Odd Lots - The US and China Are in an All Out Race For AI Domination

Episode Date: July 22, 2024

There are several sources of tension right now between the US and China. Pure trade anxiety is a big one, with the US having imposed tariffs on Chinese electric vehicles, solar panels and other important industrial components. Then, of course, there are direct geopolitical concerns, with fears over a possible move by Beijing against Taiwan. And then there's artificial intelligence, which countries all around the world see as a crucial geopolitical asset, with the potential to transform economies and militaries if and when it reaches sufficient strength and power. And so, American-based labs are going toe-to-toe with Chinese ones, investing enormous sums of money to get ahead and stay ahead in this race. But what is this actually all about? What kind of advantage does America have in the AI race and can it be maintained? How might it change under another term of President Trump? On this episode, we speak with Jordan Schneider of the ChinaTalk newsletter and podcast, as well as Kevin Xu of the Interconnected newsletter, to discuss the state of play.See omnystudio.com/listener for privacy information.

The news doesn't stop on the weekends. Context changes constantly. And now Bloomberg is the place to stay on top of it all. Hi, I'm David Gurra. Join us every Saturday and Sunday for the new Bloomberg this weekend. I'm Christina Rafini. We'll bring you the latest headlines, in-depth analysis, and big interviews. All the stories that hit home on your days off.

And I'm Lisa Mateo. Watch and listen to Bloomberg this weekend for thoughtful, enlightening conversations about business, lifestyle, people, and culture. On Saturday mornings, we put the past week's, events into context, examining what happened in the markets and the world. That on Sundays we speak with journalists, columnists, and key political figures to prepare you for the week ahead. Join us as soon as you wake up and bring us with you wherever your weekend plans take you. Watch us on Bloomberg Television.

Listen on Bloomberg Radio, stream the show live on the Bloomberg business app, or listen to the podcast. That's Bloomberg this weekend. Saturdays and Sundays starting at 7 a.m. Eastern. Make us part of your weekend routine on Bloomberg Television, and wherever you get your podcasts. Bloomberg Audio Studios. Podcasts, Radio News.

Hello and welcome to another episode of the Odd Lots podcast. I'm Joe Wisenthall. And I'm Tracy Allaway. Tracy, obviously, there's a lot of, there's an infinite number of angles to artificial intelligence and what's going on in the industry. But one thing that I find, like, striking about this that feels very different from other stories in tech.

And I don't really know what it's about is why you hear it talked about in sort of geopolitical, strategic terms and why countries, whether it's the U.S., China, but also Saudi Arabia, Europe, really feel like they need to have like their own domestic model or champion. Like it's almost like it's like oil or something where countries have at least been convinced that it's very important for them to have some sort of homegrown AI. Yeah, you're right. I hadn't really thought about it, but that's absolutely. Absolutely true. People talk about it almost as a strategic resource at this point and sort of in existential terms. And you never saw anyone talking about like the need for a national SaaS strategy or something like that. Right. That's exactly what I was thinking. Right. Like you didn't know what it's like, oh, well, we have to have our own slack or we have to have our own sales force. We have to have our own British sales force and a Saudi Arabia sales force. But there's something about AI models. And like, you know, it's sort of about the chip specifically and, you know, everyone wants. the chips, right? But there also seems to be something where like countries want to like really

develop their own models and have their own open AI or whatever it is. Yeah, I also wonder how much of it has to do with the data and the idea that like the data is what you're training the model on. And so there's like privacy concerns isn't the right word, but you know, countries want to have some control over their own data. They don't want full transparency. So I wonder how much that feeds into it. But you're right. We have gotten used to talking about the AI arms race in one way or another. And we should dig into why that is. Like, why is it such a competitive piece of technology? Totally. And I would say, you know, there's two other things here, which is that obviously the ultimate in everything these days, a lot of it is U.S. China and how U.S. models and

U.S. companies compared to Chinese models and Chinese companies, which I don't think many Americans have much visibility on, or at least I certainly don't. I don't really know what China's AI strategy is. And then there is, of course, existing policy and how that might change under a Trump administration and what that might do, both to you as China relations in general. And I don't know if that's going to, you know, some sort of shift in how Trump might think about AI versus a Biden administration. It is true that like certainly the semiconductor portion of AI is something that has been competitive for a while now. And we've seen that in both the Trump administration and continued under. Biden. So there is that kind of commonality running through it. But I am very interested in getting

more of a handle on exactly what China is doing in this space and what it sees potentially as like an actual threat here. Totally. Well, I'm very excited to say that we have two perfect guests who really know all the dimensions of this topic extremely well. We're going to be speaking with Jordan Schneider. He is the founder of the China Talk podcast and newsletter. One of my favorite sources for understanding what's going on in China. We're also going to be speaking with Kevin Chu. He is the author of the Interconnected Newsletter and also a hedge fund manager at Interconnected Capital. Writes fantastic stuff, like sort of really digging into some of the technical and tech

sides of all this. So hopefully we will walk out of here with a little more understanding. Jordan and Kevin, thank you so much for coming on Oddloves. Thank you so much for having us. First time, long time. What a treat. I love when people say that. I love hearing.

It reminds me of like listening to like sports talk radio. as a kid. I'll just start with the question. What is it about AI? And is it inherent or is it sort of the industry's marketing job? But what is it about AI that unlike a Tracy mentioned, you know, different from like CRM software, social networking or whatever, that countries really seem to prize having some aspect of it be homegrown? Yeah, I couldn't help but chuckle a little bit when you guys are doing your intro about the national SaaS or what is your national CRM since I am software investor by day. You know, it's really interesting. If we think about how AI, when we talk about it generally has been formulated, you know, we're really talking about generative AI, right?

Large language models in particular. There are a bunch of other AI machine learning use cases that we're going to, you know, just stick to the side for now. There's something very deeply cultural when you start to model the entire Internet's information when it comes to large language model, right? It's essentially a three-dimensional space where every single two, token or every single word has a line that's drawn from one word to the other that kind of formulates knowledge or formulates, you know, truth in a lot of ways. And I would say so far, given all the dominant products, chat GPT, Clyde, Gemini, etc. They've been modeling the internet in general, which has a lot of content from the American internet, if you want

to think about that, or, you know, Western European internet. And there are other versions of the internet in Japan and China in particular where the ecosystem is such a world. wall garden, that the way knowledge and culture and thoughts and, you know, history is being expressed is very different. And I think that is what triggered a lot of country to come to an understanding that this AI thing isn't just a blown up spreadsheet with good UI that you can track, you know, pipeline or, you know, have a bunch of messages going around to help your workplace go a little faster. It is a codification of your people and your history. Collective truth and culture. And I will say collectively your public discourse, maybe even an element of your private discourse

that's being published on the internet is being codified in a particular large language model. Just to add to that and bring up the U.S. China context. So, you know, if we take as a premise that the U.S. and China are in a strategic competition, what nations do when they try to, when they compete is they try to maximize national power. What is national power composed of, how big your economy is, and how powerful your military is. Now, you know, there's still a lot of debate to be about just how important AI might be to the future of militaries and the future of productivity growth. But it is the technology today that has the most potential to really change long-term military and economic trajectories. And I think we've seen over the past few years the realization maybe starting

with the October 7th, 2022 export controls, that having a completely open exchange between the U.S. and China is too much of a risk for American policymakers to stomach. And sort of in response to that, simultaneously and particularly in response to the October 7th export controls, you've seen China really aggressively start to invest and support both their semiconductor as well as a broader AI ecosystems. Wait, so I want to press on this point. Kevin also brought up, but the idea of like AI as a means of like cementing cultural and informational control, did anyone play around with the, what was it called like the study she thought app?

Oh, I never played with that. Yeah, it was like the open AI app that was trained on Xi Jinping thought. Here's the funny thing is that app is like actually kind of vaporware. We tried to find it on China Talk, which you guys. should all subscribe to and couldn't. And I think this is one of the really interesting pieces of the Chinese government and how they're specifically playing on the sort of AI model side of things. If you look at a lot of the Chinese government sponsored projects so far, particularly on the model generation or like national data set side, and you try to get to ground truth on them,

there's really not a lot of there there. But that's not to say that the Chinese sort of VC-backed model ecosystem isn't the real deal. We've seen companies raise multiple billions of dollars and be able to publish models, which are maybe six months behind the leading edge when you're looking at Open AI or Anthropic. And the sort of delta between the government back stuff that can't pay for the best engineers or what have you and these, you know, dripoo AI moonshot, Ernie at Bidu, those folks are really running just as hard as what you're seeing in the West. Let me ask a related question. But in terms of Chinese AI models, what do we know about the data sets that they're being trained on? And are there any limitations that are introduced by the fact that, you know, they are being developed within, what is ostensibly all walled, you know, the great firewall of China internet?

Do they still have access to X China data sets that they can pull in, even if those aren't being sort of publicly dispensed to normal people? So I think they definitely do. the one thing that all these Chinese tech companies, the model makers in particular, are very aware of. And I will say any tech company in China has to be aware of, is the red line that cannot cross when it comes to certain things that their product may generate. And I think this is one of the reasons why building a generative AI application or a company in China is so hard. Yes, they do have access to, you know, Wikipedia or the Common Crawl, any of the kind of the standard table stake data sets that you use to train all the large language models. so far, and then they can inject some of the data that is available inside the Chinese Internet kind of garden, a wall garden, if you will. But there are two interesting things about

their Chinese Internet, right? One is, I'm assuming everybody who's listening to this, is aware of the level of censorship that's in public discourse on the Chinese Internet. That does not stop Chinese netizens from wanting to express themselves in a certain way. So if you are an aficionado of their Chinese language, you will see a lot of, of contortions of linguistic kind of acrobats being used in all kinds of ways to publish public in China that you can convey certain meanings, but when it comes to modeling the language, actually becomes very difficult to model the accuracy of the language. So that's number one. And number two, because of, again, the red line of censorship moving at all times. So it's not like

a static target. A lot of contents actually get deleted on the Chinese internet. Things just disappear. You know, if you don't archive a certain article when you first read it, two days later, it might be gone for, you know, reasons that a lot of people can't even decipher it, even if you think about this all the time. And that reduces the quantity and, I will say, the quality of training data that exists on the Chinese internet in particular, that I think is almost a built-in disadvantage to a lot of Chinese model makers that are building their own LOM, you know, despite all odds, they're, of course, still continuing their efforts. to do so. One thing I like about some of the AI conversations is they get into the realm of

linguistic and linguistic theory. Can I just press you on that point about the nature of the language itself and the effect that that has on building a high-quality model? Yeah. So I would say two things that's interesting about their Chinese language in particular is that one, there are a lot of hominemes. There are a lot of sounds, there are a lot of characters that sound almost exactly the same. That means, you know, nothing similar. Oh, the famous example. is the whole, like, it's either a poem or a very long sentence with, like, Ma, like, horse and all these, ma can be, depending on the intonation can, like, mean a bunch of different things. You can construct a whole poem or sentence out of that single syllable. That's right. Like Ma, Ma, Ma, Ma, right, the four tones of the standard

Chinese Mandarin language. And if you just put that into the, to the tech context, you know, Jack Ma, Ma, Yu, the founder of Alibaba, has been either a celebrate entrepreneur or a villain. depending on how you think about his current public stature, right? That makes the modeling of any, let's just say, publicly available information about Jack Maugh pretty difficult when you're also building a generative application on top of it. And there are a lot of these examples in the Chinese linguistic case that makes not just the modeling difficult on a technical sense, but also what can or you cannot generate based on the red lines that you may or may not have to cross. Yeah.

So just to play devil's advocate here for a second. GPT4 was trained on like 0.2% Mandarin and is 99% as smart in Chinese as it is in English. And when you look at what some folks out of the leading Chinese AI labs are saying, like they're perfectly happy getting the vast majority of their training data from English and sort of porting it over to Chinese. So there is a little challenge there. But I also think it's the other thing on this sort of are the model is going to be lobotomized because they will be able to be too politically sensitive. I mean, look, I feel like this is basically

a solved problem. If you try to make chat GPT say something racist or sexist, it basically won't. And if Chinese internet companies have data sets on anything, it's all the posts that they've censored over the past 20 years. You know, having also tried to do this exercise of getting Chinese models to say politically racie stuff, it is a pretty hard slog. So, you know, on the plus side, I think, not all data that's going to matter for the future of AI and its economic and military impact is going to be linguistic and text-based. And, you know, being the world's manufacturing center, putting cameras into all those factories may help you train a sort of like AI manufacturing robot a lot better.

So there are real tradeoffs here. And it's not just like China is sitting behind the eight ball arms. You can get the news whenever you want it with Bloomberg News Now. I'm Amy Morris. And I'm Karen in Moscow here to tell us. you about our new on-demand news report delivered right to your podcast feed. Bloomberg News Now is a short five-minute audio report on the day's top stories. Episodes are published throughout the day with the latest information and data to keep you informed. Yes, there are other products like this from a

variety of news organizations, but they usually rerun their radio newscasts throughout the day. That's not what we do. We create customized episodes that can only be heard on Bloomberg News Now. And we don't wait an hour to publish breaking news. When news breaks, we'll have an episode up in your podcast feed within minutes. So you're always getting the latest stories and developments. Get the reporting and the context from Bloomberg's 3,000 journalists and analysts we're all over the world. Listen to the latest from Bloomberg News Now on Apple, Spotify, or anywhere you listen.

I just realized how old I am, because I remember when China censors, like, actually physically used to cross things out in newspapers. that were like imported in Beijing. This would have been around like 2004 or 2005. It used to be done by hand. One thing I wanted to ask, since Kevin brought up the sort of like vagaries of Chinese politics where one day something is really in

and the next day something is very much out, we have seen that bleed into financial markets and specifically, you know, funding for technology, primarily internet technology, SaaS in recent years, but also into housing and things like that with the three red lines policy. So I'm curious, like, what does the funding environment actually look like for AI at the moment? Is there a lot of, like, government support for developing this technology? And just while you're answering that, like, who are some of the companies that we shouldn't, like,

who are the open AI and anthropic of China that we should know about? Yeah, and how are they raising money? Like, is most of it coming from VCs or public markets? Yeah. Yeah. So when it comes to. the AI model makers, right? I'll put that in one category of new startups, which actually kind of came up around the same time as all the other ones that we already know here in the

US as well. You have Tripu AI, which is, you know, purported to be one of the company that's closest to open AI. You know, their vow to also develop AGI. They have strong academic roots from Qinghua University. You have Moonshot that I think Jordan mentioned. You have 01 dot AI, which is Li Kifu's new venture building models as well. And I will say there are probably six or seven Chinese unicorns that currently exist that receive funding mostly from your kind of Chinese standard VCs. These are VCs that are just like the sequoias and the Andresans of the world, right? You have Zheng Fun, you have a Hillhouse Capital. You have Hongshan, who was formerly a Sequoia Capital, China, and a few others pouring money into this particular

lane because one thing I think we should back up is that this is one of the very few lanes that VCs in China can actually put a lot of money into and expect the kind of return that you would expect from a pure financial VC because a lot of other parts of the Chinese tech ecosystem has been battered down quite a bit because of all the differences in the policy crackdown that happened in the last few years or so. So those are some of the big model maker players. And a lot of them attracting large corporate venture investments from Alibaba in particular, but also Huawei, Tencent, all the big dogs pouring into it very similar again to what we have here in the U.S. where Microsoft, Google, and Amazon all have like their player on the field, if you will,

when it comes to startups. And even some of the deal structures are similar. I think one of Alibaba's deal is that half of the investment they're making has to go into cloud credits to use the GPU service. in Alibaba cloud to be able to train their model as a way to entice, but also to tie them into Alibaba's own ecosystem. So actually, there are probably more similarities, I would say, than differences when it comes to the AI model makers funded by VCs.

But that's about as similar as it gets. Jordan, you mentioned the robots and the idea that another form of data is just all of the visual information that's inside a factory. And we did an episode on sort of the connection with AI and robotics, with Josh Wolf and we talked about some of these ideas. But, you know, you mentioned, you know, you're talking about, okay, they also want to achieve AGI. Are there distinct aspects, though, of the strategy that is like, oh, no, like, okay, we are thinking about AI in the chatbot sense. So that is the way most people in the U.S. think about what these AI companies have delivered. Is it the same with these Chinese

powerhouses or are they thinking about different forms of expression, maybe in a more industrial manner, et cetera, than the sort of chatbot idiom. So, Joe, one thing that this is more of a recent development, so I think just last week, Shanghai hosted its annual World AI Conference, right? And during this conference, it's a big show, and there are, I believe, 25 different brands of humanoid robots that were on display in the Expo, you know, Hall of the Center. So that illustrates, I think, to you that I think the direction, I think Tracy about this you know where's the funding going to right I think if the national

government were to have a say a lot of the funding ought to be going to AI applications that are what I call heart tech things that you can actually touch things that actually do certain things in the real world whether it's in a car factory in a manufacturing shop or even actual robots that will be in your home and I think that is the direction that a lot of Chinese companies that are not kind of in the model building world are going into to. And I think there's actually one very important kind of larger existential crisis that China is dealing with that a lot of people don't talk about in the context of AI, which is that by the end

of the century, the UN is projecting that the entire population of China will go down to around $800 million to about $1.4 billion right now, so roughly halved. And half of that 800 million people will be people who are 60 years or older. So demography is destiny in a lot of ways. And if you're Chinese policymakers staring into this abyss that we're just going to basically rent out of people, then having a lot of AI robotics is almost a national imperative for China to solve its own demographic problem with or without everything else going on around the world. Okay, so one of the benefits of having a command economy essentially is that Chinese policymakers can come out and say we want everyone to direct their money away.

from, I don't know, online internet, e-commerce platforms and into hard technology like semiconductors or developing some new chatbot or something like that. But the downside is that you can get excess capital to put it politely crowded into these things. Is that a concern when it comes to AI or is it that the technology is so strategically important and so competitive that it's not really a concern at the moment. You know, so I think it's an interesting contrast you can look at between sort of the AI model makers and the semiconductor and the broader semiconductor ecosystem. So what what Kevin laid out, which is what I see as well, is basically like a largely private sector driven thing with the same kind of boom and bust or boom and like question mark that we're current livingly living into in the West happening in China. Now, you know, what's happened in the Chinese semiconductor ecosystem over the past decade or so is really different.

So, you know, if you're talking about the U.S. Chips Act, which didn't come online until 2022, and it's going to spend roughly $75 billion. Like, we've seen that level of investment doubled or even trebled over the past 10 years in the Chinese government. You know, she is really semiconductor and hardware and industrial pill. And I think he also from a lot from a long time ago, saw the writing on the wall of these, you know, of these export restrictions coming on as as semiconductors became more and more and more strategic technology. She shishen ping sounds like he could be the perfect guest if anyone wants to disseminate that if anyone's listening, someone who's a hardware.

An open invite to Shishin Ping to talk about semiconductors. If he's hardware and chip-pilled, it sounds like he would be a great guest. So let's talk a little bit more. we know that the Biden administration has imposed along with some allies various limitations on the export of advanced chips and so forth. But why don't one of you just sort of give us the broad outline beyond or including the chip limitations of U.S. policy towards, you know, international AI policy towards China and how we're actively implementing the strategic aspect or the geopolitical aspect of AI? Again, you know, I started with this idea that AI is a strategic dual-use technology that has

big implications, both for the future of economic growth as well as the future of military power. So in September of 2022, a few weeks before the October 7th export controls, Jake Sullivan gave a speech where he said that it's not good enough to just be in the lead, but that the U.S. needs to have as far a lead as possible in critical emerging strategic technologies, AI being first and foremost on that list. So we've seen a number of different efforts on, you know, from both a data and algorithmic as well as a compute perspective to make sure that America can continue to run faster than China. So from export controls, you have limitations on the types of chips, which you can be exported into China. We've had EUV restrictions so that China can't make the most advanced chips

domestically, the types of chips that the U.S. is no longer allowing Chinese firms to buy in country. We're starting to see more and more restrict, you know, there's movement in the water around restrictions of letting Chinese firms access those types of chips abroad and with foreign cloud providers. And then there's also starting to be some interesting movement around restrictions when it comes to both data and algorithms. On the data side, you're starting to see new data walls be lifted with regards to electric vehicles in particular. We'll maybe see more coming forward. And then from an algorithmic perspective, you know, two weeks ago, Open AI said that they were nowhere longer comfortable giving API access to developers based in the PRC. And that may be a harbinger of

something larger. Congress recently gave this the power to explicitly stop algorithmic exports. So the wall slowly but surely is coming up. And the Chinese developers and semiconductor ecosystems response to this is really interesting because you're seeing very aggressive investment on the hardware side, you see stockpiling of chips and even billions of dollars of the best of what Nvidia can legally export to them. Smuggling too is starting to be a real thing. We recently saw some mainstream coverage of. So, you know, there will be kind of gray areas and people trying to play corner cases to get more technology and access. Industrial espionage as well is something that starts to, starting to bubble up a bit more. The scandal of sorts with

Leopold Ashton Brett are getting fired. for two weeks later OpenAI to bring on General NACOSONI, the former head of the NSA, to shore up their internal controls, I think is an interesting leading indicator of what's to come on that front. I'm Francine Lacqua, an award-winning journalist, and I've got a new podcast, Leaders with Francine Laqua from Bloomberg Podcasts. I've interviewed everyone from Heads of State to fashion icons about the news of the moment, but I've always been curious, who are these people as leaders?

I don't think there's one right way to be a leader. Make decisions. A poor decision is always better than no decision. Listen to new episodes every other Monday. Follow leaders with Francine Lacroix wherever you get your podcasts. One of the things that you sometimes hear about restrictions on, for instance, semiconductor exports to China is the idea that by cutting off the country from all this external technology, you're only going to accelerate its own technological progress and its own development of leading edge chips and things like that.

How much credence do you lend to those types of arguments? I think it's a moving target right now. I think logically you would expect China to put all of this resources into building the best GPU possible, right? If we all think AI is going to be the next big paradigm shift for national competitiveness, not just market competitiveness. Yet right now, because of, I will say, the effectiveness of U.S. expert control, two rounds of expert control and also working with allies around the world to solidify expert control,

it's actually been a very difficult time for China to catch up from the hardware side, both in terms of designing the best GPU possible, but also to manufacture the chips in bulk using its own chip foundry, SMIC, because China's been cut off from TSMC to be able to meet the demand that China itself, its own companies need, to build its own product, let alone expanding to other parts around the world. And that's why you see all these behaviors still

of either Chinese developers using VPNs and other ways to access open AIS APIs or you have these sort of smuggling of prevented or prohibited media chips into China to be able to keep the building of the infrastructure going. And I think that goes back to what you were talking about, Tracy, earlier, is that there is a limitation to how much the national government wants certain things to happen and whether that could actually happen or not when it comes to especially building hard tech in China when it comes to AI infrastructure in particular. We'll see a lot of pretty poor examples of government money being wasted in chip manufacturing already in the last few years before AI was even really a thing. And now it's in a very tough spot, I think, for a lot of people in China to be able to overcome the current hurdle.

And I think this gap, this specifically hardware GPU-imposed gap, will only widen when Nvidia mass-produced. It's a blackwell chip, which is the next generation of GPU that's going to be produced in bulk starting next year, which has a huge performance kind of, advantage, roughly 32 times more than the A-chips, which is what open AI use to train GPT for, to be able to close that gap. So I think, unfortunately, for the Chinese technology builders, the gap will only widen. Jordan, I think you've talked about this before as well. The idea that if China is restricted in the amount of compute that it has access to, maybe there's a scenario where its own AI models just become really, really efficient in one-weigh. way or another. Yeah. I mean, this is, you know, when you're under constraints and there's money to be

made and, you know, engineers will find a way, right? So I, you know, on the compute side, right, you know, it's interesting thinking about Blackwell. Only about 10% of the gains that it is making from the past generation came from node size. The vast majority of the gains that they've been able to squeeze into the Blackwell system are from stuff like, like optics and interconnects. And that stuff, high bandwidth memory. That type of stuff is a lot more difficult to export control than just one machine coming out of the Netherlands. So, you know, it's difficult.

There's tens of billions of dollars of R&D that's been on this stuff. But it's sort of an open question to me, how difficult reengineering all of the innovations that went into Blackwell is going to be versus reengineering an EUV machine, which like seems to be next to impossible. This is an important point. Just to jump in here. Like when people, when in pop culture, when people talk about chips, it's almost always like node size and smaller and smaller. I like the idea of like vibrant pop culture of semiconductors. I'm not sure we're quite there.

But this is how like ignorant people like myself, if I think like, oh, well, you know, the cutting edge chip is just in my mind smaller nodes. But this is like a really important point that you hit on that when actually when we're talking about, say, Nvidia GPUs, there are other dimensions that are actually much more important. node size. That's right. And, you know, if you look at all the new products or the Vidae is releasing to the market, it's not a chip. Of course, you can buy a chip, but, you know, it's really eight chips together connected with high bandwidth memory and the interconnects and all the other elements that goes into a system. And the next generation, the packaging will be 36 chip and then 72 chip. And of course, there's a good reason for that because all these models, if you believe that larger models become smarter models, then they need massive

of amounts of parallel processing to be able to train all these model efficiently and to deploy them officially. So that's why you tie up all these high-performing GPUs together. And I think that's what makes this hardware gap even more difficult from a system level for even a cloud level to really overcome. I think, you know, before we've been talking a lot about the chip war between different countries, I think we're vastly quickly evolving into a cloud war between different countries competing on AI and technology in general because it's all about the cloud and the system and how officially can you acquire and then utilize them that makes a big difference. Tell me if I'm wrong here, but my sense also with like semiconductor rivalry, maybe to a lesser

extent, but certainly with the AI rivalry is that it's mostly a competition between the US and China. Like as far as I can tell, Europe isn't necessarily obsessing. about developing national chat GPT versions as much as, you know, the U.S. and China are. Why is that? So, you know, I think from a, from the semiconductor perspective, right, like, it's really China and the rest of the world. There is a very integrated global ecosystem in which you have European players, East Asian players, South Asian players, the U.S., that are all kind of working together to push,

you know, to push node sides and improve the sort and develop the sort of, and develop the sorts of technologies which are going into something like Blackwell. And China, you know, has kind of been cut off, or is at some levels being cut off from that broader ecosystem and having to start to domesticate a lot of the technologies that they were able to formally rely on from other countries. Now, Kevin, I don't know if you want to talk to the like every country with its own model and cloud thing because I think that's a really interesting. Yeah, I'm curious about that. Yeah, I think, you know, not to beat on Europe too much, but I do think given just a regulatory, bias, right, towards releasing new technology really quickly in general. It just isn't a very

hospitable environment to be able to build a product like ChatGPT, which was really released, I believe, on a whim or on a quick competitive note, it wasn't something that's super thought through, but because of the environment that we have in America, we allow that sort of things to happen, and then we think about regulation later. Coming back to the national angle, I think there is a strong sort of motivation coming back to the cultural point that every country does want its own national champion to some extent. I think French France has mistrawl, which it kind of puts up as his own national champion. It's a very strong open source AI model maker that has a lot of strong relationship with Microsoft as well. I think there's another company out of Germany that also is a model

maker. So there are a few examples out there. Even Japan has, I believe, Sakana AI, which is its own startup national champion to model Japan's culture or truth or knowledge or whatnot into a national model. And India has a few others as well. And the Middle East is another very important region that is playing an increasingly large role. I think as the government actually pushed out some really good open source model about a year ago called Falcon, that series, which has been able to put the Middle East on the map when it comes to national development of AI model. So there is a lot going on, but certainly not nearly as much as the U.S.-China competition besides that. Actually, you just reminded me, this is something I've always wondered as well,

but what is the state of open source in China? Like, is it as much of a thing as it is in the states or elsewhere? What does that actually look like in terms of sharing code and things like that? So it's very much a thing. In fact, so I used to work at GitHub, which is one of the largest code repository of open source code. and to drive international expansion. So I got to work with a lot of Chinese open source developers. And it's actually one of the very few, I would say, vibrant communities

that really embrace a lot of openness, transparency. They have this really nerdy holiday on October 24th, which if you know binary is, you know, two to date. I thought so to attend it actually. But basically, it's called Programmer Day and it's a completely grassroots holiday in China that celebrates open source development. They have a huge party, a huge convention.

They talk about open source sharing. They all use GitHub or Hugging Face or all your usual open source collaboration platform. So the open source kind of movement, if you will, has really deep roots in China. And that has been around for a couple of decades already before we have this kind of open source, close source model dichotomy that's happening in AI. In fact, a few of the nonprofits that are kind of government affiliated in China has released open source models as well as a way to showcase what they have done as a way to kind of both compete but also to collaborate with academics and researchers elsewhere. Yeah, you know, just coming a little more on the government open source connection. You know, it's easier to be like open source is a good strategy, particularly if you're behind. We've seen a lot of investment on the hardware side with Risk Five as well as on. as well as as Kevin mentioned on the software side.

And, you know, I think it's still an open question, but a lot of, there's a lot of chatter on the Chinese internet about just to what extent the Chinese model makers are dependent on meta continuing to drop open source models in the Lama series. And, you know, we may, if and when Zuckerberg decides that he's done playing the open source game, we may see a very different trajectory for the Chinese model makers if they're not able to kind of, you know, get their algorithmic and hardware shops in order before that. So we just have a few minutes left. But, you know, we talked about the current state of U.S. restrictions toward China or how the U.S. thinks about AI and the geopolitical context.

There are chip export controls. There's more being done, you know, when it comes to cracking down an industrial espionage. There have been a lot of stories, including several broken by Bloomberg, about, you know, academic collaboration and, you know, trying, limiting who can talk to whom. And we know that there's a lot of issues in academia related to computer science and tech and Chinese researchers in the U.S., etc. We have this framework, but it might all change because the president, the party of the presidency might change in a few months. What do we know, if anything, about the Trump view towards AI and AI in the geopolitical context? So I think the interesting about Trump is that during his presidency, his first term, he actually did have worked on national AI initiative of something, which I believe is also chaired by Eric Schmidt, the former CEO of Google.

And he signed on to some kind of a national initiative plan in 2020, but it was, you know, middle of COVID. None of us had any time to pay attention. And it just kind of like didn't go anywhere. But he did have some kind of history with AI. But based on the most recent kind of national. national, the Republican National Convention's platform just being released. One of the lying item is that

they will overturn Biden's AI executive order so Trump can put his own spin or his own branding or his own substance on AI when he, if he does become a president again. So we can expect a lot of kind of volatility

on that front. And another thing that I would say that I think is the most consistent thing about Trump is his unilateralism, right? He's always been about America. first, American only. And right now, there is a pretty good rhythm of global AI governance that's happening, starting with in the Euclid, the Bletchley Park Declaration to a summit in South Korea earlier this year, to then another summit in Paris in February of next year, where you have

world leaders, including China, and all the other relevant departments and ministries around the world coming together to talk about AI governance in a global sense. were to put my money in the betting market about Trump, I think he will probably blow a lot of that out of the water to go to his unilateral instinct and really put America stamp on this. And it's hard to predict what that will look like. Trump went on the all-in podcast, maybe the podcast that must not be named on the show. No, it's fine. It was really remarkable because they asked him about China and AI. And he had like, you know, a moderately sophisticated answer, which was sort of shocking. He, you know, identified that it's a race between the U.S. and China. And he also said that

energy is really important to data centers, and we got to make that happen. You know, if you edited down his comments from five minutes to like 30 seconds, it was pretty coherent. But it was a remarkable thing that, like, he was talking to someone about this and they got like two halfway decent talking points in his head. So, you know, how he's going to approach technology in China is, I think, is a really open question. On the one hand, we had the. this very, like, well-thought-out, sophisticated playbook that he ran to basically get the world to turn on Huawei. On the other hand, you know, we had a lot of weird Trumpy stuff. Like with ZTE, the Commerce Department was basically going to kill it. She called him, said that it was going to cost too

many Chinese jobs. And then he went on and tweeted the next day saying too many jobs lost in China, like, we're going to give ZTE a reprieve. So he's really of two minds on all this stuff. Maybe someone talks him into being scared about inflation. Who knows? But I think there's a very wide range of outcomes, both on his domestic approach to artificial intelligence, as well as he's, as well as how he's going to think about technology competition in China. You said something there that actually I wanted to ask, and this is just about the sort of technological arms race again, but obviously within the U.S. and the data centers, there is a lot of anxiety about access to electricity and how Are we going to supply that? And we all know the U.S. is not in a good state these days.

It's like building things and adding a lot of energy capacity. When you think about U.S. versus China in that respect, could that be a comparative advantage for the Chinese companies? I mean, I know China continues to add more nuclear generation. And they're probably more liberal in terms of adding more coal power and other things like that. Could they end up with an advantage based on their electricity grid on solving the power system? side of AI? You know, this might be a bit of a red herring. I am like cautiously optimistic that Republicans and Democrats will end up kind of giving some, you know, NEPA exceptions for these data centers, which like everyone is now agreeing is this like incredibly important national strategic

resource. I mean, I think the bull case with China, right, is if they're stuck with less efficient ships and just have to build data centers twice the size that are twice as expensive, then like, yeah, maybe they can like throw on a few more coal plants and nuclear reactors together. them to get them going. But I am I am cautiously optimistic that InVideo will continue to make increasingly power efficient chips and you know state local and national regulators will understand that this is something worth bending the rules on which is sort of what you've seen on the sort of semiconductor fabrication build out over the past few years with the chips act. And if I can add one more point to what you just said Jordan, I think a lot of the focus that we've

talked about has been the the algorithm, the model, how, how are you, how are you? for the GPU is, but a lot of engineering sources has been devoted and will continue to be to reduce the power-hungriness of AI model. Because one of the big things that's a hugely hotly debated topic is that all these tokens that we're generating for all these apps are just way too expensive, right? The expense comes from the compute and then the power that's being consumed to it. So there will be a lot of engineering that is going to accomplish when it comes to reducing the cost and the need for energy over time. So I don't think the energy trade is a long-term investment thesis, but maybe a short-term arbitrage as far as how much extra power we need

just to feed the hungry AI dinner center when all the engineering is going to reducing that dependence because everyone sees how that could be detrimental in a lot of ways. Jordan and Kevin, thank you so much for coming on Odd Lots. That was fantastic. Our goal for the episode of Understanding This More, I feel I do. I do. So thank you both so much. for coming out. Yeah, that was great. Tracy, I thought that was great. I thought, I mean, there were just a bunch of things that, like, I, you know,

even setting aside the geopolitical aspects that I just didn't know about what the AI, you know, space looks like in China. And that was very helpful right there. I'm sort of obsessed with the programmers day now. I want to learn more about the holiday. Do you think all the open source programmers, they, like, all get together to celebrate and do a dance, maybe a bite dance? Yeah.

Oh. Boom. Come on. Okay, Tracy, you get that. We'll leave that one in. Thank you. The producers will add that cricket noise that we have on hand. I am also interested in the holiday. I do think, so Kevin's first answer, I just thought was like really clarifying about why a country might think that having a national version is important. Right?

Because if, and I hadn't really thought about it in those terms, because on some level, right? Like it's easy enough. Like if it's, you know, Lama, you can use that anywhere on, you know, various chips and stuff. But the idea that like the AI is going to like codify truth as it means in the context of that country. And, you know, of China, their version of what's true, Kevin gave the Jack Ma example, which is that one day he may be a hero and one day he may have a villain or maybe a villain. Like, that's always changing. And so there may be some impulse to have the model sort of correctly reflect what is state policy. But this idea that like it captures, you know, creates this version of the national identity culture and language. It's very interesting. No, absolutely. And I think China is

sort of the extreme example of that. And that's sort of what I was alluding to. Yeah. Not very well in the intro when I was talking about data control. But like there is this cultural element to AI where you are using something like chat GPT or maybe not at the moment, but you could imagine a future where you're using something like chat GPT in the same way that you're using something like Wikipedia, where it's almost like the codified public version of events, right? And so if you are a government like China, why wouldn't you want to have some degree of control over what that's spitting out and the narrative that it's building? I was also really intrigued that, you know, we talk about chip wars all the time, but the idea that there is a distinction that is useful to make

between a chip war and a cloud war, and that, okay, there is, you know, we all know that, like, supposedly, you know, smaller nodes or whatever are better, although I think in the context of GPUs that's less important, but that so much of the technology is about the interconnect. And we talked about this with Corweave, that there is a distinct element to building up a very powerful GPU cloud versus a traditional cloud because of the cables and the packaging, et cetera. and so the idea that it's about more than just like the chip itself and, you know, the size of that chip, so to speak. No, absolutely. And one thing that really stood out from that conversation, and I'm thinking back, this actually happened when I was in Hong Kong and I was, you know, coordinating some of the news coverage there. But I remember the big crackdown on the sort of like retail internet companies like Alibaba.

And I remember at the time, policymakers were very, very specific. and very deliberate and very open about trying to direct more capital, you know, less disorderly capital flowing into SaaS, more capital going into hardware and making things like chips. It was almost, you know, for lack of a better tech analogy, it was like flipping a switch, right? They basically said like no more money going into retail tech, everything going into semiconductor manufacturing and all that stuff. So it was interesting to hear like both the challenges and the opportunities from that perspective in China. By the way, I don't know if it was the financial times, but that was the first, whoever came up with Chats-She-P-T, that was very clever. That was very

well done. I first saw that in that term, that word that in the F-T, but that's a pretty clever term. That was a good one. I'm disappointed to find out it's actually vaporware, although no, I know, I know. Perhaps not surprised. Okay, shall we leave it there? Let's leave it there. This has been another episode of the Oddlots podcast. I'm Tracy Allaway. You can follow me at Tracy Allaway. And I'm Joe Wisenthall. You can follow me at the stalwart. Follow our guests, Jordan Schneider.

He's at Jordan, S-C-H-N-YC, and check out his China Talk newsletter and podcast. And Kevin Schu, he's at Kevin S-Schoo, and check out his writing at Interconnected. And follow our producers, Carmen Rodriguez, at Carmen Armin, Dashel Bennett at Dashbot, and Kel Brooks. Thank you to our producer, Moses Andam. And for more odd lots content, go to Bloomberg.com. slash odd lots, where we have transcripts, a blog, and a newsletter. And you can chat about all of these topics 24-7 in our Discord.

Discord.g.g. slash oddlots. We have an AI room in there. Maybe when this comes out, we'll see if either Kevin or Jordan or both would be up for doing an AMA in there. The day it comes out, I'm sure people have a lot of questions. So we'll try to make that happen. And if you enjoy odd lots, if you want us to, I don't know, start a pop culture, semi-conductor spin-off show, then please leave us. us a positive review on your favorite podcast platform. And remember, if you are a Bloomberg subscriber,

you can listen to all of our episodes, absolutely ad-free. All you need to do is connect your Bloomberg account with Apple Podcasts. In order to do that, just find the Bloomberg channel on Apple Podcasts and follow the instructions there. Thanks for listening. This is Tom Keene, inviting you to join us for the Bloomberg Surveillance Podcast. It's about making you smarter every business day. I'm Paul Sweeney. We bring you complete cover of the U.S. market open. We cover stocks, bonds, commodities, even crypto, all the information you need to excel. And I'm Alexis Christophis. Bloomberg Surveillance also brings you the analysis behind the headlines. We do that through conversations with the smartest names in economics,

finance, investment, and international relations. We do all this live each and every weekday that bring you the best analysis in our daily podcast. Search for Bloomberg surveillance on Apple, Spotify, YouTube, or anywhere else you listen. On the East Coast, listen at lunch. And on the West Coast, listen as soon as you wake up. That's the Bloomberg Surveillance Podcast with Tom Keen, Paul Sweeney, and me, Alexis Christophorus. Subscribe today, wherever you get your podcasts. Bloomberg Surveillance, Essential Listening, each and every business day.