lot of sort of foreign SME and the U.S.

24:19is still buying its chips primarily from Taiwan. But like every other country in the world is going to have to make a decision or try to plead their case to China, you know, to Huawei or to the White House and NVIDIA that like, oh, please, like, I would like to buy the chips, please. And then I'm going to have a local kind of like cloud provider. And then I'm going to fund an AI lab to train a model and build companies on that. Or, you know, other companies will say, like, I just want some data centers here. If they're AWS, that's fine. Or you say, I don't really care. Like, just give me access to open AI, you know, Claude and OpenAI, and that'll be that'll be it.

24:58Right. And coming back to my first discussion for the sort of Internet boom of the 2000s and 2010s, you basically had, you know, a little bit of this like data localization stuff. But for all intents and purposes, it was American companies on American chips by American cloud providers that like run ran the world ex-China. So, you know, Google and Facebook, Uber, Amazon, whatever. It was like, it was just like the U.S. But, you know, what started to change, I think, was in the mid-2010s when the strategic importance and national importance of data, particularly data-based citizens, started to emerge amongst policymakers.

25:39And you get to the stage that by 2020, more than 100 countries have got their own data privacy frameworks. We always think about GDPR, but, you know, Kenya had one and Brazil had one. they were expressing some desire of classic Westphalian sovereignty in this new domain against the backdrop of essentially having been happy for the Americans to do the running for 30 years. Yeah. And also you had this time period where no one was pushing back all that much about the sort of censorship decisions of Twitter and Facebook. But like that started to change once governments all around the world realized how important this was.

26:20And then, you know, you have the Indian government like throwing Google employees in jail until they like take links on YouTube down because they're worried about, you know, X, Y or Z things. So like, I think it is inevitable that countries will end up wanting to put their spin on this. But sort of what is what is also interesting to me is like right now, it's still early enough that the incumbents aren't so entrenched that you can't imagine a new cloud provider emerging or imagine a new model developer existing or imagine, you know, a new company that's being built on top of these models. Right.

26:58And so it seems much more fluid than like, oh, I guess we're stuck with Facebook and YouTube. We're going to have to work these American companies like now we can still maybe try to squeeze the Americans into into building sort of higher value stuff. I just returned from Abu Dhabi this morning. And of course, the UAE has gone all in on AI as a technology for the nation, as a technology for its industry and economy. It's building huge data centers. But critically, to your point about new model providers, they have released in the last two or three weeks, a really performant open source reasoning model that is extremely light, measured by billions of parameters.

27:42And, you know, it is interesting to see that, to your point, nothing's yet locked down. There are certain things that feel rather more locked down. So NVIDIA's chips and TSMC's fabrication and ASML's lasers. But above that layer, there is still quite a lot to play for and quite a lot of space then for countries to think about what does sovereignty mean in this environment. We'll have to see. I don't know. it's kind of fun and also depressing if you're like working the Abu Dhabi, like, I don't know, tech policy department or whatever, because on the one hand, yeah, there's room to play where there really wasn't if you're talking about this from 2018.

28:21But it's also like you're competing against these companies which are valued in the trillions of dollars at this point. But, you know, we don't know. I mean, I think this goes back to the question I asked you about AI that you so neatly dodged, which is that we don't know how the industry structure will play out. We're full of surprises. The internet had provided a ton of surprises. Remember, it was always going to be about decentralization and it ended up being incredibly centralized. It was going to be about the hour of spring and people speaking with one voice and it ended up being anything but that.

28:56So one of the things that I've got some experience and some scar tissue from is trying to understand exactly the path that a technology takes from here to there. And as I said, you know, China seems to have a different approach to AI, which seems to focus on deployment rather than really, really breaking the frontier. I mean, is that a fair way of characterizing it? I'm not sure. I think, like, it's harder to push the frontier when you don't have as many chips. So if you don't have as many chips, you'll do some other stuff. I mean, the amount of like hand wringing I've read from Chinese CEOs saying we need to do more like basic research.

29:39Like we can't just let the let the foreigners like steal a march on us. I mean, there is angst about this fact that all of the breakthroughs seem to come out of California as opposed to, you know, Zhejiang province. That I think the sort of tech leaders as well as, you know, Xi himself and the heads of sort of Chinese science bureaucracies aren't necessarily comfortable with. I mean, I think there are some unique dynamics in China in terms of like the fact that it is a protected market, right? Like OpenAI and Anthropic, their models, like you cannot have like, well, I mean, people are like fucking around and trying to get around it.