SemiAnalysis Founder Dylan Patel on New AI Regulations, Chinese AI & xAI's Surge to Hyperscale

Well, if you want to maintain hegemony, it turns out like, what did we do during World War II? Well, we like, Henry Ford, here you go. You're going to be absolutely freaking rich because you're going to make all the tanks, right? This is sort of the same thing you're doing with AI is like, here you go, Satya Nadella. Like, you get to do this in AI, right?

Dylan Patel has become one of the go-to thinkers on all things hardware and AI with his writing over at SemiAnalysis. Today on Unsupervised Learning, I had an awesome time chatting with him on a host of different things. It was particularly fun because we talked right after the AI diffusion rule came out. So I got his live reactions to what it means for hyperscalers, what it means for China, the UAE and others, and what the future of geopolitics looks like given these regulations. We also talked a lot about some of the key questions in AI. What will happen with open source models? What are NVIDIA's long-term competitive moats? And we also talked about what's required to build really large clusters. We had some amazing stories from xAI's 100,000 GPU cluster.

xAI kind of came in and was like, well, we can't find a data center anywhere. What they did is like the most Elon awesome thing ever, right? Which is like—

I think folks will really enjoy this conversation. Without further ado, Here's Dylan. Dylan, thanks so much for coming on.

Yeah, thank you for having me. It's a very cozy, comfortable place.

I can't believe we got you like during AI diffusion rule week. I feel like this must just be madness right now.

Yeah, I mean, uh, the fun story was, uh, it was supposed to be originally Monday. Uh, they dropped on Sunday night/Monday, right? And so I was like, I was up all night working on it and had to cancel, and we, we shifted it back a few days. So thanks for being flexible.

And you shipped a pretty long piece in the interim.

Yeah, yeah, it was, it was great.

I'm curious if you just give our listeners some like Quick context on what's going on on the regulatory side with the AI diffusion rule, and then maybe given the you know goals of what they're trying to do, what kind of grade would you give it?

Yeah, so I think originally right the October 2022 regulations were primarily on the semiconductor industry, but the the wording there was like not minced at all, right? It was we want to regulate AI. We believe it's going to advance rapidly.

It's October 22, which is like pre scaling pill pre ChatGPT, but they were scaling pill, right?

Exactly. And and like you know if you get a chance to talk. Talk to some of the people like you know Ben Buchanan, who's like the special advisor in the White House. Like he's like, "Yeah, no, no, no. This is exactly why we did it. We knew it was coming." And it's like, "Oh, so the government is actually quite competent." And it's quite interesting because you know they they they hit really hard. They have the goal of like the U.S. must be ahead of China on AI, right? Because the next five years of progress, next three years, next two years of progress. Are going to shape the next, you know, century of like hegemony for the world. Right. And so this is like taking an axe to like try and stop Chinese progress. As far as the grading, right, October '22 was really well-intentioned from if you, if you have this worldview, right, that, that AI is, you know, in the next 5 years going to transform the world. You know, if you have the view that it's going to take 20 years, then it's a very different like story. And actually these regulations are actually pretty bad. because they're going to limit US competitiveness long term. But at least over the next 5 years, they definitely will keep the US further ahead. So October '22 is really well-intentioned, but there was quite a few loopholes. So they did another round in '23, then they did another round in December. And all of these have slowly patched over loopholes. There's still some major ones, right? Namely, Chinese companies can get GPUs from foreign firms, right? So if you've seen the fuss that Oracle's been throwing, They've just been like ranting about it. Well, it's like look at who their largest one of their largest cloud customers is. It's ByteDance, right? Or you know, hey, you can build a data center in Malaysia now. All of a sudden, it's fine, right? And so Malaysia from 2024 to 2027, not the country itself, but companies operating there mostly Chinese companies, many of them claiming they're now Singaporean companies, right? Like the largest operator in China of data centers in China, GDS. moved to Singapore and says they're Singaporean now. It's like, but now they're— these companies are building 3 gigawatts of data center capacity. To put that in context, at the beginning of 2024, that was roughly Meta's global footprint. Yeah, right. So it's like, oh wow, like the entire world's like social media is running off of, you know, what Malaysia is putting up in 3 years.

So what happens to all these like, you know, Malaysian data center builds given like, you know, obviously a big part of this regulation is both tiering countries and then also applying much more regulatory oversight of like who's actually training models within your data centers. Obviously, all this planning has gone into building these things, but now it seems like the rug's kind of been pulled out in some way.

Yeah, I mean, this is definitely the most far-reaching regulation that I've ever seen, right? Probably there were big— there were bigger things during the Cold War or World War II. I'm not sure. But like, this is very far-reaching in that like it's regulating clouds overseas, right? It's regulating foreign companies. It's, it's limiting what they can buy to a huge extent. And so Malaysia has a huge amount of data center capacity that's being built. We'll see if they continue building it, right? I think most people will still continue building it. Companies like Microsoft and Oracle have significant assets being built in Malaysia. They probably won't stop. But other companies, what happens now? At the end of the day, the US is still very limited on data center footprint as well. We're building as fast as we can. There's a lot of regulatory bottlenecks. There's a number of different things in that nature. But, you know, what happens as you deploy, you know, American companies deploy AI globally, right? It's like, is OpenAI going to serve GPT in Japan and Taiwan and South Korea? Or like, do they just run it out of Microsoft data centers in Malaysia? Maybe, right? Like, this could be a reasonable way to like subsume this. But like, the really important thing here and the sort of, you know, even if you're like very pro-America, anti-China, there's still huge negatives to this regulation in that like, a ton of buyers can't buy, you know, that data center capacity in Malaysia and actually deploy, right? Like American companies can't do that, right? And so, you know, Oracle is a good example, right? They're one of the parts of the rule is you can only have 7% of your data center capacity in a country, in any country that's not like a US ally.

Yeah, right.

These tier 1 countries. And Oracle was planning to have like 20% of their capacity in Malaysia. So now it's like, what do I do with these data centers that I was going to build? I'm obviously an American company. Fine, I won't rent it to Chinese companies, but like, what do I do now? So there is a bit of like, you know, the companies that have so much capacity in the US are the only ones that can now take over these data centers in Malaysia and build them up.

Right, because of the 7% rule.

Right, because of the 7% rule. So like, it's like Microsoft, Meta, Amazon, Google, right? These 4 companies have, you know, 70+% of their data center, AI data center capacity in the US. And so they can like deal, and they're building super fast in the US, they can deal with like, Hey, I'll take I'll take 500 megawatts. I'll take a gigawatt in Malaysia, and that won't break me out of the ruling. Right? Everyone else is kind of like SOL. So now you've like reduced competition massively in the market. Right? And you've created sort of a monopoly. And so this is sort of like why these regulations, while they are like yes, if your belief is like this whole AI like acceleration and like you know you need to combat China, this is very important. U.S. must maintain hegemony. That's one side of the fence. The other side is like. Well, if you want to maintain hegemony, it turns out like, what did we do during World War II? Well, we like, we're like, Henry Ford, here you go. You're going to be absolutely freaking rich because you're going to make all the tanks, right? Or whatever, right? Trucks. This is sort of the same thing they're doing with AI is like, here you go, Satya Nadella. Like, you get to do this in AI, right? And so there is a bit of like antitrust, like, you know, I like decentralized power. You like decentralized power. You like innovation. And does this like stamp out innovation in some ways. Uh, you know, probably not for American, like, startups, but for, like, infrastructure hardware, it definitely is, like, pretty, pretty bad.

Obviously, there's been, you know, like, uh, proliferation of players like CoreWeave and these folks. Like, how do these regulations— like, I mean, ultimately, you know, I've heard you say it obviously favors, you know, the, the giant tech companies, but presumably, like, they can have footprints in these tier 1 countries as well. Like, what happens to the kind of, like, you know, the mini cloud ecosystem?

So I think, I think CoreWeave has rapidly like become quite close to the hyperscale level, right? Like, and so, and then their, their next set of buildouts, like large buildouts are, you know, continued in the US, of course, but then also Europe. Um, I think they could expand overseas for a while and still be fine. They're gonna have to hit certain regulatory, like, you know, levels for, you know, being able to monitor their customers' workloads and all these, like, you know, just when you throw up— The more requirements you add, The less people who can do it.

Yeah.

Yeah, exactly. The rest of the cloud ecosystem though, right, excluding CoreWeave, like is very much hit hard, right? There's a lot of clouds that are in foreign countries that are trying to build up. There's sovereign AI, which is like, you know, a lot of startups in the Bay are serving sovereign AI firms in Malaysia and Singapore and India and the Middle East, whatever. That's their biggest customer. And it's like, these guys are heavily, heavily impacted by these regulations. There's a lot of, Cloud companies that are, I guess, tier 1 is the hyperscalers, tier 2 is CoreWeave and Oracle, and then below that is everyone is just really hit really badly. I think the startup ecosystem is going to have to learn, what do I do if potentially my big Middle East customer or my big Singaporean customer or my big Korean customer, whatever, maybe not Korea because Korea is allowed tier 1, but one of these countries, sovereign AI countries, are they— am I now regulated? This is actually potentially bad for startups in some way.

It also is diversity on the foundation model side, right? I guess on the sovereign AI point, maybe contextualize before this regulation what was happening both in kind of the data centre build-out world that's government-fuelled, as well as the kind of training core models for a specific country or geography.

It was basically the Wild West, right? The US had an executive order, a while back. It was basically just like, if you're building a model that's 1E26 FLOPS, which is very large, that's twice as large in terms of FLOPS as Llama 4.0 5B.

Yeah.

So presumably very few people have done that yet, but it's like a notification requirement. That's it. But that's in the US, right? So if you're in China and you're doing this, or if you're anywhere else, it's fine. Of course, there were not really data centre buildouts of the scale to hit that quite yet, but they were coming next year and the year after, right? Or 2025, 2026. These buildouts were coming. And so there was not much regulating foundation models, regulating other countries from building foundation models. Obviously, OpenAI and other companies were told, like, hey, I know the UAE is telling you they're going to build you a massive cluster. You're not allowed to do that. Right? There's obviously things like that tacitly under the table happening, but there were no, like, here are the regulations. So if you were a company in China, you would do whatever you want. And if you're a company elsewhere, you do whatever you want. And you could rent GPUs from wherever you wanted, and you could build whatever you want. And this smacks it down, right? You can't export model weights outside the US, outside of trusted clouds, which is, again, the hyperscalers, right?

Yeah.

Of foundation models that are really the biggest big ones, right? Think Llama 4 and GPT-5 and that kind of stuff. Or you can't— there's various things around protecting against synthetic data generation, which all the Chinese companies are doing a ton of synthetic data generation of GPT-4 so their models can be good. This is part of the dirty secret about why, yes, Alibaba and DeepSeek are really good, but they're also just generating a bunch of data from GPT-4 and using that to post-train partially. There's a lot of regulations around accessing clouds. And so a lot of clouds they didn't have a core, like, when you look at the big tech companies, they've sort of all paired off, right? Meta's themselves, Google's themselves, but then Microsoft OpenAI.

Amazon Anthropic.

Amazon Anthropic. Elon's world is xAI, Tesla. The lines are blurry, right? X. The lines are very blurry between these companies. But these are sort of the hyperscalers in the US. But in China, Alibaba was doing good in AI and they were building out. But besides that, there's a bunch of new players. There's Moonshot, there's DeepSeek, and there's all these other players that are popping up, but they don't have a hyperscale partner. ByteDance, of course, right? And ByteDance does, they do run a lot of their infrastructure, but a lot of it they rent. What do they do? Well, they decided, well, let me just like, not DeepSeek, but ByteDance is just like, oh, I'll just rent from any GPU cloud out there, right? So there's random clouds in Europe, there's random clouds all over Asia, right? And it's like their business case is like, you're gonna make, a decent profit, right? And you're gonna build GPUs and rent them to me, right? And I'm not— you're not gonna have any observability in my workload. I'm just gonna do whatever I want. Most of it is just serving TikTok, right? Let's be clear, it's not actually like illicit, but you know, the potential is there. And they're obviously working on language models and generative video models, and there's always the like fear that they're gonna manipulate our teenagers' minds and, you know, destroy our children.

Not after Sunday, I guess.

I mean, they're still gonna be able to do the latter thing, right? But, but they're heavily, heavily, heavily limited on you know, the size of clusters they can get now.

Yeah.

And, and who and what and the notification requirements and like the companies have to be able to observe their workload, which is like bad security. So it's, it's a really, really big blow to, of course, ByteDance and many Chinese AI players, but it's also a huge blow to all these random cloud companies whose business case was like, I'm gonna sell to ByteDance, right? Or I'm gonna sell to this Chinese company.

So where, where does Chinese AI go from here? And like, what, you know, do you see any kind of like, if they're looking around for loopholes, like what, you know, what is the path forward?

There is, there's one obvious loophole, which is, um, well, there's like strict caps, right? Like each country can only buy 50,000 GPUs for the next 4 years. And it's like, that's kind of nothing when Nvidia's making, you know, 6 million plus this year.

Totally.

Right? Um, it's like, huh? Um, so, so each country can only buy 50,000 GPUs, but then there's still loopholes, which is if you buy 1,700 or less, that doesn't count to the 50K. Um, so, so the obvious thing is you just spin up a bajillion, you know, shell companies and you just buy, you buy 1,700, you buy 1,600 GPUs and, and then you route 'em to China or whatever, right? So there are still some loopholes, but generally, this is much harder to do. I think more so is just like China has a ton of amazing engineers and they will have to innovate, right? Like DeepSeek is amazing at engineering, right? So it's like they're crushing companies with similar levels of compute, right? Because they're just engineering better, but they're not quite like OpenAI Anthropic level. Now they need to take this same Compute for labs continues to scale like crazy, and theirs cannot scale nearly as fast. So it's like we not only have to be better at engineering, we have to be way, way, way, way better at engineering.

With this new push to test-time compute, is that feasible with a huge compute deficiency?

Yeah, so I think there's some interesting aspects of test-time compute, right? It is way more compute-intensive than people think. People are like, oh, training is dead. And it's like, No, building a test-time compute-capable model requires tons of training.

You've written some really interesting data about this.

It's not just pre-training, it's post-training, right? So you have to generate a tonne of data, you have to throw most of it away, you have to verify the data to make sure it's actually accurate, right? The chain of thought or the reasoning chain. And then you have to train the model, right? And you have all these reward models. It's very complicated post-training stuff. And it's very compute intensive, actually. So you can make reasoning models today, right? Like O-1 and DeepSeek's R-1, I think it is. And Alibaba's reasoning model, I can't remember the name off the top of my head, but these models are like— the way you should think of it is scaling laws are log-log, right? So you can scale data, you can scale parameters, you can scale— and it's like, well, both of those are sort of like petering out because data's not scaling fast enough.

Yep.

And just infinitely growing parameters doesn't give you anything, right? Or it gives you diminishing returns. But we've scaled up this rung many times, right? The training runs that these frontier labs are doing are billions of dollars, right? And next year, probably tens of billions of dollars, right? It's scaling very fast. You can only get a logarithmic improvement on that front every so often, right? Whereas test-time compute, we're at the bottom rung of the ladder.

Right.

And it's like, oh, I can like climb up this ladder really rapidly because, oh, right now we're spending hundreds of thousands, millions, you know, tens of millions, hundreds of millions, billions. Right. And like the compute deficiency doesn't like matter until you get to the billions-plus scale.

Yeah.

Right. And so there are many rungs that they can, you know, out-engineer with the limited compute resources and sort of catch up. But there's also the flip side of like, well, Anthropic, OpenAI, DeepMind, etc. Right. XAI, all these guys are— and many startups, in fact. Right. Like Mira Startup and and so on and so forth, are all scaling up this ladder really quickly as well, as fast as they can. And so it's interesting to see what's going to happen. I think that that's on the training side. And then on the inference side, it's like if you actually do a query of O1 versus GPT-4o, it's like, yeah, it costs like $0.20 for a query and it costs like $6 for a query.

Right.

It's like a dramatic difference. Now, obviously the quality is better, and that $6 is still way cheaper than paying someone who makes way more than $6 for that work. But it is like, and it's scalable, right? Whereas human resources are less scalable. It's kind of like, this is a huge difference in cost. So that $6 is all inference. And it's like, well, you're limited on GPUs you get into China. You're limited to 50,000 in all these countries that are friendly to China or even neutral to China.

Yeah.

Right? So it's going to be very hard to get that scaling of compute on inference. Right. And you can throw as much money as you want on training, but to actually change the world, right, again, if you believe AI is going to change the world, then you have to spend a ton on inference. Right.

Totally.

Even if the margins are low. Right. Even though Anthropic's and OpenAI's gross margins on inference are like 75%, a lot of these other companies' gross margins on inference are like 10, 20%? Even if it's 10, 20% gross margins, to amortize the cost of a $5 billion model, okay, you need to do $10 billion of inference revenue. And it's like, okay, well then that's like $8 billion of hardware cost. And that hardware cost of $8 billion is actually depreciated over multiple years. So in that year, you bought like $30 billion of hardware. Congrats. And that's way too much. That exceeds the caps dramatically. So I think, you know, test-time compute, they'll be able to like deliver impressive results probably. The question is if they'll be able to scale it and sell it. And so that's like the issue. And like on the flip side, it's also like China, the cost of developing software in China is like way cheaper than in the US, right?

Yeah.

They just have way more talented engineers, software engineers. And so, you know, their development practices and all these things are different as well. But every company there, there's a lot of platform SaaS doesn't work in China, right? This is pretty well understood. In the VC world, it's been amazing for VCs is that platform SaaS type stuff has been awesome because you can sell to all these companies because you're developing it in one place and you're selling it to everyone, right? Whereas in China, it's like, okay, you're developing it in one place, but everyone can just develop it internally and it won't be maybe as good. It won't be best in class, but whatever, devs are cheap, right?

Which kind of seems like the future we're headed to with all this AI coding progress.

This is what I'd be curious about your take.

Yeah. No, I mean, I definitely think it's a huge risk for these SaaS providers, right? Um, you know, the ability— I mean, obviously Klarna and folks have talked about it very publicly, but I think the kind of moat that was spinning up these applications in the past is, uh, is certainly a lot weaker now. And I think, like, the ability also, you know, that combined with the ability to more easily migrate data out of whatever system of record you're in, um, and basically all the, like, painful stuff that probably stopped someone from ever really upgrading their software in the past is now just, like, a relatively simple thing to do with a bunch of, like, AI agents you throw it at, go away for, like, a few days, and there it is. Yeah, it's a real threat to us going forward. But a lot of countries have, I think, clearly from the beginning, have realised the importance of having this expertise and models in their own country. But if anything, I feel like this regulation makes incredibly clear— Yeah, I mean, like, if you look at UAE, this is exactly what's happened.

And I think it's been pretty well reported in the media. Like, at one point, G42, which is the UAE sort of champion of sovereign AI, funded by all sorts of, like, UAE money, and it is based in UAE. And they, they, they were at one point like threatened of being cut off. And they were like, the CEO was like a Chinese guy from China. And it was like CCP links potentially, blah, blah, blah, all this stuff. And then like, you know, it goes like all the media reporting on this goes silent for like 6 months. And then the next, you know, thing to come out from like whether it was Bloomberg or Reuters or Wall Street Journal or whatever was like, yeah, G42 and Microsoft are going to do a partnership. And it's like, huh? I wonder like what happened here. Like, is it purely organic or was there like some sort of like tacit like, well, if you're just going to run all your own infrastructure, you know, maybe there's something bad here. But, you know, now that we know Microsoft has partnered with you, you know, it's good. And like, you know, now when you look at it, it's like G42 and the US government are actually like talking all the time and they're like, like G42 will be a big cloud player and they're deploying GPUs in the US, even not just in, you know, the UAE, but in Europe and the US.

So it's like, through these hyperscalers, I guess you can enforce the American sphere of influence of, like, what, you know, both obviously, you know, oversight of what, you know, models are out there, and, but also, like, you know, ultimately, like, who gets out. You know, I think right now there's, what, some human rights terms involved in, like, the tiering of the companies, of the countries. And you could imagine, like, all sorts of ways that administrations use this stuff going forward.

Yeah, I mean, like, the Biden— the regulation itself does mention human rights. Obviously, the new administration is coming in. If you had me guess, I'd say, like, 95% probability the Trump administration just keeps these rules and maybe like, like obviously tweaks them to some extent. But the human rights thing, they're probably going to throw out and like, you know, and switch for like, you know, American economic sphere of influence or like industry or like whatever, whatever the priority of like, you know, purchasing energy and weapons from the US government. Right. Like these are the sort of things that might matter more for the Trump administration versus like human rights policies might matter more for the Biden administration. But at the end of the day, yeah, it's like a weapon for sort of the arsenal of democracy. This is like a regulation. So I don't know if it's like—

Totally. But I mean, taking a step back, I feel like you've talked before about, and I think it was maybe in the context of the chip bans, about this Goldilocks that you're trying to strike of preventing China from training their own cutting-edge models, but not making it so barren that everything gets built up within the country itself. And I wonder, back to, I guess, the grade question, does it feel like this hard-handed approach, is that the right way to further these policy objectives? Jacob here, taking a quick break to say that if you're enjoying this conversation with Dylan, we'd love if you could share it with a friend. We're always trying to get the word out about unsupervised learning, and we have an absolutely fantastic set of guests coming up these next months. Now back to Dylan. Does it feel like this hard-handed approach, is that the right way to further these policy objectives?

So I think October '22 and then the 2023 regulations, they were more like, we're going to try and limit you. But they didn't fully limit China building their own domestic supply chain and all these things, right? Because the intention is you can't build your own domestic supply chains and you can't access it, but you can access it at some tier, right? Now, with these most recent regulations, it's like no access at all to this stuff, right? You have order of magnitude less compute than any of the US labs, right? That's what we're trying to enforce on you. In which case, obviously there's going to be some leakage. And so maybe it's not a whole order of magnitude. Maybe it's like, 75% less compute, whatever, right? It's still a humongous amount. So now it's like China only has 2 options, which is like build your own supply chain or, you know, and if we limit that effectively, right? Like, so like Ben Thompson has this take and he's like, yeah, if you, if the regulatory regulations are so effective on preventing China from building their own semiconductor industry, or, you know, it's sort of like, or they like build their own semiconductor industry and it's like competitive, then like action on Taiwan is much more likely, right? Whereas like if they feel that they're still benefiting from Taiwan, right, they feel that they still can build it up and get access through Western sphere, right? This is like the sort of like geopolitical stuff I have no idea about, by the way, right? Talking out my butt. But like there's sort of the like, you know, the Goldilocks zone of like, yeah, you don't want to ban them too hard. You don't want to let them just have everything because then that like induces risk, right? But if there's like that Goldilocks zone of like you just can't, right? Like, you're kind of like, you're behind, but not so far behind. There's hope of catching up, then it's maybe fine.

Well, I guess switching gears to, you know, the cluster buildouts that we're seeing today, um, maybe just give some context to our listeners on, you know, obviously you're, you're the guru of, of tracking all this, like the size of cluster buildouts we have today, those being planned, and maybe just talk a little bit about kind of the biggest blockers that we have to larger clusters today in these tier 1 countries.

Yeah. So, I mean, when you take a step through sort of like GPT-4, 2020, 2022. 2022, sorry. Roughly a few thousand GPUs, A100s, and they trained this model, changed the world. Now, what's the next step up? You got to look at from 2 to 3 to 4. There's a couple orders of magnitude of compute increases. Now, with GPT-5, quote unquote, or Orion or whatever's coming out, they're working on it. This is a scale of Hey, well, we can't necessarily go for order of magnitude more GPUs because it's just too expensive and there's a lot of difficulties with building that out. So we go from 20,000 GPUs to 100,000, right? And so xAI has built 100,000, Meta's built 100,000. Both of them have actually built more than that in one spot to do training. OpenAI's plans are also, and Anthropic, so on and so forth. Anthropic has 400,000 Trainium server GPUs coming up this year. Or Trainium chips, not GPUs. But people are in the hundreds of thousands range for this year. But what's built today and what's training models today is on that order of magnitude of 100,000, right? And the GPU that's being used is the H100. So you have 5x the GPUs and you have 3x the performance per GPU, roughly. So then you're at like 15x more compute. And we'll see what comes of this, right? The thing is, You don't build a cluster or run a model and immediately get something great out of it, right? You have to do tons of experimentation, and training takes months, and post-training takes months, and safety takes months, right? It's not like snap of the finger, you get the model right now that you've built the cluster. And so there is a bit of a lag, right? And so this year we'll see models come out that are from that sort of, quote unquote, last generation cluster. But the next generation one is being built now. And what's this next generation one? It's like, Hundreds of thousands, right? And for context, right, like, you know, each GPU, right, once you include networking, building, all this sort of stuff, maybe is, you know, the GPU itself of H100 is like $24,000, but once you add everything else up, it's like $40,000, $45,000 per GPU all in of everything, right? And so the cost of like 100,000 is like, like a lot, right? It's like $5 billion for 100,000 GPUs.

Yeah.

Right? And so Elon's round was $6 billion and He built a 100,000 GPU cluster, right? Obviously there's other costs, right? They had to buy data, they had to do employees, and they had to convert a factory into a data center, right? And set up all this crazy shit that they did, right? Which is like, I'll talk about that in a second, but there's a lot of stuff that needs to be done to get a model trained. And now that Elon's built the cluster, it doesn't mean a model's going to come out I'd be surprised if they have a leading model even this quarter, right? Um, it's gonna take some time. And they're— and, and so like, what is, what is, you know, that's on the training front. Um, and, and these next generation clusters that are gonna be built this year, not $5 billion, they're, they're, you know, $15 billion, right? You know, and obviously you can't scale the money up, you know, like insane amounts, but 3x, right? And, and obviously—

In the context of that, Anthropic raising this like $2 billion at, you know, in, in this current VC round seems perplexing.

Well, so what's nice is, you know, Anthropic, OpenAI, Anthropic especially, can, you know, and OpenAI and, and, and, you know, some of these companies that are the startups, right? Uh, can, can raise a round and that money can only be enough for maybe a year, year and a half of GPU rental, right? And that's a rental.

Right.

They're not the one buying the GPUs, right? So Amazon might be the one spending, you know, $5, $10 billion, but they might for one year rent it for $3, $4, right? Um, and you know, when you look at the revenue, the credits they get from Amazon plus the round they raised, right? That's enough to pay for something like that, right? Now, the, the, and, and the Trainium server is not $10 billion because it's like a cheaper internal chip for Amazon. But like, you kind of get this like benefit if your cloud partner really loves you, right? And in the case of like OpenAI and Microsoft, kinda, but they were able to make a deal, strike a deal with Oracle, right?

Yeah.

Oracle's spending like $10 billion plus for them this year to build out data centers and GPUs, right? for OpenAI, not Microsoft, interestingly enough. Microsoft is doing stuff as well, but they're not charging $10 billion this year, right?

Of course.

They're charging a few billion dollars this year. So there is that benefit, right, that you don't need to necessarily— you might have signed the dotted line and Oracle may have taken the credit risk, right, of like, will OpenAI and this contract exist in 3 years? I mean, you and I believe yes, but this is something that you have to consider as a possibility, no. And so there is that risk. And the numbers that companies are raising are not enough to pay for the whole cluster. But that just means Anthropic's going to raise even more.

Of course.

9 months from now or a year from now, whenever they release their reasoning model. It's a constant fundraising thing, right? That's like Sam's strength is he's a god-tier fundraiser, right? And so that's one aspect of it. The other aspect is when you look at the hyperscalers, right? Meta, Google, Amazon, XAI, all these guys are building their own data centers. They're the ones delivering the CapEx. It's really hard to get the electrical infrastructure, the substation, build it all out, get it to work, get all the chips to be networked properly, deal with all the failed chips. Because when you're talking about this number of chips, actually quite a few are failed. A lot of them are silent failures. There's a lot of difficulties in getting a cluster to work. And so the big data center guys have known this, the hyperscalers, XAI kind of came in and was like, Elon's like, well, we're just going to do it. And it's like, well, we can't find a data center anywhere because all of them are already taken, right? Right. For the timeframes we need. So what do we do? Right. And like what they did is like just like the most like Elon awesome thing ever, right? Which is like they found a random appliance factory, closed appliance factory in Memphis, Tennessee, right? Located next to like a giant power plant, a water treatment facility, like a garbage dump, like Like, just like literally like what most people in SF would be like, this is the worst place on fucking earth, right? Like, not, not really, but like, you know, you talk to people at XAI that actually go to the data center, like, this place kind of sucks, right? Like, Memphis has really good wings, by the way. You know, like, there's good food in Memphis.

Great food across the board.

Yeah, great food in Memphis, but like wings, barbecue, et cetera. But like, you know, it's like, okay, we bought this appliance factory, but why did we buy it? Well, one, there's a gigawatt natural gas plant.

Yeah, that helps.

Right next door, right? 2, there's a natural gas line, a main that they tapped and they set up their own generation capacity on site. You know, 3, they're upgrading the substation, take more power from the grid, right? So they're like doing— they set up their own like fricking mobile generators to like create power, setting up Tesla battery packs. They're now planning— they've filed permits to build their own fucking power plant, right? Like on site, not just like mobile generators, but like a massive natural gas combine, you know? So they can do a million GPUs, or is what he claims, right? Or what he says he's gonna do, which, you know, we'll see. Um, I won't bet against Elon, right? There's, he's doing all these like things to like get the data center now, right? Because there's no power, you know, the substation wasn't ready, right? The power's right there, but the substation was all, was not, was too small and it was being upgraded too slowly because, you know, US power grid upgrades take a long time, right?

Cool.

Take at least 6 months a year. They had already thought about it, but it was only like 6 months a year. Like, okay, we'll just generate power on the site. Right? So they tap the natural gas line that's next door and they set up all these mobile generators, right? Oh, the power from these generators is pretty dirty. And GPU training is like, the moment I stop training, power just goes to zero. Or maybe if I'm doing a gradient update now, all my GPUs go idle and they're exchanging weights and it's like, oh shit, power went to zero. And then it's like, oh shit, power went to 100 megawatts. And this sort of stuff can blow up grids. And so Elon's like, oh yeah, we'll just throw a bunch of Tesla battery packs. They're doing all sorts of crazy things. Oh, how do we cool this facility? Because it wasn't ready to be able to cool this much amount of stuff. Oh, we'll water cool everything and then we'll rent a bunch of chillers, water chillers, like many of them restaurant grade, like fricking container things and just place 'em outside and like, cool. Right? It's like they're doing all this like crazy stuff and it's like way more complicated than this.

Yeah.

Right? And it works, right? It's like, it's like very impressive, um, how they deal with all the problems. Uh, I guess one more funny story is like this power problem I mentioned. Meta, they accidentally open sourced this code. Um, it's literally called power plant no blowup. It's a flag. Um, and so when you're doing the gradient update, right?

Yeah.

When you're like exchanging weights, Instead of having the GPUs go idle, basically when you're sending all the weights in the network, instead they just have them do fake matrix multiplication.

Just so it's not sitting there.

So the power stays stable, right? Because otherwise the power plant might blow up if you go like this. And it's like, what the hell? So this is the crazy stuff that's happening. And it's like, there's so many complicated things going on in these $10 billion buildouts that people are doing.

Oh, fascinating. If you were running the government right now, I mean, obviously it seems like the energy side is a massive policy and just like bureaucratic blocker to more of these data centers happening. Like, what's the best thing that government can be doing to unblock this stuff? And, you know, is every future data center build-out like this going to have to hack together Elon style?

You know, uh, I mean, I mean, like, to be clear, right? Like the, the, like number of like normally built, conventionally built data centers is like this. It's just like, what we want is like straight line up, right? Like, you know, um, and so, you know, for example, uh, gas generators are sold out for 2, 4 years from GE.

Right.

Like, it's like, oh, wow. Vernova, I guess now, right? Substation equipment, same thing, right? It's like sold out for 4 years. So like now you have to get like creative, right? And so like even though it's like growing really rapidly, so like some of the stuff is like held up by like, I think in one part of like the Midwest, it might be Ohio or Indiana, the, the cost, like if you build a new power plant to transport that power on the grid to a customer costs more than the cost of actually generating the power. It's like, what the flip? Right? So the grid needs huge investments, right? There needs to be a removal of a lot of the environmental regulations. But also, to an extent, there needs to be more vibe shift of ESG-type stuff, right? And this is the attitude of people at AI labs. I don't know if it's actually true or it's the correct one, but maybe you just say, screw it, to build AGI faster, we do it with natural gas. And then because we're doing it with natural gas, and then AGI will like create enough economic wealth and prosperity that we can just do carbon sequestration and it will all be good, right? It's like this is a pretty like wild take, but this is like this is what some people believe. And and to some extent, Meta has kind of you know either for either you know they laid off their whole like DEI thing. They they uncensored politics. There's been a massive vibe shift. One of the things they're doing is like they're setting up two gigawatts. Right. I told you earlier, 4— sorry, 3 was like their total global capacity beginning in '24. They're setting up 2 gigawatts in Louisiana alone, basically all powered by natural gas. And it's like, yeah, now they can get in 2 years, right? And they've like figured out how to get the substations and all that, whatever. They've gotten all the supply chain stuff and it's like, yeah, if we just throw ESG out the window, can go faster, right? And like you're doing it in states that don't have as much environmental stuff, Texas and Louisiana and so on and so forth.

Right.

Many other states all over the country, but it's like, that's part of it, right? Now to this moment, right? Google and Amazon are still quite committed to their green pledges. Microsoft is sort of in the middle, and Meta and xAI obviously don't give a fuck, right? So it's like, or they don't, not that they don't give a fuck, right? Obviously xAI, Tesla does way more for green than what they're destroying, right? So it's like all sorts of, it's complicated, but is there a way to change the vibe on those 2 and therefore they move faster, right? Or is there a way to Make it so you don't have to do crazy stuff like grids are upgraded, power generation is, is, you know, like onsite power generation, like all these different things like help build solar and wind and batteries faster and cheaper in addition to doing all the gas, right? There's a lot of complicated things to do on the energy side, and there's ways to still make it green and fast, right? Like you could just throw up a lot of solar, you could throw up a lot of wind. And, you know, the power generation of like renewables are like this, but wind and solar do not— don't actually like correlate that well. So then you're like, most of the times you have power, obviously you have excess power sometimes, but most of the times you have power and then you throw on like a small amount of battery and now it's like almost all the time you have power, but there's still like times where the sun isn't shining and the wind isn't blowing for a long enough period that your batteries are drained. So now you have some backup gas, right? And it's like, these are ways to like make it greener and still do it fast. Yeah, I think, I think like grid and industrial policy around that is like the biggest like thing that needs to be done to accelerate this.

Kind of like the domestic version of now that we've done the, the international regulations, it's the next, next area of focus.

You would think, right? It's like, like I myself have like far more of like an attitude of like we should win harder rather than we should like try to make them lose. Yeah, right. Like that's sort of like I think most people think that's a better ideal and what America stands for more. But we kind of like— yeah.

So we have these 100,000 GPU clusters, obviously, now where we're going to see the output of these models at some point in 2025. And I assume that will obviously impact a ton of the future investments based on what we see there. But obviously people have to invest way ahead of that. And so we already kind of know what we're getting in for buildouts in this next year. As you kind of think, just kind of putting your betting hat on for 2 years from now, 5 years from now, like how big clusters do you think we'll have?

2 years from now?

Yeah.

Um, I would, I would imagine it'd be like, you know, like there's a couple things, right? Like, you know, you could say it by number of chips, but it's not fair, right? A100s were like 400 watts. H100s are 700 watts. Blackwell is 1,200 watts, right? You know, sort of the power per chip is scaling rapidly. So saying the number of chips is not fair. Yeah.

I guess energy.

Right. So energy is like a good metric because it's like how much useful work you get per energy. Is like growing rapidly, right? But then like the amount of energy you're devoting to this, right? So, so 150 megawatts is these 100K clusters, right? We're seeing like, you know, 400, 500 megawatts clusters sort of being built out now, right? Like, you know, and then, you know, in 2026, I'd imagine it'll be like gigawatt-scale clusters, right? Or that's what it looks like based on, or, you know, 1 gigawatt in 2026-ish. And Meta's like trying to do like 2 gigawatts by early to mid-'27. So it's like, and others are all like sort of doing similar stuff in different parts of the nation. It's like these buildouts are rapidly like sort of escalating to like earth-shattering amounts of like power in one spot, right? So, you know, and then like obviously the like dollars per chip and then the chip's power, all this, these are variables that change over time. But yeah, the buildouts are scaling, right? When you go from like 20,000 GPUs, 400 watts each. That was on the order of 20 megawatts to 2 gigawatts from 2022 to 2026, '27. It's like, holy crap, you just scaled that much power. Totally. In 2 orders of magnitude in like 5 years.

So if we're sitting here at the end of '27, how much money has OpenAI raised?

I think it'd be really difficult to say. I really do think it would be north of $100 billion. Right? That's like, yeah, it's pretty clear that they've been trying to convince all these sovereigns to invest and they've been passive on getting them. But I think you're going to see O3 release, you'll probably see another reasoning model release, you might see GPT-5 release. It's like, how are they not going to raise a ton, ton more money at a ton higher valuation? Revenue skyrockets, right? It's pretty— I think that's what's going to happen. Yeah. And I think the similar stuff, obviously OpenAI will lead the pack, but others will also get similar trajectories happening, like Anthropic, xAI. I think the amount of capital that will be deployed is going to be insane.

On the model side, obviously we've seen this massive scaling of test-time compute. It seems to work really well with easily verified data, coding, math. I think there's this big question of what other tasks will fall into these test-time compute models. What's your thinking on that?

So I think, I think like, I really like, like when OpenAI released the like 5 levels thing, I felt a lot of people were like puzzled, like what the fuck is, you know, it's like reasoning and then, and then agents, like, 'cause everyone's been, you know, in the startup ecosystems, like chatbots, agents, right? Um, and I think that it's like, I think it makes a lot of sense because test-time compute, like you're, you're like verifying your outputs and you're making sure they're accurate, right? So I think, I think like, and these, these like test-time compute methods, right, of synthetic data generation. and then throwing away a lot of it and verifying it and using it. This is where other modalities can come in. So computer use is kind of predicated on all these synthetic data generation pipelines and test-time compute working. Agents are all predicated on the reliability of a model getting high enough to where you can actually get the answer reliably and chain multiple together. And it's like these different paradigms, I think, will stack together. It's like you can't do computer use without test-time compute. You can't do agents without test-time compute. And when you combine test-time computer use agents, it's like, oh, wow, now it's autonomously working on the computer without even having to manually code in all the APIs that it's accessing and all this. Right? I think what tasks it can do will branch out drastically once you get the core of reasoning sort of not solved, but at a really strong point. and the pipelines for doing all this stuff, right? And so I think that's probably— now, what kind of use cases that is, is like, I think it's like whatever SaaS business model, of course coding is great, but it's like, what about software engineering, right?

Yeah.

Because coding is very different than software engineering. It's like customer service, great, chatbots do that kind of, but not great, right? Oh, searching through documentation, right? RAG is okay, but actually you need to RAG is really bad because there's no way to verify that you properly pulled in the right stuff, right? Your retrieval actually worked. People can be like, oh, our retrieval is really accurate, but it's still not accurate enough for an agent. And so when you stack these things together, I think it's really any information task. I don't see why it wouldn't work. Now, obviously, what gets done in '25, '26 is very different than what might get done in 2030. I don't know if I'm the best to say, what specific use case, though, right? I think I can just see sort of the writing on the wall is clear.

What about the open source side? I mean, obviously open source models have been relatively quick to catch up to the GPT-4o type level models. I think there is a big question with these test-time compute models, even in the regulations themselves, right? I mean, I think the way they framed it is like open source models, all fair game, and there's much more— whatever the open source model's capabilities are up till then is fine. And then basically It kind of almost assumes there's going to be some massive gap between closed and open models that just increases over the next few years. Is that how you see things too?

I think so. I agree. I think Meta will release Llama 4, and it'll be better than GPT-4. Obviously, you can say open source models have closed the gap, but it's like, does anyone have voice mode? Does anyone have quite the level of inference cost that Anthropic and OpenAI do? No, not really, because their models are are really efficient. O4o is much smaller than LLaMA 405B in terms of active parameters, way smaller. So there are still open models that are behind, maybe not in capabilities entirely, but in some ways. I think what actually ends up happening is that reasoning models are not necessarily going to be fully open sourced. Even like you see the DeepSeek model, you see OpenAI's model, Meta will release better models as long as they think it it makes sense to, but will they only release chat and completion models? Or internally they have their own code model that is trained on their own code database in addition to all the stuff in Llama. And it's like, well, that's not being released and that one's far more capable, right? At least at coding, right? So there's things like this that I think it might point to, like, okay, Meta's not going to just open source everything.

Right.

I don't think that necessarily means that you're putting together the reasoning system, right? Or the agent system, right? It might mean, here's the base weights. Congrats, we didn't do any of the reasoning training. Or like, here's the instruct model. Or maybe we even do some reasoning, right? But it's like, I guarantee you, if Meta has the best model in the world, they will not open source it. I would bet my whole wallet on that.

I guess, can it feel like, hey, they're 6, 9 months behind with a reasoning model? Will there be an open source equivalent to O3 in 6, 9 months?

Yeah, I mean, that's a great question. I'm not exactly sure. I don't think so, actually. I think O3, especially depending on what level of compute you throw at it, if you looked at the ARC Challenge, seems unlikely to me. But at the same time, Google's already got a reasoning model. Anthropic allegedly has one internally that's really good, better than O3 even. But we'll see when they eventually release it. It's like, what actually ends up happening is going to be very—

It feels like the whole question is how much CapEx is Meta willing to spend to you know, to keep up with these things.

I mean, I think, I think they'll keep up on compute. It's like the science is hard too, right? It's not just like, you know, you know, find my shit.

Well, at some level of compute, I mean, the numbers you're counting, if OpenAI by the end of '27 has raised $100 billion, um, that's a lot of money to put toward, uh, free open source models.

Yeah. I mean, like, but like most of, like, to be clear, most of Meta's GPU purchases are for like recommendation systems on Instagram Reels.

Yeah, other than that, obviously.

Right? Like, it's like, you can say like, Zuck can be like, oh yeah, I bought this many GPUs on like some stream or whatever. But then, When you dig into it, it's like, well, more than half of those are for recommendation systems. Chill. You're not spending $30 billion on LLM infra or transformers. You're doing this on all your products, right? Yeah, I just think Meta won't open source everything they do, right?

Yeah.

And I think that Llama 4 will be open sourced, but the moment that they have something at such a capability level, I don't see why they would open source it because they get benefit from— they get to attract a lot of talent. They get to have the whole community tell them what they're doing wrong with their models, right? You look at Llama 2, Llama 3, people are like, oh yeah, you're doing your RL wrong these ways, right? You're lobotomizing the model in these ways. And people are like— and Meta is learning from that, right? And they're attracting all this talent that if they have the best model, they're not going to do that. Or if they're like 6 to 9 months behind, Maybe they're 6 to 9 months behind on capabilities, but then the cost is way higher for inference. Or maybe cost is the same on inference, but their capabilities are worse. And it's like, if you look at any specific benchmark, it's like, oh, 5 percentage points, great. But it's like those benchmarks that are in the 70s and 80s are irrelevant. It's about like, oh, you're like 10% on this benchmark versus 50%. And when you look at SWE-Bench and things like that, it's like—

Oh, deltas are getting really big.

Yeah, the deltas are really large, right? Or like ARC, right? It's like, oh yeah, there's only like 2 models that even score any reasonable amount, right? It's like, these are the sort of things that I think matter. It's like, can you do a task or not? Versus like, oh yeah, it's 5% better at this task.

As you think about the broader, all the kind of big questions you're thinking about, I'm wondering if there's a set of 2 to 3 questions that are like most— if you could just zoom forward 2, 3 years and get the answer to them, I assume one of them will be the performance of these models on the 100,000 GPU clusters. But what else do you feel like are the big questions that if you could flip, if you had a time machine, go forward 2, 3 years, you'd most want to know the answer to?

I think a lot of it would be how they do stuff in the model level, right? Because there are a lot of outstanding questions on, with model development, how that changes hardware development, how that changes networking development, how that changes data center buildouts. Do you need all the data centers to be right next to each other? Do you need them to be Can they be geographically spread? What level of bandwidth do you need between them? What model research stuff comes out? If a year ago you explained reasoning and test-time compute and how that works to me, I think that would've been huge.

Right.

But I didn't know that was even the question to ask. Okay, I knew synthetic data was a thing, but I didn't think it was that huge of a deal. It was a big deal, but that was more for distilling stuff. But it's not. There's a lot of other use cases. So it's like, I don't even know what question on the model side is the important one. What really gets agents to work reliably and well? What is the trick that gets computers use models? Because Anthropic's computer use model today sucks, right? But I mean, it at least is something, right? No one else has something. But now that you have something, it's clear to see how you can climb up and be like, oh, this is good. So what are the tricks that people figure out and how does that impact, right? What is the average, when you're doing inference, what is the average sequence length now, right? What's reasoning or what's test time compute in terms of search look like, right? These are questions that I don't even know what the answer would be. And I don't know, I guess I'm definitely not smart enough to know what question to ask in 2 years, besides, what is the secret? What is the biggest development? Right? It's like, ah.

Well, with this development, I guess, of more compute being used to create synthetic data and inference, that obviously has massive implications on the data center hardware side. It feels like we might be able to get to potentially more distributed—

Well, for example, right, when you're doing synthetic data generation, you're verifying which reasoning chains that you've generated are good, you're throwing away a bunch of them, and then you're training on them, and you're grading them with all these reward models, and you're training on them. How much of this is online versus offline, right? Can you fully separate the synthetic data generation from the post-training RL with all the reward models? No, you can't, because you need to update the model that's generating the reasoning data constantly.

It all needs to be in the same place.

But then how often do you need to update it? Do you need to update it every 30 seconds, right? If you look at training of LLaMA, I think the gradients got updated every 15 seconds, right? And it took a few seconds, and then you're training for 10 seconds or 15 seconds or whatever, right? And it's like, oh, okay, you're spending this much time in network versus compute. And think about how you need to design your network now. But what is that ratio? What does that mean on reasoning when you're generating that data? How much time are you generating data and then doing post-training, and then before you update the model and send it back to the data generation nodes to do it again and grading their outputs? This stuff, it's kind of unknown to me and probably unknown to anyone besides OpenAI and Anthropic and SSI.

Yeah.

Like, what is like actually the future and what is, what matters there? What can you, like, how would you design a cluster to do, to accommodate that? Or how would you design your network? Or can you be not co-located, right? Maybe, maybe just really, really good fiber layouts between these 2 data centers is good enough. Or maybe they do need to be co-located with super high bandwidth, right? Like these are sort of questions that like, I don't know, right? Yeah, the paradigm exists, but like there's so many details. Totally.

No, I mean, I think it, and it obviously has implications. I feel like you must get asked all the time. And I think you've talked about very publicly on other podcasts, I'm sure people are like, well, Nvidia, paint the likeliest picture where they lose, or what are the actual long-term moats here? And I know you've talked about obviously the hardware, software, networking side of things. I'm sure you've seen a bunch of hardware startups and folks going after different parts of it. What of that is most interesting to you?

I think there's a couple classes, right? There's the old AI hardware startups, right? Cerebras, Groq, Amanova, TensTorrent, Right? Quote unquote old, right? They've been around for a while, and then there's a new age of them, right? Like there's there's like you know Etched and Positron and Maddox, and you know there's a number of other ones. There's a couple that are like some pretty high profile ones that are I think still under stealth, so I won't say anything. There's quite a few AI hardware startups, and their approaches are quite different. The way they think is very different. What they're focusing on is very different. Like each each hardware company has quote unquote a gimmick, right? Because like you're just not going to beat Nvidia by engineering better than them on all axes and doing stuff the same way. So you have to do things quite differently. Meanwhile, Nvidia every generation is making large enough changes to the architecture. So it's not necessarily the same, but they're making large enough changes to their infrastructure. And so the question is whether or not What are the quote-unquote gimmicks people are doing? How are they solving them? And do those give them— does that differentiate them enough? And then at the same time, you have the big problem of all the models are being developed on Nvidia hardware, cognizant of what's going to happen with the drawbacks and benefits of what Nvidia hardware does and what's the next generation one going to do. And so all the model research ideas, like, hey, this research idea would work, but it's If it runs really inefficiently on GPUs, I'm not going to pursue it because I don't actually care about how many operations I did. I care about how much time it took. And even though that is theoretically 10x less operations, but it took 10x longer because it's just an algorithm that doesn't work well on GPUs, and all of a sudden it's like, this is not worthwhile pursuing.

Yeah. So the entire research landscape is—

Yeah. So Nvidia's dominance influences where research goes.

Right.

And so you have to be different, but you can't be too different because then you're like, well, then models didn't develop the way I wanted them to and now I'm SOL. Could you have a model that's extremely far fewer parameters but needs way more bandwidth and could run entirely on chip potentially? But how are you going to train that model? And it's like, well, I'm going to train it with Nvidia GPUs, which doesn't make sense because I can do something better. And how am I going to deploy it? Well, I'd rather have it be able to be deployed on all the GPUs I've already bought. So it's like you've got the chicken and the egg kind of problem too. I think some people are explicitly targeting inference and training, and that'll be interesting because when you talk to Nvidia, you talk to the chip teams at various hyperscalers like the TPU or Trainium or et cetera, one of the labs has an AI chip team, all these companies, All these people believe that training chips are inference chips, which is very funny. And I think it's because they see the flexibility, they see the workload changing. So does an inference chip, what does it materially mean? Because there are a few of those in works, right? By the time you get your chip out, maybe you projected a 5x cost difference or 10x cost difference versus Hopper.

They've already captured it.

But now Blackwell, Nvidia's pitching 10 to 15x improvement in cost. It's like, Well, they're massaging the numbers, marketing. Actually, I think they said 30x in their GTC, but massaging the numbers, 10x, 15x even, could be reasonable or not. Depends on the workload. Standard chat inference, it's probably not that, but for medium-sized and small models. But for really, really large models or reasoning models, I do think it's going to be like 10, 15x. So it's like, oh, wow, okay, so you've captured, you've narrowed the delta massively. So that's like a big, uh, that's a big like question also is like, so, so what is interesting is like, yeah, a lot of these companies are taking approaches that are super cool, right? Like, and the 3 that I mentioned, the new age ones like Maddox, Positron, and Etched are taking approaches that are super cool, right? And could work out, uh, or models could develop in a way that isn't good for them, right? Um, so that's, that's the big question, right?

I think you've called Trainium like the Amazon Basics of chips. And so, uh, you're obviously the hardware guru. But like, one thing I've just been curious about is like, to what extent is Anthropic's use of them like a sacrifice versus, hey, it doesn't really matter.

Um, yeah, so it's the AmazonBasics TPU because like its topology of networking looks exactly like a TPU. And like a lot of different design choices they have were like, this is just like TPU stuff, right? Um, obviously they're doing a lot of their own independent stuff. There's a lot of unique things, um, that they've implemented, like stochastic rounding and all these other things. But like, it's like, these are unique things, but it is the AmazonBasics TPU from Google, right? Google TPU. And so because it's worse, but it's cheaper. And yeah, now Anthropic, they barely use Trainium today, right?

Yeah.

They barely use Trainium in Inferentia. They've primarily used Hopper. Hopper at Google, Hopper at Amazon. They did use some TPU at Google, of course. But now this year, it's very clear they've bet all of their eggs into the basket of Trainium. How much of this is forced, right? Like, you know, how do I do— did I need to do this to get Amazon to invest in me and like make me their partner? Partially, yeah, absolutely, absolutely, right? But did I do this out of my own volition because it actually made sense? I think partially, yeah, because like the Amazon Basics TPU gives you, on some metrics, it gives you the best performance per dollar, right? And those metrics are like memory bandwidth and memory capacity. per dollar. It actually has better than any other chip in the market, including TPUs or GPUs, right? Because it's like some of the supply chain choices that Amazon chose, right?

Yeah.

And things around that, right? So it's like, what— I don't think it's as simple as Anthropic was forced to. I think they had to weigh the benefits, right? They're like, we could partner really closely with Amazon. We could raise more money from them and do this, or we could raise a little less money and and try and get GPUs from them. They had to partner with someone to do the CapEx, of course, because the whole rental versus CapEx of the cluster. So they definitely had to deal with that. And then they had to have the distribution channels of Amazon. They needed those things. But I'm sure they could have struck a smaller deal with Amazon and gotten GPUs if they really wanted to. Now, was that better or was it better to take more money, go Trainium, go all in, Do try and emphasise your future model architecture on the benefits that Trainium offers you versus GPUs, because there are some. Majority, no, they're worse, just straight up worse, but some benefits exist. So that's the sort of question that— it's a good question. And I think mostly it's hand-wringing and a bit of like, yeah, we can engineer our way out of this.

Does it make sense for OpenAI to build its own chip?

I think it does. It doesn't make sense if you think their growth is linear, it's going to go from— what do you think OpenAI's revenue this year is going to be? If it's less than $10 billion and the year after that less than $15 billion, should not make a chip. But if you think their revenue this year is going to be north of $10 billion, by $10 billion I mean run rate, because we're at a VC, right? We don't think about things in— no, run rate is more important. I think it's just simply more important.

Whatever the higher one is, we'll take it.

Yeah, exactly. Exactly. I mean, it's more indicative of what the current state of the business is, right? So if they go well north of $10 billion on run rate, and the year after that, what are they at? North of $20 billion? The costs start to scale rapidly, and it makes a tonne of sense to make their own chip, right? And so just because they have the chip team doesn't mean that they're not collaborating super closely with Nvidia and working with them on new architecture, right? There's a wall in OpenAI. The people working on the chip do not get to work with Nvidia on what their next generation architecture is. Right.

Well, and I think Nvidia is no stranger to this, like, coopetition. I mean, all of its major customers.

Exactly. Exactly. Like, Meta's doing it, Google's doing it, Amazon's doing it. Right. It's like they're used to it, and their attitude is very much like, good luck, we're going to crush you.

Right.

Like, we're just better. Right. And so far that's worked. Right. And I think their attitude to OpenAI would be the same. And OpenAI's attitude is always like, well, how much do— and it's the same as the hyperscalers, right? Well, if we spend a few hundred million dollars, billion dollars a year on this, it gives us a call option. Also, we can always just use it as a negotiating chip, right? If I spend a billion dollars on my own chips in '26, '27, but then the alternative is— and I'm not really deploying them in huge scales, I'm mostly still buying Nvidia chips, but if I got a 10% discount on Nvidia chips, it paid for itself.

Yeah.

And it's like, does, does Trainium help Amazon get a 10% discount or a 5% discount on $30 billion of GPU purchases? Like, okay, great. It's like worth it. Right. And it doesn't even like, you could like just not do anything with it and it would be worth it. Right. Like obviously it's a game of chicken. Like Jensen like knows how much you're making and all these things. But yeah, I think, I think that's the big, um, question.

Yeah.

And I think, I think from my perspective, it absolutely makes sense, even if it's not successful, to do this. And if it's successful, it's humongous. Yeah.

There's this whole crop of companies too that popped up, like, you know, the Fireworks and Togethers of the world, you know, that are, they're obviously just, you know, focused on inference optimization today and places to run a bunch of models. What happens to that market long-term?

I think that enterprise AI is, is actually got really decent legs in some regards because of, again, all the unique data that enterprises have and unique use cases. And enterprises like to think their data is good, but now most of the times it's pretty dirty and garbage, just to be clear. But now with the synthetic data pipelines, reasoning, all this sort of stuff, people can actually figure out how to generate a bunch of data that's specific to a business, and verify whether it's good or not. And then they can actually improve the models and they can do reasoning on a much smaller scale for enterprises. I think this works hugely, right? This is massively good, right? Now, obviously, Fireworks and Together, I think, are like— there's like 20 inference API providers out there that are just taking open source models and serving them. You obviously named the best 2 in class, right? 2 companies that build their own inference engine, are very efficient, much more efficient than VLLM or TensorRT LLM, all of these sort of things. Have good go-to-market, all these sort of things. These companies are good. I think they will be good because they're going to continue to be, A, Meta is going to open source models, B, China will continue to open source models, C, enterprises will be able to work with people to make better reasoning type models. Right now, Together helps people train models, right? And so does Databricks and all these companies help train models. And there's many other sort of services. Or people can partner with a startup to help them make a reasoning chain.

It's so interesting because I feel like the whole Mosaic pitch was like, train your own models. That went out of fashion for a bit because the general purpose models were just getting better at everything way faster than a Bloomberg GPT was. But now with these reasoning chains and the need to kind of verify this, you actually might be back in that world.

Yeah, I think Mosaic really lost its kind of— and Databricks kind of lost its footing for a bit. It'll be really interesting to see if they can get RL and reasoning to work, because I think that And I'm sure Ollie at Databricks kind of knows this is the thing that would make training for your customers make sense again. And maybe not pre-training, right? Because that's what Mosaic was doing. Mosaic was doing really, really good pre-training. They had the best model in the world twice, 2 different instances, open source model for short periods of time, but they had the best model in the world for short periods of time twice, once with MPT and then once with DBRX. But they've kind of fallen off because it's like, well, I can't really compete and spend the CapEx that Meta is, or even Alibaba. But now again, it's like I take the open source pre-trained model, I generate all these pipelines for reasoning, synthetic data, and my customer's data. Again, it kind of sucks, but now I can verify it in some way, figure out the recipe, and actually apply it to everyone's business use cases. That is potentially— and these people need to serve models, right? So it's not just Databricks Mosaic ML that's going to do this, but everyone's going to do this. There's going to be startups that do this as a service, like Adaptive ML. They're going to do this as a service, and they've actually been doing it as a service for a year now, which is really cool.

Yeah.

These people can partner with enterprises. They can do white glove services. They can do consulting services. Whatever it is, help them build models that are uniquely unique use cases for this stuff, right? And then those will need to be deployed, right? Or companies will want to do it on their own, right? And there's always going to be security stuff. So I think Together and Fireworks will be the partner of choice for inference of these models, right?

Mm-hmm.

Even if it's not necessarily on their GPUs, it's like, let me do a virtual private cloud on Amazon with your software running the model, right? Because you've built all these variable batching and all these various things, and it's way more efficient. Right. Than like the public software.

I'm gonna ask like a very basic question about CoreWeave, uh, just, uh, for, for my own edification. Um, and that's just like, what made them so successful? I mean, from the outside, sometimes it feels like there was this decision from Jensen to allocate GPUs among a bunch of players. They were kind of rewarded in that. Um, and I think it kind of goes to, and you're an expert on this, like the extent to which some of these mini clouds can differentiate beyond like just access to the GPUs, which it felt like in the first bit was, was one of the big things.

Yeah. So I think like part of it is, I think there's 3 factors, right? One is the allocation bit, right? Like Jensen is no idiot. His customers are all trying to build chips. They also rent most of the chips to other people. You want to have more competition, right?

Yeah.

So let me spark competition. Let me do a small investment in like 4 different clouds. I think they've invested in 4 different clouds so far, and they haven't been large investments. They've been pretty small, but like it's sort of given other people confidence like, oh yeah, they're gonna get allocation, right? And then those GPUs are allocated there. World's so tight in GPUs, They're able to rent them for a lot of money, make money, and get the engine going on a cloud, right? So CoreWeave now has like 200K+ GPUs, right?

Yeah.

A lot of GPUs, right? They're doing billions of dollars of revenue, right? It's not mundane business now, right? And they're worth whatever, like $20 billion or something, I think on the last round, whatever, right? And they're going to go public this year probably at more than that, right? Well, more than that. It's like, this is a real business now. now, right? How do they differentiate, right? One is, yeah, they still continue to have access, newest generation, they're still going to have access. Maybe Google gets, for every 10 that Microsoft gets, CoreWeave maybe in the beginning can get 3. Now at some point, the production capacity is not 13, the production capacity is 100. And again, I'm just pretending there's only 2 people. Now Microsoft gets 90 and CoreWeave only gets 3, right? So now it's 93, right? Whatever, right? But at the beginning of a new GPU's useful life, it is worth the most because at that time you can rent it for way more. And then over time it narrows as more and more competitors enter the market and there's more and more of them. That's one. 2 is the speed of buildouts, right? Microsoft's data center organization and rack and all these things like server organization, same with Amazon, they're custom. They're tens of thousands of people. They've always been used to modifying servers, customizing them to squeeze every penny out, right? And in the era of like GPU compute, it's like, well, what if I just do what Jensen says, right? Like I build, you know, I might make some tweaks, I might put a little bit less memory, I might change the storage around, I might change the network a little bit slightly. But like in general, what if I just take the reference design and like modify it a little bit? Now I am paying a little bit more, right? You know, that, but then I have the like time to market, right? Like the moment the GPU's being produced by TSMC, it's packaged, it can go to the server maker, reference design is ready to ship to me, right? Whereas like Amazon, you know, especially Amazon, their Blackwell rollouts are much later than many of the other players because they have to do all this customization.

Yeah.

Right? And then like lastly, like sort of there's like, I think the other aspect is data center capacity, right? We talked about this ESG stuff. CoreWave gives like, they're very, they've been very aggressive always on like renting GPUs, getting credit at like what at the time seemed like crazy terms, double-digit loans, what the flip, to deploy GPUs. It was the right decision, right?

Yeah.

Now they're established, and so their loans are single digits again, higher single digits. But it's like getting data centers is very hard, and they've been very, very creative. Initially, they were just getting anything and everything. That capacity all dried out, and the hyperscalers woke up, and they went to try and get everything and everything. But then CoreWeave started going to crypto companies, Right?

Yeah.

They went to a couple different crypto companies, offered to buy one of them out, offered to— and then eventually just struck a deal where they're retrofitting all their data centers. Like, yeah, throw all this crypto crap away. I'm going to convert your thing into a data center. Right? And one of these crypto data centers has a 200-megawatt natural gas plant in the middle of 7 buildings. Right? And each of these buildings is a data center. So they simultaneously— and it's like a natural gas plant. Meta, what they're doing is like, okay, the gas plant is 30 miles away, and I struck a deal with this solar power company that's putting tons of solar panels in a similar distance away, but I'm not necessarily consuming those electrons, so it's not clean. In the case of CoreWeave, it's like, no, no, no, I don't care.

Yeah.

The power is here right on site. I'm converting these crypto data centers to real data centers. There's a metric called PUE, power usage effectiveness, is basically how much power are you pulling from the grid/power generation versus how much are the chips consuming, right? Or the servers, right? And so the ratio that hyperscalers try to design to is 1.1, right?

Yeah.

Only 10% of the power is lost in transmission and cooling, right? But this crypto data center, once they convert it, because it's a really dumb design because the crypto people didn't know how to build the data center properly, right? Even though it's being retrofitted, it's still going to be like So it's like you go from 1.1 to 1.4, it's like, oh, if it's consuming 200 megawatts, what was 20 megawatts of power being burned on cooling and electrical transmission and conversion now is like 50, right? 60, 70, right? And it's like, okay, this is a lot of inefficiency, but it's like, screw it, we need GPUs now.

It was worth it, I guess. Yeah, the inefficiency both on the financing side, on the power conversion side.

And then they do certain things really efficiently. They're just a small organization that's well run, right? And their software, their cloud software for GPU rental is literally better than Amazon's and Google's. It's actually hilarious. We'll have a piece about—

What makes it better?

Just the way they do their managed services, the way they manage their network, the way they do storage. There's a number of things here that are just objectively better with CoreWeave. And it's like, what? Um, how did you guys build this?

I never thought the hyperscalers would lose at that game.

But it's like, it's like, well, you have this legacy and, you know, innovator's dilemma and you don't build a brand new thing that's purpose-built. You like morph and you have all these like weird requirements for random customers, like random defense company. You also need to make sure your GPUs are rentable to them or like, you know, and it's like CoreWeave is like, well, let's like, they're starting the problem anew and they have a brand new team, clean slate, really good engineers, and they're building what needs to be built, purpose-built for this business, right?

Yeah.

Um, and they don't have like a team of like 2,000 people who work on storage architecture that used, that used to be primarily targeted for these kit— this use case. And accelerators are different, right? Like what you need in AI, not exactly analogous. It's similar, right? Like, you know, but they're trying to morph this massive code base to that rather than like ground up build something new, right? And it's like, this is the—

The case for every startup in the AI world.

Yeah, like, I think, I think like there is innovator's dilemma with the big companies.

Yeah. What are like, I mean, you obviously must get, see all sorts of kind of interesting investment plays on this massive hardware buildout going on. There's the obvious stuff. There's data centers, liquid cooling. Any like second or third order effects you've seen where you're like, that's a clever way to play the, uh, the massive data center buildout?

Yeah, I think like, I mean, like some of the stuff is like, uh, yesterday I had like an hour long discussion with a guy who's building a data center in Armenia because there's a nuclear power plant It's like, great, this is awesome. And he's doing a venture fund. He has a venture fund in New York, but he's kind of just like, screw it, I'm going to work on this pretty much. It's like, oh, this is awesome. So there's silly stuff like that. Not silly, it could work, right? Although the regulations kind of screw him. I talked to him about that.

Yeah.

But I think a lot of interesting plays around networking and optics. There's a lot of stuff going on in networking and optics, which is a big bottleneck. especially as we continue to scale the density of GPUs and the amount of data that needs to be communicated between GPUs as models get, A, larger, and B, the context lengths get longer and longer because of reasoning. This is sort of like— and search, all these things require more interconnected GPUs. I would say that there's a lot of interesting stuff around transformers, right? I told you they're sold out for 4 years. Well, there's all sorts of interesting startups working on solid-state transformers and partially solid-state transformers and all sorts of hardware stuff that's sexy. There's funny stuff. There's cool companies on carbon sequestration because yes, Meta and Microsoft, Meta especially, and xAI have said screw the environment, right? Not screw the environment, but we're going to build AI as fast as we can. They're at some point going to be like, yeah, yeah, let's be green again.

Right.

And it's like carbon sequestration stuff wrapped up in a data center play is actually pretty interesting. It's like there's all sorts of interesting stuff around storage, what companies are doing there, because video models are going to require a lot of interesting work there. It's purpose-built solutions for the plethora and sea of innovation that's happening and targeting all these things. I think if you look at every layer of the stack, there's more innovation happening today, not just in each stack, but also bridging multiple layers of the stack to come up with a huge improvement. Some of that is happening in the big companies, some of that's happening at startups. I think some of these startups will get acquired by bigger companies, some will actually find a product-market fit, right? I think some of them will license technology, whatever it is, I think there's a lot of hardware technologies that are interesting. And then I think the really fun area is software infra.

Yeah.

Right? So software infrastructure is also getting very diverse because not everyone can just build their own stack. I think that where people are very in tune with the hardware, understand how to make things run fast, also in tune with what's going on with models, and you bring your data and all your capabilities, and I bring you something that lets you build and serve your actual use case, right?

Yeah.

So software infra was not always sexy. It was always like SaaS was more sexy, or some sort of I mean, I guess software infra is to some extent SaaS, but not really. I think these are the sort of the businesses where there's a lot of innovation happening as well.

The infra layer has just been so hard to invest in because the models change so quickly that any kind of scaffolding or things you're building for this current generation of models, you're like, is that relevant in 6, 12 months?

This is why some companies have bet on things that were further out, right? Like I mentioned Adaptive, right? They were working on pre— it's like a few French guys who moved to UAE and they built the Falcon model in UAE, right? And then they moved back to France, and now they're in New York. But while they were building the Falcon model for UAE, they were doing pre-training. They did a 180 billion parameter model to Chinchilla optimal. It's fun, right? It was a good model when it got released. It was just like, whatever, right? Like, Llama just came out too, right? It was really good, but it was like, yeah, they were immediately like, pre-training doesn't matter, right? It's like, we're going to work on synthetic data and reasoning. It's like, what? Okay, sure. And then for a while I was like, what are you guys doing? I don't know. And then all of a sudden I was like, oh, wait, wait, this is right.

You saw something.

Yeah. It's like there's a lot of companies that are like doing something that may not matter today and you just have to learn, or like doesn't seem to be the focus, the popular thing today, but could be and will be in the future.

100%. What about, what do you think of like these AI hardware efforts? Like folks that are using, you know, AI for chip design or, you know, circuit board layout or some of these other things?

I think it's really interesting that YC all of a sudden like decided to invest in like 12 different AI for chip design companies all at once, uh, in, in the last few batches. Um, a couple of them are cool. I actually even invested in one of them myself personally. Just to be clear, I'm not hating on the idea. It's just like very interesting that this happened. Um, but like, you know, AI chip design is very expensive, right? Um, it's very difficult. The number of engineers working on chip design has not grown that meaningfully over the last 2 decades, right? In the, in the US.

I feel like this is the start of like every AI app pitch. It's like whatever, insert your profession here, is, uh, stagnating in growth.

Right. But the output has been dramatically improving, right? And that's because of EDA software, right? Electronic design assistance. And EDA software, as much as you can hate on it, as much as you can say it's like old shitty software, it's actually really awesome because it's like the productivity gains that chip designers are getting year on year are meaningfully large, right? And that's pre-AI. Now Cadence and Synopsys and Siemens are all— Mentor Graphics, right? Are all investing in AI for chip design. But there's also a lot of companies doing it, right? Nvidia has written multiple papers about their internal tooling, and they've hidden a lot of it too. They haven't said a lot about it. Google, same way, right? This is absolutely going to be huge. Now, is AI going to chip design chips right away? No. But is it going to be a force multiplier for a very in-demand profession? Yeah. And the other thing is today, a lot of workloads just have to use CPUs, and then GPUs are the general-purpose parallel computing platform, right? Now, obviously, they're generally making more and more enhancements for certain types of AI models, but GPUs are not good at running CNNs.

Yeah.

Right? And this is a conscious design decision by Nvidia, right? They're okay, they're good, but they're not amazing at running CNNs like they are at transformers, right? And this, I believe, is a conscious design decision. How do you go from CPU, which is the catch-all general purpose, GPU to general purpose for parallel processing, to what is the next step? And Nvidia is branching their GPU architecture out. They have different architectures for automotive versus gaming versus pure data center AI. So they're not just sitting flat-footed, but it's like, how can we proliferate architecture to target use cases better? And today, for it to make sense for you to design a chip for a TAM, that TAM needs to be billions.

If you can bring down the cost.

Right. So Google made a YouTube chip, right? Because that was a billion-dollar, multi-billion-dollar TAM for them alone internally. That's when ASIC made sense. But when chip design costs can drive down, and so AI for chip design will do this, right? I just think it won't be like, it's not going to go from 0 to 1, it's going to be like a force multiplier. And I think it's one of the areas I'm super, super excited about, whether it be floor planning or RTL generation or many other things like verification. Verification is like half the spend of a chip design. It's quite funny, more than half the cost is making sure what you designed actually works. I was like, okay, interesting. Um, like there's tons of like stuff there, um, that I think, um, could be done. And yeah, it's like, yeah, I'm very excited about that space actually.

Well, though, I feel like this has been a fascinating and really wide-ranging conversation. We always like to end with a standard set of quickfire questions. Um, and so maybe to start, uh, what's one thing that's overhyped and one thing that's underhyped in the AI world today?

Overhyped, RAG. Um, underhyped, various other forms of like semantic search. Um, and like retrieval. Um, and there's some interesting papers that came out recently. Yeah.

Will model progress, uh, this year in '25 be more or less or the same as in '24?

More.

What's your weirdest prediction on like the implications for all this AI progress on the future?

I think, you know, there's a lot of like worries that like there's gonna be huge inequality. Um, I'm actually like on the flip side, I think that like, blue-collar type jobs that were generally hated on were like, are going to do well until robotics comes, which I think is pretty soon. But, you know, so for that, but like, yeah, but like, no, but like, I think that like generally the poorest people in the world will also improve in quality of living massively. Right. And it's not going to be like a haves versus have-nots, ultra elite versus not. I don't think that's going to be the future.

Yeah. I mean, I guess you better hope the inference cost on these reasoning models comes down.

Yeah, I like that. The— what is, what is OpenAI's thing? Intelligence per dollar, like maximizing intelligence per dollar. And it's like, well, right now it's really freaking expensive, right?

Yeah.

And I expect, I expect the models to get even more expensive, right? Like a query from like GPT-5.05 or whatever, right? Like reasoning version of GPT-5 is going to be like dozens of dollars, if not hundreds of dollars, right? But it's good. The work it puts out is going to be crazy.

Do you have a go-to query you use when like a new model comes out, test it out?

Um, it's funny. I actually asked them about semiconductor manufacturing because I know the training set quite well. I've read a lot of the papers, and then I see if it actually understands it. And a lot of times models get it wrong. And today, even to this day, Claude is the only model that you can, once you query it enough, can get the patterning stack for lithography somewhat accurate. And that's just because no one publishes on it. So it has to understand the papers and infer what people are actually doing, which I thought was very interesting.

That is a cool way to do it. And we've talked about a bunch of different parts of the AI startup world, but any spaces within the AI startup world we haven't talked about that you find particularly interesting or exciting?

I think there's a lot of cool stuff happening on the distributed training side, right? Like with Noose and with, you know, although I'm not, you know, like, you know, there still needs to be a lot more to prove what they did is good and real, but very exciting. And then Prime Intellect, same thing. And yeah, they also are doing really cool stuff. I think distributed training and inference is gonna be really cool. I think that, I think that a lot of these multi-agent reasoning startups that started a while ago, but have actually been focused on reasoning, like Synth Labs and Adaptive and people like that, rather than agents and pivoting to reasoning, because everyone's doing a reasoning startup now, right? I think that's a really cool area. And I really just like people who, in startups that have been singularly focused, forever on the same thing. And they're right. It's just like, you don't know when they're going to be right. And it's sort of like me loving chips so much. I had a good business, but it did not pop off until AI. There were a few people working at my company, not 20. So it's sort of like same thing as being so internally focused and good at it, but then all of a sudden the time does come and you were right. Those are the kind of startups I like the most. Not necessarily people who are outsiders to a field and just want to revolutionize it.

Well, we always like to leave the last word to you. I can't imagine there's much of a Venn diagram of folks that listen to this podcast and aren't subscribed to SemiAnalysis. So I assume some folks are at least at a high level familiar with you, but would love to just, you know, where can folks go to learn more about you, the work you're doing, any plugs you want to make? The floor is, is yours.

Yeah, I would say if you want to see the serious side, go to the website, check out the newsletter, or check out the institutional sort of products, which is like the main business, like the data sales and consulting. And, you know, we sort of like have industry experts all the way from, you know, former industry experts all the way from like, have worked at ASML and like Lam Research and Intel all the way up to like Microsoft and Nvidia formers and like, you know, everything in the middle. And then also like half the company's like ex-hedge fund people. So it's a cool mix. And then we also have— and then I would say if you want to see the like lighthearted bit, right, where I will tweet jokes like, hey, you know, AI regulations came out on Monday. Arsenal for democracy. Oh look, Israel-Palestine ceasefire two days later. Wonder how that's absolute joke. But you know, that's what I post on my Twitter, right? Like so, you know, like I think like you know there's a there's a range and that's Dylan five two two p, right? So I think that's sort of the the shilling I'll do for myself.

Amazing. Well, Dylan, thanks so much for coming on. This was awesome.

Yeah, thank you. Thank you.