Sovereign AI: Geopolitical Strategy & Industrial Policy for Countries 3-193, with Anjney Midha, a16z

Hello, and welcome back to The Cognitive Revolution. Today, we're exploring a fundamental question that governments around the world are wrestling with. If the United States and China are going to continue to dominate the fundamental AI inputs of talent, data, and compute, and thus own the AI capabilities frontier, what exactly should everyone else do? My guest is Anjana Mitta, general partner at Andreessen Horowitz. Who brings deep software industry investment expertise and a remarkable ability to channel diverse national perspectives on the opportunities and challenges of AI to this conversation. We begin by considering what sovereign AI means to the enterprise, which I honestly hadn't planned on coming in, but does make a lot of sense considering the degree to which companies are both driving investment and adoption around the world and informing their respective governments' priorities. Here, it seems regulatory questions around which companies have to share what information with which governments under what circumstances loom so large that United States-based hyperscalers are beginning to face new competition from national champion AI data factory companies like the ones recently announced in France, the UAE, and Saudi Arabia. With the conversation grounded in that reality, we then turn to what countries should do and and are doing. On the talent front, we agree that there's really no way for most countries to compete at the frontier today. But Anjanae does argue that governments should play for longer timelines and try to both partner for access to the fruits of frontier labor now and build the complementary and last-mile local delivery talent that they'll need to be successful on a 10 to 20 year timescale. On data, Anjanae makes what I think is really an inarguable case that all governments should be tokenizing their cultures. And we then debate whether they should expect better results by partnering with frontier developers or by trying to own fine-tuning and application development locally. Though fortunately for them, as long as frontier open models continue to drop, those options are not mutually exclusive. On compute, it seems that most countries will be asking themselves a complicated mix of questions, including what level of independence do we really need and want, and what mix of partnership and local buildout is required to get there? Just how reliable are the U.S. and China as partners, and do we have to align with one or the other, or can we play both sides? Obviously, that last question has become much more complicated recently, as the United States has shaken the foundations of its global alliances while in the midst of intensifying strategic competition with China. Still, aside from a few countries like the UAE and Saudi Arabia that have extremely deep balance sheets and can offer regulatory arbitrage in exchange for being cut into the global AI buildout on a strategic level, My sense coming out of this conversation is that for most countries, some limited investment in critical infrastructure would be prudent. But beyond that, they're probably better off letting the market develop naturally and accepting really quite a bit of dependence on the United States. At the very least, I sincerely hope that the US proves to be a good bet as an AI partner, first and foremost when it comes to responsible development of the frontier technology itself. To be honest, if it doesn't, I feel like the rest of the world might have even bigger problems to contend with. Obviously, this is a super dynamic space, and this was a great conversation. I really think that the empathy Anjanae demonstrated for so many different positions that national leaders find themselves in, and the build-buy-partner framework that he brings to this analysis, are valuable contributions, and I look forward to doing it again soon. With that, here are a couple of quick disclaimers from me and from a16z, which, as you probably know, recently acquired the Turpentine Network. First, I personally am solely responsible for all content on The Cognitive Revolution, including guest and topic selection, the questions I ask, and the editorial commentary I offer. For this episode with Anjanae in particular, I received no consideration to have him on the show. And while I did follow my usual practice of sharing questions in advance and allowing Anjanae to review our edit and request any additional cuts, no topics were off the table and nothing meaningful was cut during the editing process. And second, this is directly from a16z. This information is for general educational purposes only and is not a recommendation to buy, hold, or sell any investment or financial product. Turpentine is an acquisition of a16z Holdings, LLC, and is not a bank, investment advisor, or broker-dealer. This podcast may include paid promotional advertisements. Individuals and companies featured or advertised during this podcast are not endorsing AH Capital or any of its affiliates, including but not limited to a16z Perennial Management LP. Similarly, Turpentine is not endorsing affiliates, individuals, or any entities featured on this podcast. All investments involve risk, including the possible loss of capital. Past performance is no guarantee of future results, and the opinions presented cannot be viewed as an indicator of future performance. Before making decisions with legal, tax, or accounting effects, you should consult appropriate professionals. Information is from sources deemed reliable on the date of publication, but Turpentine does not guarantee its accuracy. With that, I hope you enjoy this exploration of how companies and countries are navigating vexing strategic AI choices with Anjaney Mitta of Andreessen Horowitz.

Anjaney Mitta, general partner at a16z, welcome to The Cognitive Revolution.

It's great to be here. Thanks for having me.

I'm excited for the conversation. So today we're going to be talking, and I'm sure we'll touch on other things, but the core theme is this notion of sovereign AI. And I think it's actually the first time that I've covered it in 200 episodes of the podcast. So that's probably on me for being neglectful of what I see as an increasingly important concept and idea that, as geopolitics is literally exploding all around us, is certainly going to become more and more important. My card's on the table. Obviously, I think AI is really important. I think governments should be taking it seriously. I think they should be thinking about how to invest in this. technology wave and bring the best of it to their citizens and also protect their citizens from the worst of it. And yet I find myself a little confused by this concept of sovereign AI, and I'm hoping you can help shed some light on the subject for me. So maybe for starters, what is sovereign AI and what is kind of the big argument for it? And I have a couple of different candidates, but I'd love to hear how you think about the big arguments for sovereign AI.

Yeah, no, It's a good question. I think we should start with sort of first principles, right? Um, one, I think we should acknowledge that nobody has a working, uh, a canonical definition of sovereign AI that seems to be consistent across different regions.

Yeah, we don't even have AI agents defined, so, uh, exactly there.

Yeah, so we are definitely in this weird regime where you have nation-state leaders and CEOs of the world's largest companies talking about how important sovereign AI is. Um, yesterday, Satya, I think, tweeted that they're launching their European sovereign cloud. And if you look on the website for a definition of what they mean by sovereign AI, you can't find one. Right. And I think that this is one of my icebreakers now when I get together with policymakers or leaders who happen to be finding themselves in the position of not just being geopolitical leaders, but now actually having to have an opinion on AI. So I often ask them, well, what's your definition of sovereign AI? Which is like the policymaker version of the what's your definition of an AI agent test, right? And it's a great icebreaker because you ask 5 different people, you're going to get 5 different answers. And so let me try to paint a picture of what I think it means to people and then what I think it should mean. What I think it means to people is basically it's become synonymous. If you're a nation-state leader, if you happen to be responsible for the future direction Of the atomic unit of a country or a large company, I think it's your way of communicating that you want control over your own destiny as much as possible as AI plays out. If you're a technologist, if you're an engineer, you're a researcher, I think it tends to mean more a collection of services and models that run locally on-prem without having to rely on cloud, traditional cloud infrastructure. That is then beholden to some set of geographical governance. So I'll continue playing out what I hear from different people. If I'm talking to a CIO of a company in Europe and ask them, what is sovereign AI? What I often hear is, to me, it's a way to have our workloads, AI workloads, located in a way that I know it does not have to comply with backdoor information requests from foreign governments that I don't have to comply with if I don't want to. Some of those are American regulations, some are Chinese, some are Singaporean. But if you're a company at any kind of scale and you have customers in various regions, it's a complete nightmare for you if now your AI workloads are located in a place that make you have to comply with the regulations of that place. So if you're a CIO, that sort of sovereign AI is your way of saying, look, I want to know where my workloads are and what regulations they comply with, and I want control over what regulations I have to comply with. If you're a nation-state leader, like a prime minister or an elected official, then I think it stops to mean something concretely technical and starts to mean something more cultural.

Right.

Where if you say, look, a large percentage of my population is going to be talking to AI models as their daily companions, a large percentage of my country's mission-critical industries are going to be running on AI workloads. And not only are these models computing infrastructure, technical infrastructure, but they're also a form of cultural infrastructure where they're trained on training data. They have a set of post-trained values in them, and I want those values to reflect our own and not some adversarial country. I think sovereignty for them means the independence and the ability to control what value systems that cultural infrastructure reflects. So again, I think it means different things to different people. But if you're asking for a sort of subset, I think it's a set of technologies and services that both convey the technical independence of traditional cloud infrastructure, but also the cultural independence of having to adhere to some other country's value systems that don't align with yours. Does that make sense?

Yeah, I think that's pretty interesting. I hadn't even really considered it from the perspective of just the enterprise, but certainly it does make sense to want to be mindful of what government backdoor sort of requirements you might be walking into without even realizing it if you're not thoughtful and take care to minimize that kind of stuff. I had been thinking about it more from the nation state or cultural perspective coming into the conversation. Another way I was kind of thinking about it is just in terms of inputs, right? I mean, if the inputs to AI are data, compute, and algorithms, you could kind of say that the data is obviously the sort of cultural stuff that can go into a model. The compute is the data centres, and those are very sort of physical, real things that have a very specific place in the world.

Right.

And then on the algorithms, I would kind of equate that to talent. If I'm thinking from an enterprise perspective, I was actually just talking to a friend yesterday who works for a kind of mid-sized boutique technology consulting firm that serves enterprises. And he was asking me, how do I think about these sort of data issues? Because the people at our big enterprise customers, they're like very conservative, tons of lawyers. We have a real hard time getting them to take any risk.

Right.

They're going to want to bring everything on-prem. They're going to want to have everything in-house. What do you think about that? And my response was, I would try to push through it. It was kind of like, if I'm an enterprise, I think I would— or if I was advising an enterprise, I think I would try to get the legal team comfortable with the Microsoft terms of service and feel like you probably already trust some big tech provider to have some custody of important data. And this maybe isn't necessarily so different for your queries to be going to the Azure cloud versus your own cloud.

Right.

In fact, they might even be more secure there in some ways, depending on what your threat model is relative to hacking. Like, what's more likely to get hacked, your stuff or Azure? I don't know, but It's certainly not obvious which is more likely to leak. So how do you think about that? And with Microsoft taking something to the European market, for example, and saying, like, okay, hey, we've got data centres here and it's all local in some continental sense. Like, you don't have to worry about American jurisdiction or Singapore or what have you.

Yeah.

Does that— would you kind of come down on the same place and say, like, these enterprises are maybe a little overly focused on controlling their own thing relative to using the best provider and, like, getting the benefits sooner, perhaps?

Yeah. Yeah. So this is, it's a good question. It, it, it, here's a framework for how I tend to think about it, which is, um, I like red teaming this sort of supply chain that, that delivers the product or service that you're depending on, because an enterprise is just like sort of a Silicon Valley word Because everyday people don't use that word, right? If you ask, what is an enterprise really? It's basically either a business or an organization that makes large purchasing decisions on behalf of users. If you kind of trace back the first principle, the rise of this concept, the etymology of the word enterprise, it started to become a thing in the 2000s when the idea of selling something and using something kind of diverged, right? So you were often selling a piece of software technology to somebody who actually wasn't the end user because that person was the purchaser or decision maker on behalf of the organization, right? So that would be sort of classic top-down adoption or diffusion of a technology. Then there was a sort of violent disruption of that with bottoms-up SaaS, right? Where— or software as a service, where Any individual within an enterprise, a company, an organization, even a government could say, hey, there's this cool new tool called Dropbox or Slack, right? And I can just go swipe my credit card and start using it. And then the rest of— and my colleagues can do that if they like it too. And then at some point, you know, there's enough of us using it that makes sense for us to adopt that across the entire organization. But that was a bit of— that was a huge shift in the way technology and software was diffused. So traditionally you had this top-down buyer model, right? And when you say enterprise, I often think of governments as the largest enterprises. So you often have the minister of IT for a country deciding what stack the country should run on, right? And you have that. And I think what we're seeing right now is a Back to the Future moment where in the pre-cloud era, right, where you had and prepackaged software, where you had a lot of on-prem infrastructure, you had CIOs and CTOs making decisions for the entire company. We're going to use Linux. The entire company is going to use Linux. And we're going to go to Red Hat and buy a bunch of licenses to allow us to use the enterprise version of Linux. And then we're going to buy, you know, really the consulting and the support that came with that contract to manage the deployment, the complexity of deployment. Of that system. And then the second thing we're going to be buying is basically insurance, which is when the system inevitably fails, who's down the line? Who's providing the indemnity and the hook? Who's accountable to raise their hand and say, we will fix it. This was our fault. We'll fix this problem. And so I think when you— any conversation about the enterprise is the framework I try to use is what are they buying? Are they buying technology or are they buying insurance? And in the top-down era, pre-software as a service, pre-cloud, you were often buying sort of solutions, right? Like you were buying technology and then you were buying insurance from the same person that was bundled. That's why we had the rise of someone like Red Hat. That got completely unbundled in the SaaS era. So you had— because individuals were buying the technology and then the risk was moved, or like the insurance was moved to the CIO or the chief compliance officer. And so, and then we've been in this sort of like scramble, like this unstable equilibrium, I would say, over the last 10, 15 years, where you had individual users within the enterprise adopting the technology just because it was the best product that they wanted to use, like Dropbox or Slack or whatever you call it, or GitHub, right? And then you had CIOs constantly chasing them to try to buy insurance and say, well, wait a minute, like this tool isn't SOC 2 compliant with the risk posture of the company. Or in some cases, when you're buying for the entire country. So who's on the hook? And then that resulted in the rise of hyperscalers or cloud companies like Azure basically saying, hey, we'll sell you insurance. You should work with us, especially for infrastructure products, because we'll indemnify. We'll provide you the security, the indemnity. And so when something bad happens, we'll raise our hand and say, you, the CIO, you're not on the hook. you can blame Azure. And so, you're buying job security for yourself and your team. And that was an extraordinary offering. If you look at, like, AI, GenAI terms of service today versus 2 years ago, Azure now indemnifies a bunch of the AI services versus 2 years ago, there was no indemnity offering, right? So, I'd like to— if your friend is asking you, hey, what should I use, Nathan? Like, And should I trust Azure? The first thing you need to ask is, is he trying to buy technology or is he trying to buy insurance? And I do think we're in a Back to the Future moment where open source— what I'm seeing is certainly in regions like Europe is the largest enterprises, especially the ones who want sovereignty like we've described before, that independence over the supply chain, are turning to companies like Mistral to say, we want you to provide us the technology and the insurance. So you come help, you, you like, you figure out what the right implementation should be because it's just way too complex. There's too many models coming out every week. We have no idea how RL, you know, RL from verifiable rewards works. We don't even know what a reward model is, but we want you to come automate. Let's say I'm, I'm the, uh, let's say I'm CMA CGM, right? The 3rd largest shipping company in the world, $70 to $80 billion in revenue.

A year.

And I know I want to automate the entire port operations and cargo operations for my company, but I have no idea whether I should be using like Mistral 7B, Mixtral, DeepSeek R1, a reasoning model, non-reasoning. My head is spinning because every week there's something new coming out. And at the same time, so there's— I need somebody to sell me the technology and to manage the complexity of the technology. And I want somebody to say, don't worry, I've got the— I'm managing your risk for you. When it comes to anything going wrong, whether it's cybersecurity risk, copyright issues with the outputs, whatever it might be. Right. And I think we're getting back to the future era where the largest enterprises are turning to one or two companies to provide all of that. I'm often in the US, I often run into Palantir as sort of that provider for mission-critical workloads where they manage the technology deployment and they provide indemnity. And so I think it depends on whether your friend is basically wants to buy Azure insurance, or do they want the best technology? And often right now you're not getting both of those from the same place if what you're looking for is frontier open source models. Hey, we'll continue our interview in a moment after a word from our sponsors. Does that make sense? Yeah.

I mean, that all sounds to me like it bottoms, bottom lines to like, Probably you should go with Azure and not, like, what would be the argument for actually trying to bring all this stuff on your own enterprise on-prem servers?

Well, one, if you don't trust Azure, right? If ultimately Azure has to comply with government regulations that today are governed primarily by American law. And so if the workloads that your friend is running are in mission-critical industries outside the US and their customers would not be comfortable with backdoors to the US government, then he should probably not go with Azure. So for example, the US Cloud Act governs all workloads run by US cloud companies. So that is one way to break it down, right? Is do your friend's customers care about the sovereignty of their workloads. Would they be fine with the US government having total access to those workloads? And if the answer is no, then that's one reason that they should be going with their own infrastructure. Yeah.

So I guess to maybe try to summarize that again, it's like the trend seems to be pretty clearly toward a rebundling of technology and advisory services and this sort of insurance component. But because of some of these laws, like the US Cloud Act, as a European company, for example, you might want to go with a European provider of that bundle as opposed to an American provider of that bundle.

Right.

And obviously there could be some nuance in terms of exactly how the US Cloud Act applies internationally. I'm certainly not an expert in that, but I can understand why you might be more comfortable going with a Mistral as opposed to a Azure based on that, right? While still basically seeking the same bundle.

Correct.

Yeah.

Now what's happening, and this is hap— so, so that is playing out, I would say, in many of many regions across the world, right? Where, um, you have an effort like, uh, Stargate here. It's a new kind of fully integrated cloud offering that starts with a single AI provider. Integrating the chips, right? The data center that then runs the chips, the compute, the compute part of the stack, the entire model part of the stack, as well as the end applications like ChatGPT, right? So that's certainly running, working in the US. And for the first time, I think we saw that 2 weeks ago get internationalized with the deal that OpenAI did with the UAE, where they said, we're going to offer ChatGPT to the entire country. And the government will pay for it because it's essentially free for all residents of the country. But the idea that now those workloads are running on the OpenAI offering for countries versus being run on Azure is the first clue that there's a rebundling of the cloud happening where hyperscalers are getting unbundled and rebundled by AI companies. Right. Where in the previous era of— in the rise of the hyperscalers, the rise of the cloud, The primary organizing function was the marginal cost of compute. And by compute, I mean storage as well as CPU workloads. And by centralizing almost all of these workloads in one place, the economies of scale were really, really hard to fight. So let's put aside for a second the whole insurance argument that we talked about earlier. Just from an economic perspective, it just didn't make sense for most companies to run their own workloads.

Right.

Which is why the classic technology purchasing decision is no different. It's build, buy, or partner, right? And in the pre-cloud era, you just bought, you had to build your own technology, your own infrastructure, your own data centers. And then when the clouds came, started presenting you with the value proposition of like, we can just offer you all of that, but cheaper and abstract away the the complexity for you so you don't have to manage your own infrastructure. Most enterprises, most countries and companies chose to buy instead of building. Until then, a few came out on the other side who were such large customers, or they had regulatory reasons where they had to then go back to building, right? And I think that's what's happening now is a country like the UAE is basically saying, we'll just buy the sovereign AI stack. And we trust OpenAI and we love them and we'll just buy it. There are other countries and regions that are going, no, we're going to build. And it's not just countries. Some companies as well are saying, we'll just build. We're going to own our own chips. We're going to want open source models running on it. And we're going to handle the fine-tuning, the post-training, as well as the deployment. We'll take some open source application, our own version of ChatGPT, and we'll handle that. We will deploy that for our civilians or our customers. And so I think you want to kind of analyze, like in those regions, they're not purchasing Azure, right? They're purchasing a full stack rebundled offering, or they're choosing to build that full stack themselves. So in the case that, you know, this is what Mistral Compute is, which is something they announced last week, where NVIDIA and Mistral are building a cloud. This is a new animal, right? Where it's NVIDIA chips. So these are American chips. Being racked in a data center, which will be Europe's single largest data center locally, with a cloud platform on top that runs both Mistral's models as well as other open source models, with then the customer's choice of whatever application they want, including both open source clients versus not. And so that's a— it looks more and more like an open scaler that's rebundled cloud offerings along the dimensions of open source.

Right.

workloads first, that where open source workloads are the first class citizen versus, uh, what has been the dominant paradigm for the last, uh, you know, 15 years, which has been clouds bundling, providing cheaper and cheaper marginal cost of compute to customers for centralized workloads, but are then proprietary, are run primarily on proprietary technology.

So let's maybe shift to the value to like countries and citizenries. Yeah. And maybe I'll go in order of talent, then sort of data models, culture, and then infrastructure.

Yeah.

On the talent side, it seems like Mistral— am I saying that right? You're the authority on this. They seem to be one of a very small number of sort of national champions that has continued to pretty much keep up with the frontier. I mean, I wouldn't say they're quite at the frontier frontier, but they're not far behind. And certainly there is a critical mass of talent there. And that is something that is unique in Europe and close to unique around the world outside of the US and China. Right. So I guess that seems like a win sort of just on its own merits, right? Like, if you're a country and you're like, well, who knows what might happen? Again, geopolitics is exploding in all sorts of ways all over the place. we'd like to have a domestic talent pool that if walls really come down, we can still do this stuff within our own borders, with our own people, and feel confident about that. That seems like a pretty clear value prop, right? I do wonder, how many countries do you think could actually achieve that? It's notable that Germany doesn't seem to have a similar thing. And you can go down the list of most populous countries, and there are a lot of countries with a lot of people that don't have anything like this. Do you think they should be trying to create them, could create them, or would you worry if you were advising Russia or Brazil or big countries with lots of resources, would you be worried that they would just be kind of wasting money on something that's never really going to amount to anything?

Yeah. Okay, so this is a good question. I'm going to go back to a I'll answer your question with a story, which is, um, there, there was a moment in post-World War II when a lot of folks were looking at the, the way the diffusion of modern— another technology was changing the world, which was modern finance. So post-World War I is when the Bretton Woods Conference happened, right? And a number of countries met up to talk about Adopting the US dollar as a sort of global reserve currency. And then post-World War II is really when the adoption of that started accelerating. And a lot of countries got together and said, wait a minute, this dollar thing is really working. And that's resulting in 2 clear power centers. There's America that is the purveyor of the single global reserve currency. And then there's Europe, Which is turning out to be one of America's largest trading partners. Of course, this was when we were seeing the rise of the Middle East and Russia as well because of the petrodollar. And a bunch of other countries were going, if we're not on the other side of a trade partner of the US, well, what do we do? We're just going to get left out in, in the modern finance era. How else do we amass reserves of our own reserve currency? And so most countries navigated that build by or, or partner trade-off differently, but it resulted in the rise of a new kind of nation that I call essentially a hypercenter partner nation, right? Which is, this is a country that may not have the size of the big guys, may not have the talent of the big guys, but understands how to insert itself in the flow of value in the global economy. And that's what happened with countries like Singapore. Singapore's a tiny island nation, right? With fewer than 3 million people. So did not have a massive talent base, did not have vast oil reserves like the Middle East did to trade with the United States, and was not a net producer of currency like the US is. So how did they navigate that? Well, they said, what we're gonna do is, is make ourselves essential to the flow of global dollars going from the US and Europe into the East and provide a stable, a stable rule of law regime. Uh, you know, Lee Kuan Yew did a whole ton— the founding father of Singapore did a lot of work to basically make Singapore have the lowest corruption index in the world, the highest, uh, or the lowest cost of doing business index, and the highest ease of doing business score in the world. And that resulted in them amassing this extraordinary wealth, just creating an extraordinary base of wealth for the country. And at this point now, Singapore has one of the highest GDP per capita populations in the world and like one of the top 10 sovereign wealth funds in the world. And so my answer to your question is basically you either have to build, buy, or partner. And if you, if you are either a hypercenter with the compute reserves and the talent, like America, China, and now increasingly Europe, or you have to find a way to partner and insert yourself in the flow. of tokens in the global AI economy. And I think that's the race we're seeing play out right now. The result is that you're essentially directionally correct. I don't think most of the world's economies have access to the kind of frontier pre-training or post-training talent that Mistral has. The vast majority of that talent is concentrated in the US. Europe is one of the few places that has it. And most countries outside of Europe, China, or the US either have to build, buy, or partner that talent.

And boy, is it getting expensive. Zuckerberg is going to be hard to outbid, it seems.

That's right. I mean, Meta— that's why when you say enterprise, I think of nation-states and Fortune 500 companies as essentially in the same atomic unit. So if you're competing against Mark, like Zuckerberg for a researcher, you cannot bring a knife to that bazooka fight, right? And outside of the US, Europe, and China, there's not that many private companies that can fight that battle. So you, it ends up being nation states, right? And that's why there's so many countries going, wait a minute, this is a, if AI is a core piece of national infrastructure and we're bidding on talent against Meta, then we can't leave it to our private sector, which is tiny, to fight that fight for the country. We need to— back it up with essentially sovereign dollars. And that's what's happening, right? If you look at the rise of G42 in the UAE, right, which is a kind of— is an infrastructure company, but backed by the sovereign wealth fund. They are playing the kind of role in the economy that AWS or GCP plays here, but they're entirely state-backed, right? So governments have had to step in because the sheer capital required to play at the frontier is outside of the purview of the buying power of most companies.

Yeah. So would you analogize then what the UAE and perhaps also Saudi Arabia are doing right now to the Singapore play from years past?

100%.

You're sort of saying we're going to find a way to insert ourselves into the global flow of tokens. 100%.

Except they don't need their Singapore moment right now because they've got oil. So what they're going through is their petro— if you trace the history of Singapore, What's so interesting is they navigated the rise of the petrodollar through the concept of being an entrepôt where they said, okay, we don't have our own oil, but we see this massive flow of one resource turning into another, which is, you know, petroleum turning into dollars on the other side of the world. And the only thing they had going for themselves, it was their location. And so Lee Kuan Yew did a bunch of strategic kind of partnership deals, very reminiscent of how Sam Altman and OpenAI have been doing a bunch of ecosystem deals where All of the crude from the Middle East, or the vast majority of the crude oil, would be shipped on containers to Singapore and then refined in Singapore. A value-added tax was added and then shipped out to multiple other Western countries that would buy. And that's how Singapore navigated the petrodollar regime. The UAE does not have to do the same thing because they've got the petrodollars. What they're trying to do is convert their petrodollars to petroflops. Right. By buying vast amounts of GPUs, basically compute infrastructure, primarily from NVIDIA, and saying to the world, well, come run your workloads here. That's the whole G42 pitch, right? Is over the next 30 to 50 years, let's convert our petrodollars to petrocompute and attract the world's largest customers to run their compute workloads here. And then the vast majority of our GDP should be decoupled from the price of an oil barrel. Like, if, you know, the more time I spend in the Middle East, I realized that the way we follow interest rates here in the US is how they follow the price of a barrel, right? And entire investment projects are greenlit or not depending on the price of an oil barrel that week. And they would like to decouple their future, certainly their technological future, from the price of oil. And compute and AI inference is that path for them.

Let me come back to that infrastructure in the Middle East question in a second. And I have a number of questions on that.

Hey, we'll continue our interview in a moment after a word from our sponsors.

But just to take one beat on the data side and the linguistic and cultural and that sort of value side, because we've covered talent. And I think we're basically on the same page that it's going to be very hard for almost anybody to amass enough talent density to really compete for frontier AI. That's right. It doesn't mean you don't want to develop your workforce generally, but that's sort of a frontier versus diffusion question. And all these companies or countries should be promoting diffusion, but that's quite distinct from trying to amass the density to compete at the frontier. Then you also hear this argument around, well, we want the AIs to be good in our language, and we want them to reflect our cultural values and all that kind of stuff.

Right there.

Again, I feel like if I'm advising, let's say, 80%, maybe even 90% of the countries in the world, and I was like, okay, you've got your data, your culture, your language, you've got all this stuff. How can we make sure that you are best served by AIs with respect to all that makes you unique and special.

Right.

I think what I would say for almost all of them is do a massive data collection and curation project, and then literally just take the data on a silver platter to all the hyperscalers and say, hey, please include this in your future training runs, because we want your models to be fluent in our language and rile about our idiosyncrasies and talk to us the way we want to be talked to and reflect our values back to us. Would you give the same advice? And here I'm thinking, like, Mexico, Argentina, Australia, you name it, right? All these sort of middle-tier countries that we've kind of established probably can't compete at the talent frontier. But there's this notion that if they don't do that, or somehow that they might miss out. And my suggestion is just like, They want data. You have data. If you can bring it to them and allow them to use it, they'll happily incorporate it. And then you'll get kind of what you were looking for in maybe the most efficient way. Well, okay.

So to be clear, I don't believe that the countries can't compete at the talent frontier. I just don't think they can compete today. They need a short, medium, and long-term plan to get there. Right. So in the short term, look, look, at the end of the day, there's no, there's no secret recipes for RL. Like everybody knows how to— Mistral published Magistral. their reasoning paper, DeepSeek put out the recipe. So the recipes are known. I would say AI, there are some tips and tricks that you learn from, especially when it comes to the infrastructure required to do online RL learning, that are hard to learn. But there's no fundamental reason why Singapore or the EU or Mexico could not long-term build out a talent pool that can do pre-training and post-training of their own models if they wanted to. Over the arc of, over a 20-year arc, if they started investing today in the right partnerships, the right education institutions, the right infrastructure, there's no reason they can't fundamentally can't get there. Right. So the question is, what do you do on a short, medium, what's the roadmap to get there?

Yeah, I'd put the singularity somewhere in the like 5-year timeframe. So the 20 years sounds long to me, although Obviously, I don't know for sure that that's going to happen.

Yeah. So that's it. I think that's one way to organize the timelines, right? And there's certainly some countries that are more bitter lesson and AGI pill than others, right? So the UAE is more AGI pill than Mexico today. So the UAE is saying, yes, it's hard to plan outside of 5 years. And so let's do whatever it takes to build out our infrastructure and our talent pool here. And they're choosing to partner with OpenAI. They're choosing to partner actually with multi— they're trying to partner with everybody because their belief is like they have a 5-year period to establish their independence before the singularity arrives. The countries that are less AGI-pilled, that are less urgent about AI, are planning on like a 20-year horizon, right? But let's say for a second that the singularity arrives. I'm quite sympathetic to the idea of the gentle singularity. Right. That it's not some violent rewrite 5 years from now that suddenly magically the entire economy collapses and then you're no longer relevant. I still think if you're Mexico, you want to plan for a 20-year horizon on which your young population can graduate to having useful jobs and purpose in life. Right. And if a big percentage of that is doing, you know, useful post-training work on AI models or doing integration work between AI Like what, you know, whatever the frontier models are and integrating those into the rest of the economy. That talent pool is certainly, you can certainly today, I would argue even that talent pool is really hard for them to get, but you have to invest in your capability to build out that talent pool over 10 years. For example, I do think one of the most sought after skill sets today is the equivalent of a forward deployed AI solutions engineer, right? Someone who can bridge both worlds. Is that what we call them here?

That's my own personal branding.

But right, but it's someone who can take a model and then integrate it into the right workflow. Often that happens in, in some real-world task, right? Often that requires doing like, like genuinely knowing how to do variable reward design. And hope maybe, you know, maybe all of this will be automated 5 years from now. But as a country, I think it's possible to do what countries like India did in the '90s. with auto manufacturing, right? So in, in 1992, India started liberalizing and said, well, we would like to become— go from being a socialist country to being a modern capitalist country. We don't have— and we think auto manufacturing is one of the frontier economies of the sector, uh, of, of the world rather. And we need to develop that capability. We have— we just don't know how. We can't leapfrog to doing that overnight. So let's partner. And so they invited companies like Suzuki, Who came in and set up these joint ventures locally where they trained up the local population. They created this company called Maruti Suzuki, which was responsible for producing the vast majority of cars in India for like a decade. And over time they transitioned the running of that industry and that sector to local talent. By the way, the US did the same thing with oil in Saudi, you know, in the '50s and '60s with Saudi Aramco. That was originally Shell. Right. The Standard Oil Corporation of America went and did oil exploration in the Kingdom of Saudi Arabia. And then over time transitioned that to a local talent base called Saudi Aramco. So I think that's the healthy way for the US to bring its allies along into the singularity is to do joint partnerships, kind of what NVIDIA is doing with Mistral in Europe, to bring along allied partners, because if we don't bring them along, they're going to turn to adversarial countries like China to do that. And so I just want to be clear, in the short term, that talent race is intense. That doesn't mean in the long term you can't partner with the right folks to set you up for a better future at the talent frontier. And I don't think it's just a talent thing. I think that extends to all parts of AI frontier progress, whether that's data or compute. So to your question more directly about data, I think what you should be doing is, if you're a country, certainly figuring out What are ways to get your culture tokenized? If your culture is not tokenized, then fundamentally you're reliant on some other country to tokenize it for you. And so I think tokenization is the most underdiscussed part of sovereign AI, right? Is get that corpus that reflects your culture and values, tokenize it, and then figure out how to do pre-training on that corpus. And I think there are several countries actually racing to do that today, certainly for non- for languages that have a different root than sort of base European languages. Does that answer your question? I don't know if that was—

Still, it does. Although I still am kind of like, where are they really going to get the best results? And it still feels to me like just bringing all that data on a silver platter to the leading developers, and you could maybe do both, but if you were Brazil, for example, and you're like, okay, we've got all this massive data and whatever, and now we've curated it and tokenised it. Now we could do one of the great things about information. And also one of the, I think, maybe a disanalogy between some of these earlier industries, auto and oil, is like, this stuff is so intangible, and to some degree, some of these things kind of only have to happen once, whereas you have to actually stamp out every car. The cost of duplicating a trained model is obviously vanishingly low compared to the upfront cost that goes into it.

Right.

But if I'm Brazil and I'm like, okay, I've done all that work, now I can do all these things. I can go to OpenAI and give them our data. I can go to Anthropic and give them the data. I can go to Google and give them the data, and I can try to develop a national champion. I would be pretty confidently willing to bet that Brazilian national champion is going to be worse on a 2 to 5 year timescale at least, using that data and actually even serving the local culture, as opposed to the hyperscalers who I just think are going to be basically better at everything. And I don't know if you would take the other side of that bet, but it seems hard to win.

I think it depends on what your definition of a local champion is. If the local champion is the provider of Brazilian language models to the vast majority of its civilians or mission-critical industries like defense, healthcare, and finance, the value lies in the last mile, not the pre-training. So at that point, yes, they should absolutely not try to develop their own pre-training capabilities in the short term, right? Because I think that just gets commoditized. They're much better off finding a pre-training partner. And ideally somebody who's an open-source pre-training partner, because then they have a ton more customizability over the post-training, right? Because once— if you have access to the weights, your ability to do weight adjustment and on-policy updates to the weights, if you have that direct access, is dramatically higher than if you don't. But I completely agree with you. In the short term, you should find a pre-training partner instead of trying to replicate that. But the last mile distribution effects you have in your country, the last-mile advantage you have of knowing the behaviors of which companies are the best positioned to consume that technology, how to productize it for them, how to do RL certainly on non-verifiable rewards, which is really where this stuff becomes really fuzzy and you need local kind of handholding, you're tremendously more advantaged than OpenAI or Anthropic. Have you ever talked to ChatGPT in a language outside of— in any local language?

Minimally. I speak enough Spanish to experiment with that a little bit.

My Spanish is terrible, but I'm a native Hindi speaker. And when I talk to ChatGPT in Hindi, it speaks like an American tourist visiting India with an American accent. And its diction is different and its linguistic choices are different. It's like a tourist. And so for any consumer use case in Brazil where your civilians don't want to talk to a tourist, they want to talk to a local, or a mission-critical industry where the last-mile integration you only trust a local partner for, you're best off developing a local champion that's the AI partner for that, right? And then I completely agree with you, pre-training should be a partnership effort with whoever is the right technology partner. that has it. And there's, we've talked about how there's only 5 or 6 teams who can do that today. But over the long term, I think what happens is, you know, knowledge diffuses out and that company, which is today your local AI champion, which is largely a last mile provider over time, will learn the knowledge required just like with AI, you know, with Aramco and refining, uh, or with Suzuki and manufacturing. I don't think there's anything You're right that the marginal cost of producing a model is different from producing a car, but inference is an ongoing muscle, right? And continual post-training is a muscle that diffuses out, and you should have local capability over the next— in the AI economy future.

How much does that depend on the trajectory of maybe just overall capabilities and specifically closed versus open model capabilities. And obviously nobody has the answer to what that's going to look like. But one could imagine that if you take the AI 2027 scenario, a big part of that scenario is the frontier developers start to close down. They're kind of running away from the rest of the competition. Nobody can match their ever 10x-ing training runs.

Right.

And they also start to even keep some of their models secret in that scenario, which is a whole other can of worms. But I think it sort of depends on how— would you rather use O5 in Hindi, even though it's maybe not as native, or would you rather use an O1 equivalent that has that local post-training?

Look, it took 26 days to go from O1 to DeepSeek R1. So what China has demonstrated is that they're able to fast follow the frontier of reasoning within 60 days, right? And we're sitting here in June, right? It's been 6+ months since O1 came out. And of course O3 Pro has come out, but the distance between O3 Pro and the new version of R1 is not some step function that's hard The current data seems to be that the frontier of closed source and open source are basically moving in lockstep at roughly 6 months or less. And I think there were a lot of experts up in front of Congress a year ago, very confidently claiming that the US was 5 to 6 years ahead of China, which to me felt like I was being never under— I was drinking crazy juice. I was like, what planet are you on? DeepSeek had been putting out incredibly strong open source models for like 8 months before R1 came out. So for anyone who was paying attention, it was clear that the gap between closed source and front, you know, the open, closed and open source was not some, some, some like multi-year race, right?

Still, I mean, that's an interesting dynamic too. And again, no, I don't think anybody has a great crystal ball for this sort of thing, but one does wonder, it's not like these are open source projects in the traditional software sense of being like, community-driven and sort of anybody can— I don't know.

They're nation-state backed at this point.

Yeah, they're still hyperscalers who are just making a strategic decision for the moment to do that open sourcing. So I do still wonder, how long will China continue to do that? And for that matter, how long will Meta continue to do that?

DeepSeek is not a hyperscaler. So DeepSeek is not a hyperscaler, right? They are at this point a hedge fund that was Arguably under a ton of regulatory pressure to shut down their core business because the CCP believes that financial instability for the markets as introduced by market makers like hedge funds is net negative, who then turned to doing frontier AI research really well and then got a meeting with Xi to then continue doubling down on their frontier AI research efforts. So I don't think analyzing the incentive of DeepSeek as a hyperscaler is an accurate lens. I, I think analyzing their, their game theoretic sort of optimal strategy from the, from the position of a geopolitical player with the blessing of the CCP to stay at the frontier is, is more realistic. And I think they essentially, from in, in the, within the current regime, will continue doing whatever it takes to have a Chinese model open sourced at the frontier and adopted by every country on Earth. as a form of soft power, right? Like, DeepSeek was one of the most extraordinary— between the 3-way release of DeepSeek R1, the Unitree humanoid, and this animated movie coming out that— it's called Nezha 2, which was a blockbuster hit. There was a complete soft power revival in China. in the first 6 months of 2025 relative to the last 2 years, which is this malaise in the Chinese economy, right? The primary narrative being we're losing technological supremacy, we're losing cultural supremacy, and we're losing financial supremacy because the financial markets were on this downturn ever since the real estate bubble there. And whether or not you consider DeepSeek getting lucky or not, what we're certainly living in a regime now is the CCP views frontier open-source models as a core national capability. And we'll continue open sourcing that as long as they see that as a way to ensure soft power globally. Right. It doesn't have anything to do with the hyperscaler revenues.

Yeah.

Yeah. No, that's a good, definitely a good point. I feel like I have a hard time predicting what exactly they're going to decide to do, but that's a good baseline.

Um, that's, yes, yes, that's right. What's actually unique about China is they're, the way the structure works of the market is that as long as Alibaba keeps monetizing DeepSeek As the enterprise kind of distribution arm of DeepSeek, which is kind of the stable equilibrium they're in right now. DeepSeek is the AGI research frontier lab that's sucking up all the best frontier research talent and continuing to publish and open source ruthlessly to further the, I would say, the country's reputation on the global stage. and Alibaba keeps monetizing it in the enterprise, I think the CCP will maintain that as a healthy sort of steady-state equilibrium, right? Because it benefits both the party and also benefits GDP. There's now an enterprise infrastructure software deployment arm within China that knows how to do all the solutions engineering work we were talking about earlier to actually diffuse DeepSeek into the economy. Because actually, if you remember R1, It's a pretty unwieldy model. It's a very hard model to actually use. It climbed all the leaderboards and from a raw capability standpoint, it's great, but it's an MOE that's really massive and very hard for most businesses to know what to do with. And Alibaba is doing a lot of that work for them now. So that's the sort of the steady state equilibrium I see for a while.

So let's turn to the infrastructure. This is obviously what gets the most attention because there's the most dollar values and Then you've got these sort of exquisite assets that are getting plopped down all over the globe, right? I started because I'm no Dylan Patel here, but I do have access to deep research. So just to kind of get a baseline, and one of my working theories over time has been that basically AI sort of converges to look like the cloud, because in the limit, it sort of is cloud. Like anywhere where there is compute, you could run AI. And so Why would it diverge that much from that?

Right.

So what Deep Research told me when I just asked it to characterize the distribution of data centres today was that, first of all, the US has about 45% of the global data centre infrastructure. The top 25 countries account for 88% of data centres, and there are That obviously leaves the bulk of the countries with only about 10%, and many countries with no significant data centre footprint really at all. And then one other aspect or angle that I asked Deep Research to look at was how different parts of the compute stack might be distributed differently. And it basically said that compute is more centralised, the most centralised. Storage is less centralised. You've got like CDNs and all that kind of stuff. And then network is, of course, the least centralised because you got to actually get to all the end users. So I guess the big question is, in the fullness of time, meaning maybe the 5 years between now and the singularity, do you think that the AI infrastructure footprint looks meaningfully different than that? And if so, how would it end up being different and maybe why?

Yeah. So, so I do think that it would look very different in some ways, but almost identical in some others. And I'll explain, you know, which, which is which, but as we discussed earlier, I don't think there's a, like, I don't think there's going to be a fundamental change in sort of traditional non-AI parts of the data center economy. Right. So compute, storage, networking, good old-fashioned EC2 instances, good old-fashioned CPU workloads. I mean, look, those are the workhorses of the economy today. And most businesses should benefit from centralization where the vast majority of your workloads are running on Amazon East or whatever. Right. So for traditional non-AI workloads, I don't think, I don't see that changing anytime soon. I think public clouds have been fantastic for most software and technology adoption, and that's going to be the case. AI workloads, whole different story for 2 reasons. One is if you look at the, if you X-ray a data center, a data center is not a data center anymore in that it does not refer to the same like bill of materials that it did just 5 years ago. Right. Something like 60 to 70% now of a data center is GPUs. This is a huge change. Whereas like 10 years ago, less than 10% were GPUs. Right. So this, this is why Jensen, I think, is on a whole, um, you know, is on a, on a whole like tear to try to, to get everybody to realize he's calling, he calls them AI factories. Right. He wants to call them AI factories for a reason, because he thinks this is not just marketing. He's like, guys, under the hood, the data centers look fundamentally different now. We should not be calling them data centers anymore. The bill of materials has changed completely. The workflows they're running has changed completely. So for that part of the new data center economy, it's a full, like it's the, it's the most aggressive sort of rewriting of the cloud infrastructure business that we've ever seen, because If you don't have the frontier model that developers want to use, they literally are switching clouds, right? This is why arguably Amazon had to invest $8 billion into Anthropic when they were losing like $100+ million contracts to Azure in 2023, because they didn't have a frontier model on Amazon infrastructure, right? They literally just didn't have a GPT-4 alternative. Now, I was an early investor in Anthropic, and so I got a chance to see some of that, the calculus behind the scenes, and it was very, very painful. The public cloud workloads were being moved on the basis of who had the best AI model. And that's why we're currently in a race where Google is trying to integrate as deeply as they can across the entire stack, right, between Gemini and TPUs. Amazon is in a race to try to integrate across Anthropic and the Trainium, their custom silicon, right? And arguably Microsoft is trying to work on custom silicon as well. They have had an 8-year effort there. I think they're behind, but Satya would like to control his own destiny, right? They don't really have their own frontier lab. I would say today that is their equivalent of Anthropic, but they're trying to build that capability in-house. And so I would say that the data center economy is not going to change very much for traditional workloads. For the AI workloads, we are in a full arms race. And it seems like, I think tightly integrated workloads within the Google stack and within the Amazon stack will basically be like the Gemini and the Anthropic economy will continue to be highly centralized workloads. That begs the question of what happens to NVIDIA, right? Because if the largest proprietary models are fully integrated on custom silicon with the clouds and they're going full stack, then NVIDIA's core business is at threat, right? Because by the end, if by the end of this year, 2 out of the top 3 models, right, Gemini and Anthropic, are running on non-NVIDIA hardware, that's a core existential threat. And to Jensen's credit, I think he saw this. on the right, you know, saw this, like, saw the writing on the wall and started investing in a whole ecosystem of new AI-first clouds, like CoreWeave, like Nebius, and most recently, like Mistral, right? And so I think you'll have— where I think we will go is basically 3 or 4 different types of clouds. You'll have the closed scalers, like the Google Cloud and the Amazon Cloud. And then you'll have the open scalers like Mistral Compute, which is a close, tightly integrated offering between an AI frontier lab like Mistral and NVIDIA. You will have regional clouds like Nebius and CoreWeave is another one. I think there's a couple of these in the Middle East like G42. There's one called Humane that the Kingdom of Saudi Arabia just launched. And then the rest of the world is going to have to decide, do they want to be on an open stack or a closed stack? And, and I, your guess is as good as mine where things go 18 months from now, but I don't think Jensen's gonna, I don't think Jensen will just give up the entire AI factory economy to the custom silicon winners of the cloud.

Right.

And, and you know, what's beautiful about him is he's very much of a, because he's such a deep technologist, he's like 4 steps ahead. Like he knew his biggest customers were also trying to kill him. Starting a decade ago. And this is why he's been such an aggressive partner to frontier AI startups, right? Because he sees them— as long as a company, a developer like Mistral, a frontier research team like Mistral keeps reaching for NVIDIA chips, then NVIDIA will see a healthy ecosystem continue growing globally. I think the natural arc of centralization, like you said, is very strong, but it's not unchallenged. And then the primary players, I would say, the live players, right, are now the AI startups that are changing the destiny of these workloads. Like, if I told you a year ago that Mistral was going to be running a compute with tens to hundreds of thousands of Blackwell chips serving workloads in Europe. If you had told me that, I would have laughed, right? But that's literally what they announced last week. So these new open scaler wars are going to be one of the primary stories that get written in the next 18 months. And it's not, I don't, I don't think it's a foregone conclusion at all that the clouds end up winning that, the traditional clouds. But they can, they can have the CPU workloads.

So if you're a country that maybe hasn't had data centres on your sovereign territory historically, right, I guess I'm wondering, what is the case to try to get them on your actual soil now? And I don't really know country by country, but certainly we could go find countries in Africa where it's just like they just don't have it. And then maybe there's, maybe, I don't know, I'm guessing here, but maybe Mexico, for example, probably doesn't have near as much data centre presence in Mexico as they maybe use and are probably just effectively importing that across the network from the United States. Again, I'm guessing about that, but seems like a pretty good guess. Should those countries sort of just say, okay, well, it's worked for us thus far for kind of traditional compute, we may as well just kind of let the market run its course again this time, and we can import. Who cares if we're calling across a border for inference? We're not really in a great position to play that game anyway, so fine. Or should they be saying, this is different, we have to have GPUs running on our own sovereign territory because something? It seems like that impulse is out there, but I'm not quite sure Why or what they hope to gain from it or protect themselves from by doing that?

This is the idea of cultural infrastructure, right? If you don't have the infrastructure locally, the compute, at the end of the day, hosting and running inference for your models, then you're beholden to somebody else to run that infrastructure for you. And that means from a supply chain and strategic autonomy perspective, if that compute's not in your country, you're dependent on those API calls that you're shipping, you're hitting, to come back with full uptime and not returning a 404 error if some government that governs the infrastructure that your workloads are hitting turns off. I mean, I don't think it's net new from a— if you see AI workloads, particularly inference, as core national infrastructure the same way you view parts of your defense infrastructure, then you just have to decide, well, which parts of this, you know, supply chain do we need strategic autonomy over? We're going to set that up locally. And which parts do we not? Use that, you know, use that from partners. So like, look, the vast majority of the world's governments don't— countries don't build their own jets, right? They don't build their own tanks. They buy them. But then there's some parts of their supply chain that they go, wait a minute, actually we do want strategic autonomy over this subcomponent of the aviation industry. And then they build that out locally. Some governments, by the way, did have realized that they depended too much on foreign governments for critical defense parts of supply chain. This was the case with Europe, right? For the vast majority of the last 70 years, Europe has basically consistently outsourced core national infrastructure To other parts of the world, they have now since corrected that with an eight hundred billion or are starting to correct that with an eight hundred billion dollar defense bill, saying we we're going to have to fight you know for Ukraine because the as Vice President J.D. Vance said at the Munich Security Conference, the Americans aren't going to subsidize this for us anymore, right? And we can't and and so we need to do our part of funding and spending for our defense if we want to be continue being part of. NATO. And so I think if you view compute from the lens of sort of core national infrastructure and cultural infrastructure, it comes down to like, yes, I could ship my workloads off to somebody else, but what happens when they turn it off? Is that okay? Right. And some, for some workloads, I think the answer is yeah, sure. Like, you know, if, look, if 30% of my country's citizens are using a cloud in some other country for entertainment, If they turn it off, that may not be strategic autonomy, but for mission-critical industries like defense, like healthcare, we probably need strategic autonomy. And so I want those compute workloads to be running locally. Does that make sense?

Yeah, it's a tough logic right now. I'm a proud globalist, I think, and I feel like as we head into this AI transition, which, you know, I hope is gentle, but I'm not so sure. I kind of wish we were maintaining or even deepening the level of integration across countries and sort of increasing the barriers to conflict. And this seems like a classic prisoner's dilemma sort of situation where from every individual perspective, there's an argument for we don't want to be the ones dependent.

Yeah.

But there is like a common good in mutual dependence of the sort of Classic. Like, I think it's now been violated once or twice, but like traditionally no 2 countries with McDonald's had ever gone to war. There seems like there's something there that was like worth preserving. And in Chile, we're kind of liquidating that. And I agree with you.

But look, it starts from— let me put it this way. If we want— if that's a future we want to live in, then we need to acknowledge one, That the Chinese adversarial threat is very real. And then we as America need to bring along our allies and be stable, reliable partners, technological partners. But if we're not going to be sending those signals, then what more can we expect than the rest of the countries to try to develop their own plan? That's why I'm a huge proponent of a Marshall Plan for AI, by the way. Right. I do. I think it's a huge win for America that Mistral in Europe is going with American chips and not Huawei chips because they could, like the European continent has been courted by the Chinese semiconductor industry, like, like never before in the last 18 months. And yet they're choosing to proceed with American semiconductor partners. This is a huge win for globalization. There's a huge win for America. Um, and I think that's the template that we should want, like, globally, right? Open models running on American chips is a great template for global integration like that, of the kind you just described. I just think we have to acknowledge that the model layer as it currently stands has a ton of vulnerabilities if it ends up being centralized and closed source. And does not confer the benefits of sovereignty, of independence, and of self-determination, of cultural independence that most countries, most, at least most wealthy countries want. And so I think the sooner we acknowledge that, that they want sovereignty over the model layer, but they're totally happy partnering with us over the semiconductor part of the stack, I think the sooner, you know, we rush to at least 2 stable blocs in the world. Because, you know, it doesn't seem like anytime soon China is interested in, in pausing sort of their rush for frontier AI progress.

Yeah, I wanted to get into these recent deals, the, especially the ones in the Middle East. Uh, but the one with Mistral in Europe is also really relevant. I guess my question on that is like, do the Chinese really have chips to sell? I mean, I know that they have like 5G. infrastructure to sell. And that's been a whole battleground as well. Who will use Chinese and who will not? There's been pressure applied in all sorts of different directions.

Right.

But my broad sense in the AI era is that, for example, DeepSeek has said publicly, despite and in contradiction of sort of official Chinese narrative, that they are limited by their inability to acquire more GPUs.

Yeah.

And so I, I guess I am wondering, is it really realistic to think that like all these countries that the US is currently partnering with could actually go to China and buy? Like, my sense is just the supply is not there and their domestic market would eat up all that they can create and like, and lots more.

I think the evidence is that chip sanctions so far On China have had clearly one impact, which has resulted in an enormous sort of doubling down of local investment in Huawei's ecosystem. I think there's an alternate future where if we didn't put any sanctions on them, right, we could continue making their, their 20-year, the 20-year arc of the semiconductor industry, they're dependent on the US. I think that ship has sailed now. Partly that was started by the Biden administration, where they made it clear that China could not count on a stable supply of frontier, let's say, 3-nanometer or below chips from American vendors. But until that happened, Huawei is not even in the game. Like for frontier chips, they are now in the game. They have sovereign backing from their country. They have a ton of momentum now and motivation to develop, to close that capability gap. And they're not there today, but there's a national level race there to catch up, certainly starting with inference. Like they may not get there with frontier training capabilities for, let's say, 3 to 5 years. But, as long as the current arc of compute efficiency stays on track, I don't think there's any reason why you can't be running frontier reasoning model workloads circa 2025 on Huawei Ascend chips by the end of this year. In fact, I wouldn't be surprised if the DeepSeek R2, whenever it comes out, makes that clear that it's Ascend compatible on day one. Right. They are in a full-on tear to try to decouple themselves from American chips. And I think that's, that's the position we've created. They got, you know, the US has created for them. Um, time will tell if it was the right one, but I think if you look at the data, it's very clear that Huawei is in a much stronger position today than it was 3 years ago in terms of narrowing the gap between their chips capabilities with, with the US. And so, and that's the arc. I don't think you should do it. I think doing a static point-in-time analysis of their capabilities is dangerous. What we need to be asking is, we need to be doing like a 5-year analysis and where they're heading is to, they will be able to close the gap. And because workloads are becoming more and more efficient, I don't think it's that crazy for them to keep, to be able to decouple. At least the inference part of their ecosystem from the U.S. within 2 to 3 years.

Yeah, I've seen them put up a hospital in, what was it, 6 days or something? So, uh, the— never underestimate the speed.

Never underestimate the Chinese.

China might be able to do something. Yeah, there's a funny incoherence, I think, sometimes in our policies where we simultaneously make China out to be this, like, you know, big, terrible bad guy that's going to— I'm not even sure what exactly the threat model is supposed to be. I often ask people, what exactly am I supposed to be afraid of China for? Are my grandkids going to be speaking Chinese? What is it exactly that you expect? You can offer an answer to that if you have one. I usually don't get great answers, but we have this sort of, oh my God, China, they're so scary. And then on the other hand, we're like, but if we cut them off from this, they'll never catch up. And I'm like, I don't know which side of that It doesn't seem like it's a coherent idea, or at minimum, it seems like there's sort of a needle threading that we're trying to do with this policy where we can stay ahead just long enough to get to the singularity and they won't catch up. But a lot of things have to kind of— seems to me a lot of things have to come together just right for this policy to ultimately pay off in a sort of new American hegemony via our values being dominant in some sort of superintelligence or whatever.

Yeah, look, I know earlier I said that there's nothing secret about RL, the recipe is known and nations can catch up if they want. The corollary to that is they have to invest at the right time. And I'm very sympathetic to the argument or to the line of reasoning that AI is a tech tree and you have to get involved early enough. enough in the tech tree to develop your own capability years from now if you want. You can't just say, I'm going to bury my head in the sand for 5 years and then in the future at some date I'll show up and then we'll try to catch up on RL or AI progress then. Chips is very similar, right? And I think we have accelerated their adoption of the tech tree. What is the threat now? If we stop, if America and our allies stop progressing along the tech tree, and instead there's a future where you're entirely reliant on China's tech tree, then let's make no mistake, then our mission-critical workloads and so on are going to be dependent on whatever governance they choose to impose on the infrastructure running these workloads. And look, as much as I, from a cultural perspective, I think Or from, from a technological perspective, I think it's okay to, to admire their engineering prowess. But from a cultural perspective, it's— we have to acknowledge that this is one of the most ruthless authoritarian governments whose biggest victims are often its own population, right? And they're willing to do things that I don't think we'd ever want our families having to be subjected to. And so if we— I think if we just want— unless we are all okay with the trade-off in civil liberties and oftentimes persecution that comes with the Chinese form of governance, then we, we, I think we have to ensure that we have our own frontier stack that's independent of China. Right.

Um, yeah, I'm totally sold on that, by the way. We should build our own stuff. We should have our own fabs. Uh, you know, that makes— and by we, I mean the United States— I think that makes total sense and is almost inarguable. Where the rubber hits the road for me is more on the question of, should we try to deny them similar access? And if so, why? And how do we make that not self-defeating? And those, I think, are much harder questions. I certainly wouldn't want to cede the AI territory at any level of the stack to China, but I— Ah, I'm much less convinced that we should be trying to deny them their own version of it, if only because it just ratchets up tensions and just makes everything else harder. And I'm not sure that we're necessarily accomplishing all that much.

But no, I agree. I believe in the opposite, which is let's ship the best technology and export it to everybody so everybody adopts it, including the Chinese.

Yeah, I've made that argument too. Like, if you really want American values to dominate AI, like, why not just give them free inference? You know, now they'll probably not even accept it, but that's a whole other question.

Yes.

That's how, that's what they did with the internet, right? They subject, they created their own internet and prevented our companies from operating there. We never reciprocated that. And arguably that's why we are stuck in a situation where ByteDance and TikTok You know, is a— is more successful arguably in America than Instagram is. Different conversation. I think, you know, I, I, I, I, I, I'm certainly concerned about the arc of post-training on short-form video like that is going to eat the brains of our— the next generation in a way that I don't think we are prepared for. Even, even, even though I, I've ordinarily been sympathetic to the argument that like humans are resilient and we, we Arguments were made for previous technical waves that video games were terrible for our kids and TV is terrible for the kids. I don't think we are, and yet we turned out to be fine. I do think RL, like human preference today, can be weaponized at a speed and scale that humanity has not really been prepared for.

I think that's totally right. I mean, if you take what Facebook has done, and this might be about, roughly speaking, OpenAI's plans. I mean, they literally just hired a person to run applications who was a big player in developing the newsfeed into what it now is. But if you apply that same level of optimisation to one-to-one AI interactions, I think you do end up in a— or at least there's a pretty good chance you end up in kind of a dark place.

I wouldn't, I would not over-index on the executive hires, but I understand your greater point, which is that, you know, AI RL on human preference is a very, very, very powerful way to drive addiction to AI products. Agreed. I would not want the Chinese government doing that to our population.

And I agree that it's, um, certainly premature to judge any individual. And, you know, you could also make the argument that who better to avoid the mistakes the second time around than the person who was deepest in the details of it the first time around? So I could see that going either way. I don't know. I want to also just real quickly say, just to put myself on record, I remain very open-minded to the idea that China really is a terrible bad guy. I live in Detroit. Not too long ago, there was the story that came out of literally the Detroit airport, where somebody came in from China with some fungus sample that is like a—

That's right. Yes.

And I have been poking around trying to get a little more information on that. It's hard to get good information on such a story. One could wish that we had people in key positions that were making these announcements, like at the FBI and the DOJ and whatnot, that were Maybe a little bit more universally trusted than the people we actually have. But what I have heard has been basically like, yeah, dude, that kind of checks out and you probably should be afraid. And so I'm updating on that and I'm like, man, if they're willing to try to release bioagents that would destroy— this wasn't necessarily about to be released, but anything messing around in the domain of bioagents that destroy crops, Right. That's insane. Not only is it just like straight up evil and dangerous, but like, I have to imagine it puts their own population at risk, right? I mean, these things are not the kind of things you can just release in a controlled form or fashion and expect to only harm the other side of the world. Like, spores can float a long way on the winds, and I just think that's like totally insane. So I am, you know, for people who have framed me in the past as overly dovish on China, just know that I'm at least trying to continue to pay attention to this and update my worldview as new evidence comes in, which from time to time it does. Okay, these deals in the Middle East, I believe you were even there for some part of this conversation, right?

Yes, I was not physically there for the announcement of a number of the Trump administration's new deals there. I just have been involved in A number of the founders and companies I work with have been approached by both American as well as Middle East, I would say, data center providers, compute providers, government folks who are all actively trying to create a joint ecosystem across the 2 regions, across the Middle East and the US. And so, I've just been peripherally involved in some of these conversations. But I was not there for the trip, for the President Trump trip. Gotcha.

I guess it seems pretty intuitive. We kind of talked an hour plus ago now, I think, at this point, about the strategy that the Middle East countries are running to try to do something analogous to Singapore from the past generation in the financial system. They want to insert themselves into the global flow of tokens. That makes total sense. It's obvious to me why these countries would want to do it.

Yeah.

I also understand why the AI industry would want to do it, inasmuch as they are kind of deep-pocketed customers and ready to buy and all that good stuff. What's a little less obvious to me is what is the United States getting out of it? So one question is, what is the bottleneck that these countries are relieving for us? Is it that they can muster energy faster? Is it that they can just greenlight projects faster? Is it that they don't need these things to be profitable? And so they're willing to pay more than maybe American customers counterfactually would be willing to pay. Like, what's in it for the US? Why don't we just do it all here?

So one, they're required to do it, half of it here. There's a 1-to-1 match for all the infrastructure they build out there. They have to build out a 1-to-1 match here. So that's one. So we're getting a ton of direct foreign direct investment, ton of infrastructure. That's, that's just a baseline requirement for all deal-making there on infrastructure deals. The second is we get them to build on the American stack, not the Chinese stack. I mean, Huawei has been courting, you know, all— every one of these Middle East hyperscalers or sovereigns aggressively for the last 2 years. And I think— I, and I'm very glad that we get continued— a continued ally in the Middle East, or continued alliances. This is what we were talking about earlier. This is an extended version of our Marshall Plan. So What do we get? We get more folks in our camp than in China's. So, so I think we get, you know, we get a bunch of capital. We get more infrastructure funding in the U.S., which results in more chips and compute capacity for our companies here and our founders here. And we get more allies in the Middle East than allies with China. So I think we get a bunch of investment, we get a bunch of compute, and we get a bunch of geopolitical wins. So. Yeah, I think it was a sort of win-win in that situation. I think the hardest thing, or the thing we've got to watch out for, right, is there a defection of that alliance, right? If I was one of the regions in the Middle East, by basically doubling down on America as my ally, I'm running the risk of alienating China, right? How do I protect against America In a future where America's not, for whatever reason, not a reliable partner of mine. For the time being, they seem to think we are reliable partners, which is great. So I'm happy. But that wasn't clear, by the way, under the Biden administration. I think the Middle East was in a place where they were being denied a local AI ecosystem. They were being denied certainty of chips, and that was driving them into the arms of China. So I think things were getting pretty Dense where they were starting to have to build out, do, do sort of joint, you know, the, the thing about infrastructure is it's infrastructure is destiny. So you have to plan on 20, 30-year horizons, right? And so if you lose somebody, an ally in the first chapter, you essentially, that they're predetermining the outcome 20 years in advance. And there were moments in the Biden administration where we were basically driving the Middle East into the arms of China. and therefore giving up the 20, 30-year arc of the destiny of the infrastructure in that region to Huawei versus our guys. I think that's correct.

That's why I still come back to that question of like, just how ready is China to actually meet that demand? And I do agree with you that we should not underestimate and we should not also just take one static point in time. It does still feel like It's at least a few years away, though, right, before China would actually be exporting AI infrastructure?

Well, it starts—

ATNA for these Middle Eastern countries seems like it is a deal with China, not now, but like 3 to 5 years from now, if Huawei can scale to the point where there's surplus beyond what Alibaba and DeepSeek and all their own domestic major companies want to buy, which I assume is like a lot. So—

Yeah, totally. 3 to 5 years, but that determines the next 30 years. If you lose the first 3 to 5, you've lost the 30. That's the nature of infrastructure dealmaking, right? You lose the first mile, you've predetermined the outcome 30 years hence. So yeah, I kind of agree with you, but that is why the 3 to 5 years is so critical. It doesn't really matter when, in a sense, yes, but, you know, racking and stacking a Frontier data center with Ascend chips might take 3 to 5 years. But the point is that it's an Ascend, it's a Huawei data center. And then what happens to the buildout? What happens in with the next, you know, 500 megawatts of capacity and the next 500 megawatts of capacity? And that you have now essentially set the Middle East on a path down the tech tree. Where success is determined, predetermined over the 30-year arc in China's favor versus ours.

So aside from this competition with China angle, what would you say is like the next most compelling reason to do these deals? Like, is it because they can turn on energy super fast and we can't?

Or— Yes. So I do think they have on a per-flop basis, there's roughly an 18%, from what I can tell. I mean, some of the marketing would be like, oh, you You get 50% energy gains. But on liquid-cooled footprints, there's about an 18%, you know, once I handicap it for a bunch of other factors, I think you get an 18 to maybe 20% lower cost of energy per flop by locating a Blackwell node in the Middle East, in the UAE or the Kingdom, than you do, than locating it in the US. Unless without, without government subsidies here. So, yes, there is a lower energy footprint there, partly because a bunch of these data centers are liquid-cooled with oil. So, so, yes, there's a pure sort of energy reason to do it. Look, I also think there's a— one of the realities we have to contend with until recently with the new administration here is that the Kingdom of Saudi Arabia, the UAE, these are countries with their own form of governance where they don't have democratic elections. But they do have the ability to move quickly when it comes to infrastructure highways, right? They, they can get through regulatory red tape really fast. And now that those, a bunch of those regions are very, very aggressively pro-America, pro-modernization, they're up, they're changing dramatically the modern fabric of the society, right? Like if you, if you go to KSA, you'll often see You know, a complete change in cultural values relative to 5 years ago. The speed at which you can get frontier data centers up because of the lack of red tape relative to other regions across the world is kind of crazy. Whereas there are still parts of Europe where putting up a frontier data center can take like 2 years of getting through red tape and approvals, whereas in the UAE, that's just not the case. So, I'm just even, yes, there's a, there's a, there's a technical reason why these, the compute in these places is more energy efficient, but there's also, I think, a speed of execution thing here.

Yeah. What do you think as this develops, what is it going to look like in terms of personnel control and governance and like literally like military protection? So what I mean there is, okay, we know it's going to be in the UAE or it's going to be in Saudi Arabia, but who's going to build it? Are we going to have Americans over there doing the building? And does that create a vulnerability where for these countries, the US could at some point say everybody has to come home? Are there other governance things like I've heard, and this is an unverified rumour, but I've heard that in some part of some of these deals, There's an agreement on an off switch that is supposedly remotely controlled. I don't know if that's true or not, but that obviously would be an interesting sovereignty question. And then literally, like, missiles are flying right now, right? So we've had recent episodes of US aircraft carriers reportedly having to turn fast to avoid missiles to the point where they're dumping planes over. the side of the aircraft carrier, and data centres aren't so manoeuvrable. So I don't know, does this also sort of imply like a major expansion of the American sort of defence umbrella to protect these assets? Because if we have like gigawatt data centres in the desert in the Middle East somewhere, everyone's going to know where they are, or they're not going to be that hard to hit with a missile. And they're much harder to hit in the US than they would be over there. Right. So what deterrence do we have, and what sort of promises do you think we might be implicitly making? So anyway, there's personnel governance of the sort of off switch variety and the risk of missiles.

Yeah. So this is a good question. The framework you want to apply to this is dual use technology, right? Because AI is a dual use technology, it's often It's governed by a certain body of work that is separate from laws that govern, and defense capabilities and so on, that govern non-dual-use technology. So if you're talking about a Frontier data center that is classified for mission-critical workloads, actually the construction of these Workload, these data centers is classified information. It's confidential. It is quite hard actually to, one, for a number of reasons, like other sort of critical national infrastructure, the set of protocols that go with the setup of those data centers is different from traditional, like a traditional Google data center. So you can't just, it doesn't show up in regulatory filings. It often doesn't show up as— the construction protocols do not look like traditional data center construction protocols. And for better or for worse, right now, because the model training happens in a different place than the inference, the data centers running mission-critical inference workloads can be quite small. Like, they don't— you don't need 100,000 black holes in one place to run military workloads. Or national security workloads. So I, I want— I just want to say that from a, from a pure physical, like, brick-and-mortar perspective, they're not that easy to trace. Like, their footprint is different. They can often just be inside what looks like a department store on the outside. Then there's, uh, you know, let— there's the whole part of, like, counterintelligence and so on. Let's— so let's say there's counterintelligence from an adversarial nation that figures out where These workloads are running and they will be targeted. Then your fault— the analysis you want to do is basically what is the threat vector, right? And what is the frontier of warfare in that situation, which is if it's aerial, an aerial threat vector, then you're essentially running the UAV playbook, right? And these data centers, at least the ones that are civilian data centers in the kingdom and so on, are governed by their laws. and their military posture and their aerial playbooks. So I don't think the analysis is sort of AI-specific, is my point. That just applies to any kind of asset. And the risk of that particular asset being attacked is no greater or lower than the risk of any national security asset there. And it depends on basically how good is the government there at defending their best assets. So I don't think it requires a net new expansion. From— if you're talking about the physical footprint of these data centers, there's a whole other analysis we could talk about, which is the weights, right? And this is why open source is so effective, because if your entire national security apparatus is predicated on the idea that weights don't get stolen, then I don't know how to protect you.

Yeah, exactly.

That— yes, I think that remains, including in the United States, completely unsolved. Like, there have been You know, look, there have been Chinese citizens who have been caught boarding planes from the US with like TPU schematics. I think there was a Google engineer. I don't think there's an apparatus right now within the US frontier labs to prevent model weight exfiltration that is real. Like if our adversaries, if China wants the weights, they're getting the weights right now. Yeah, I think everybody. Yeah, I think that's at this point. Yeah. So I don't think we should. I don't think we should expect other countries to be able to do something even we can't do yet.

Yeah, all good. I mean, this has been. You've been very generous with your time, and I really appreciate it. Do you have any closing thoughts? Anything we didn't touch on that you want to share? I mean, the we've covered a lot of ground.

That's for sure. No, we should do a part two. I don't think we got to talk about Taiwan at all, which is a whole lot of can of worms. But I think that's a critical part of the story that. Depending on how, you know, Taiwanese sovereignty plays out over the next 2 or 3 years, the arc of how the rest of the compute infrastructure, the rest of sovereign infrastructure plays out will be quite different. So that's for, for next time.

Yeah. Talk about a, uh, a big domino to say the least.

Yeah.

Yeah.

That's one of those reasons that I'm still not sold on the export controls, just because I feel like if nothing else, it's clearly making it that much easier for some big move to be made on Taiwan as opposed to the scenario where China was getting the chips. But I've said that on this feed many times. So cool. Any other closing thoughts? I mean, we can definitely check in, you know, I feel like the cycle times are getting shorter and shorter. So, uh, I don't know if it's 6 months away or 3 months away when we'll have to come back and evaluate all of AI and geopolitics, but something tells me it won't be all that long.

But any other, um, No, that's exactly what you said. Everything we discussed today should be treated as subject to shorter and shorter time spans of relevancy. So see you Monday.

Sounds good. Ajit Nayak, general partner at a16z, thank you for being part of the Cognitive Revolution.

Thanks for having me.

If you're finding value in the show, we'd appreciate it if you'd take a moment to share it with friends, post online, Write a review on Apple Podcasts or Spotify, or just leave us a comment on YouTube. Of course, we always welcome your feedback, guest and topic suggestions, and sponsorship inquiries, either via our website, cognitiverevolution.ai, or by DMing me on your favorite social network. The Cognitive Revolution is part of the Turpentine Network, a network of podcasts where experts talk technology, business, economics, geopolitics, culture, and more, which is now a part of a16z. We're produced by AI Podcasting. If you're looking for podcast production help for everything from the moment you stop recording to the moment your audience starts listening, check them out and see my endorsement at aipodcast.ing. And finally, I encourage you to take a moment to check out our new and improved show notes, which were created automatically by Notion's AI Meeting Notes. AI Meeting Notes captures every detail and breaks down complex concepts so no idea gets lost. And because AI Meeting Notes lives right in Notion, everything you capture, whether that's meetings, podcasts, interviews, or conversations, lives exactly where you plan, build, and get things done. No switching, no slowdown. Check out Notion's AI Meeting Notes if you want perfect notes that write themselves. And head to the link in our show notes to try Notion's AI Meeting Notes free for 30 days.