Open source is a bit of a fake because China is paying for all the training. Okay, it's not open source like a generation ago. And the reason is because there's so much innovation and cost savings. So, what do we discuss today? Google loses two generational scientists in 48 hours. That's a tough time. DeepMind closes a $7.4 billion Series A. Is the Series A at a $50 billion price, but only China gets voting rights? Interesting. And then finally, the $725 billion question Wall Street is finally asking: who's actually going to pay for AI? There's only one thing worse than a seat-based model, and that's a model that's based on bodies. You don't get to make $10 million for working 18 hours a week. You get a watch. You get an Omega. You want an Omega or you want to be rich? Make your choice, boys. The whole reason the OpenAI and Anthropic models work is because other idiots have spent the $300 billion on their behalf. Ready to go?
Boys, it is so good to be back. Jason, you are back. It is so good to have you back from China. We're going to start with the news that I put at the top of the list, which was DeepMind loses two generational scientists in 48 hours. Namely, we have first Noam Shazeer, who was Character.AI, and then we have John Jumper, Nobel Prize winner, co-creator of AlphaFold, also leaving to join Anthropic. How significant are these moves? What should we read from this?
Look, listen, it's easy to pick a turnover in any organization, right? There's so much turnover in any organization. On the other hand, when you talk to some of the smartest engineers and developers in AI, they're really looking to be in a very specific environment, right, where they get to pursue exactly what they want to do, especially on the research side, right? Especially they want to work on what they want to work on, and the best of the best. And you know, I was thinking back in the day, I went to pitch Google for my last startup, and Vint Cerf was there, like one of the creators of the internet, popped into the meeting. Okay, I didn't get it at the time, but Google back in the day had created this environment where the best researchers in the world wanted to be there, and I think that's how they lured the DeepMind guys in, right? Was the story, right? This persistent to create this environment: look, you're going to get to stay in London, you guys are going to get to build your own thing. And I think this is probably a sign of the cracks of the realities of having to try to be number one in AI and forcing an environmental change potentially that your competitors can welcome, right? And OpenAI can say like, just come over here and work on whatever you want to work on for $500 million, $2 billion. And it's just when I talk to folks at the bleeding edge of AI, that's just so appealing. It's just so appealing to only work on what they want to work on on AI. But again, I just got back from China. I might be off by a beat, but if you want to get the best researchers, you have to give them this environment where they only get to do what they want to do.
It's funny because that feels like a one-dimensional answer, but there might be two dimensions to this. Because in one sense, yeah, you have that whole researchers just want to go do what they want. On the other hand, you listen to a lot of people who left Google, and it's a little bit of the frustration of not being able to ship. I mean, there's a lot of frustration that they had a ChatGPT alternative, and then the bureaucracy just kind of smothered the product when, you know, OpenAI just jammed it out the door and as a result took a lead on them, right? So I think that it's funny when you have a history, you know, an existing business, you're kind of damned if you do, damned if you don't. Sometimes people want to just do research, other times they want to actually get stuff done and ship, and you get in the way of that. My sense is, first of all, these two people in terms of their research pursuits are somewhat different. And that, you know, Noam obviously was at Google, did the original attention paper, left, did Character, got bought back to Google in large part to be, you know, it was a very clever acquisition, the brains behind their restarting, their, you know, after OpenAI stole a march on them. And I think it was a couple of billion dollars. So a good slug of that was going for him. So assuming four-year vesting, he's probably left half of whatever he's offered on the table, right? So that's a lot. I can imagine there it's a combination of some version of, hey, as you say, more research, more ability to do things, plus a more certain ability to ship and make stuff happen. Because I think even though we all did the last year, we did the yay, Google is amazing because unlike the other three, unlike the other hyperscalers, they have their stuff together, they have an AI story, and the stock has reflected that in the last year. It's the only one of those, it's one of only two of the Mag 7 that's up on the last 18 months.
But at the same time, when you look at things like having a viable coding model, having really kind of that next level up from just shipping a model to shipping you interesting products, the truth is Google hasn't done an amazing job, and OpenAI and Anthropic have. So if you're into product shipping, which I think Noam might well be, then I can see why going to one of those two makes sense. Totally different, you know, on the Jumper. I mean, this is someone who's your pure research scientist, around undergraduate, postgraduate degree, I think University of Chicago, all focused on protein folding. Remember, got a Nobel Prize. DeepMind has been the only company he's ever worked at since his, I don't know if graduation is the right word after you get your postdoc, but whatever, since academia, right? So basically it's been a whole bunch of academia, a whole bunch of time at DeepMind, pick up a Nobel Prize. You got to believe you're going to Anthropic because there's a story there about being able to do more research, which is almost, you know, on kind of high-end science, which again they've had, Anthropic's announced an initiative on that. So somewhat different, but I think what it speaks to, stepping back, is, you know, when you're top of the heap, you can promise everyone everything to everyone in a way that you're not. And the truth is, top of the heap right now means you're the new model, you're the new companies, you're unconstrained by history, you're unconstrained by the install base, like that old proverb joke about hell is the install base. And you got a stock and a currency that is huge, and no one's giving you grief about stock-based comp. So if you're Anthropic and OpenAI, you know, you can buy whatever you want, including people, and you can let them do whatever they want, including whatever it is they've been promised to do, in a way that you've much less constraints than the incumbents.
You know me, I would never deal in rumors. I don't do rumors, but the rumor mill that I heard was...
Come on. Come on. Bring it on, baby.
Is that candidly, Anthropic have clearly had a breakthrough that a small number of people know about, and that's why John Jumper went there.
The reason I'm skeptical of that, Harry, is the lag time between I've had amazing invention and revenue in the core LLM space is a year or two years or three years. The lag time if you have a medical invention, an idea, and turning that into real meaningful economic value is 10 or 15 years. I mean, you know, the protein folding advance has collected its Nobel Prize and as yet has not had meaningful commercial success or a drug in production yet. So I doubt it's like, oh my god, we've discovered something new and this is magic, and if you don't join in the next three months, you won't be part of this thing. My guess is the initiative, whatever initiative Anthropic has around next-generation science, is a multi-year thing, and it's just a question of where do you want to spend the next five years of your life doing research.
It is a vibe, for what it's worth. And listen, I really hate to be one of those like VC dad tells stories about their kids or what happens in their spoiled kids' school, you know, as your thesis for investing. But my son is very good at a certain type of math. I don't understand Jensen fine string theory and all this stuff, but as a college student, he's one of those like top 10 in the country. So my point of the story is the labs find him, right? He doesn't have to apply per se. So he got an internship offer without looking from Anthropic for an amount of money that when I was in college or grad school would, you know, and the opportunity is incalculable when you graduate even today. And so he instantly turned it down, right, with no job. But he's just like, I can't do enough research. I want to do my type of research for my type of math and my type of AI. I mean, and so, you know, everything that we do on this show or the other stuff, he kind of makes fun of me because he's so far ahead of how inference works, how open source works. I mean, really, you should replace me with him. But my point of the personal dad VC stories, which I hate, is he instantly just said like they couldn't create... he's not worried about money. He's not worried about any of this. If they can't create the right learning environment for him, he just won't do the job, right? And so I don't mean to map my family to two of the top researchers in the universe, but I feel the same vibe. Like in today's world, we're on a bull run like we've never seen in our entire lifetimes. There's nothing like this where a so-called startup can pay billions of dollars to acqui-hire you, and then you leave with 50% of your billions uninvested. We've never been like this. And it just creates an environment for the best of the best where you just will only do what you want to do. You just won't do the job that isn't 100% what you want to do. And remember when Harry and I were in London and we were with Mira for OpenAI, I think she said she was one of the sales leadership. She said she'd never been in the researcher building except for like one or two meetings. They didn't allow sales in the whole building with the top researchers. We don't want these pesky go-to-market professionals bothering our researchers. So anyway, I hate to be that guy, but it really resonated. Like you just, you have to like, and it's really hard for if you're not Anthropic, how do you retain, now it's not DeepSeek, but how do you retain this talent? How do you let people work on what they want to work on when you need the chat bug fixed so you can compete with Sierra? I mean, how do you contain them? And I think it's very hard.
First of all, you're exactly right. I mean, these are two of the most talented people on the planet. One of them wrote, was part of the attention paper and has made plus or minus a billion dollars, and the other has a Nobel Prize for medicine and is still relatively young. I mean, these are not top 1%, they're not top 0.1%, they're top 0.001% people, right? Probably is one of, you know, out there on the planet in terms of research. And the funny thing for Google, the so hard thing is they're struggling to keep people like that while at the same time they're not getting the tactical stuff done, you know, and just making the stuff happen to kind of continue to make progress. And the amazing thing about someone like Anthropic is they're able to do both. They're able to create these wonderful research environments for the leave-me-alone, let-me-do-research people while at the same time they're obviously executing tactically brilliantly in terms of product making stuff happen, right? That's it. It's momentum. I mean, that's my aha here. It's like everything in life. When you start winning, everything starts going your way. The people start going your way. The breaks start going your way. You know, winners win and they compound until something happens to break that chain, right? And this is just it and magnified.
Listen, Google overall I think actually has been on fire. But I think as I continue to learn, and I'm only so smart about the role of open source versus closed source, right? The most vulnerable is going to be number three. And so you really, for talent, for people, for revenue, and for your ability to do things that maybe aren't core, if you're number three as the closed-source LLM, that's where you're going to hit the most pressure from open source. And it's just, sometimes when you're in that, and I'm not saying this is correct, but sometimes when you are in that place, you feel like you don't have the luxury of letting folks do what they want to do because you're under such intense pressure. Maybe Anthropic feels like, as competitive as it is, it has a luxury that its competitors don't have.
I think that's true. And just I want to come back on the Google is executing amazing. I was not commenting here, just look objectively, as I over the last 18 months since 2025, only two of the Mag 7 have outperformed the S&P. It's Nvidia and Google, which is another way of saying, you know, people like Facebook, Amazon, and Microsoft are doing stuff, they're not relevant here. Google has done an amazing job of being relevant here, and that's a true statement one. But statement two, that's equally true, is you don't wake up every morning and say, let's try the new Google model, let's try the new Google harness, let's try the new Google coding tool. You do try Claude Code. You do try Cursor. You do try OpenAI. So the fact is they are relevant and in the frame, but they're not yet, as you say, but they are definitely number three in terms of innovation, which is a whole lot better than being Microsoft or Meta and having to say we don't have something yet, but we spent 70 billion, we might get something next year.
Right. Jason, just so I understand, why is number three the worst position and like the perilous one in a way that it's not for cloud?
There are two threads happening at the same time. On the one hand, clearly the price of tokens, token maxing, the budget issues are real. And so the amount of folks running routing models, running multi-model is exploding, right? And so at a small level, the OpenRouter and your own sources, that stuff's all a big deal. And everyone is realizing they just have to get smarter and smarter much more quickly on routing workflows to different models. Okay, that is clearly very true today. And in maybe 90 days on the pod, it wasn't clear how big a deal that was. It is across everything except the smallest of startups are routing. Okay, so what do you route to? Well, you know, if you have three vendors, and Rory's the professor here, especially with his background, you know, number two was often simpler. Number three was usually cheaper, right? Some variant of that, right? So, you know, and frankly, GCP, old Google Cloud, blew up because it was cheapest. It was cheapest. You know, a generation ago, Google Cloud struggled in the beginning, and then it was just cheapest and simple, and so people would move certain workloads a generation ago to Google Cloud. Now, you're trying to do the same thing with your massive, massive AI spend. And the question is, open source is really complicated. Open source in inference and training is not free, unlike Linux. It's not free. There are substantial costs, but it is materially cheaper, which we could talk about. So should your number three, forget about, in fact, in theory with OpenRouter and others, you could route to 10,000 models. In practice, is the number three thing you're going to figure out what's the best open-source model I can use for my application, or is it the number three closed source? Open source, there's so much going on in open source, right? It's fueling these crazy, Baseten and Fireworks and all the... there's so much innovation in open source that number three just might get swamped by all the subsidies of the Chinese government, everything else subsidizing open source, because open source is a bit of a fake because China is paying for all the training. Okay, it's not open source like a generation ago. And the reason is because there's so much innovation and cost savings. So your number three there, there just might not be enough energy for the number three closed-source model, even if Google has the billions to fund it. They do have all the money it takes to fund it, but developers may lose interest other than it being, what, a setting in OpenRouter and friends.
That's my thesis, and that's what I'm seeing too, right? And it's a big deal. And literally this morning, I got an email from Anthropic. This is their sort of shot across the bow for open source, saying your prompt cache hit rate is low. Your prompt cache hit rate is out of the blue. I don't know if you guys got this email this morning, but you know what Anthropic is doing here aggressively is fighting back at open source and trying to get you to cache your prompts, which are very expensive. And they offer such a massive discount on cached prompts if they work for you that it actually can be cheaper than open source. This didn't say, you know, fighting Gemini. This is an email they sent to their, maybe their entire base, saying cache your prompts so that it's cheaper than open source.
So that's why I think it's just hard. Number three, you just, you know, you can't be cheaper and you can't keep the researchers and the projects are less interesting.
No, but yeah, maybe I was because you... but the short answer is I agree, you're right, Jason, is that look, typically tech markets tend to, I mean, look, you don't end up with perfectly competitive tech markets in the economic sense of millions of players. You end up with small numbers in a tight oligopoly where there's a leader, a number two, and then depending on the size of the market and competitive dynamic, maybe there's a three and a four, and the vast majority of the revenue goes to one and two, right? It's just the structure in the cloud market where AWS was first, Microsoft second, and Google Cloud third. The interesting thing about Google being third, right? Well, a couple of things. One is unlike the typical third, if they were a standalone company and didn't have the Google balance sheet behind them, I think it would be incredibly tough, right? And I think implicit in that statement is it's almost impossible for a number four to, same business model, closed-source number four, to emerge and catch up at this point. I mean, Aleph had a nice piece on that six, twelve months ago. The market is set. The game is clear, right? And the only reason Google can keep on punching is because they have a whole balance sheet behind them. Because I agree with you, Jason, the other part of it is if there is a compelling alternative to this entire set of competitors that's, you know, 5x cheaper, which is what open source is, then it's going to grind everyone down. And when an industry gets ground, what tends to happen is the number one guy makes a little less money, the number two guy makes quite a lot less money, and the number three guy goes bust, right? And in this case, obviously not bust because they got Google behind. But you're right, it's a real powerful downward pressure on the profitability across all these. And that is, I think, the big story now, kind of from open source, with lots of caveats.
Rory, how do you think about sovereignty then, and like sovereign models and Mistral of the world if there's no room for number four?
There's two ways to answer that. Well, we'll answer sovereignty first and then the impact of China on sovereignty. Look, Europe is effectively saying, if someone says we need a sovereign model, rephrasing that, what you're saying is we no longer are part of the market over there in America where our product is third. We are in fact first, instead of being fourth in the worldwide market for closed-source state-of-the-art foundation models, we are first in the European market for state-of-the-art foundation models. Right? So you're effectively saying I couldn't compete as the fourth player worldwide, but I can compete, profitable for me but probably less efficient for the system as a whole, as the dominant player in, you know, Europe. And look, a million years of economic theory explains why that's a dumb idea from an efficiency perspective. By definition, everyone in Europe is getting the less good model at a higher price. But someone's deciding that there are political or national security reasons to pay that tax. And that's a, you know, political decision. It's beyond the economic analysis. It's just something that government may choose to do or not. Right? But the other part of it, and I'm going to segue to Jason's trip there, is the other part I thought the sovereignty question is the fascinating thing is all the competitive models to the foundation models, all the open-source models, bar not quite every, not quite all, but most of them are Chinese-based. And, you know, Jason, you're just back from China. You know, in the context of sovereignty and security, it is amazing that the entire open-source initiative is running on, you know, four or five models, all of which are built in China. So what was your takeaway from there?
Well, it was, you know, it was interesting. I mean, folks probably already know that, you know, when you go to China, and interesting, even Hong Kong, where Hong Kong has its own sovereignty, it can do whatever it wants. Anthropic and OpenAI don't serve there, like intentionally for security. Like they don't allow you to access Anthropic or OpenAI. You can't access it effectively. It's hard. You can do it over a VPN, but it's harder. You can hack it on your phone, but it's tough. They intentionally try to block, whatever, IP address blocking or whatever. It's not China, it's them blocking it, right? But Gemini doesn't. So when I'm in China for two weeks, it's DeepSeek and Gemini. Those are my friends. It's a parallel universe. And DeepSeek is intentionally crippled in China. It's not as good as it is here. It can't search the web and it is trained on different data, as near as I can tell. But you get it. I don't think I got the sovereignty argument until I was in China. Right? This is China basically saying we do not want to be subservient to the United States. Whether this was the original goal of DeepSeek and others and Alibaba and others, I'm not sure. I don't have my professorial background on my YouTube, but it clearly is where it is today, right? Massive government subsidies, right? DeepSeek's raising this round at $80 billion. The government's the only one getting voting shares. It says all you need to know, right? And it is an existential sovereignty thing. The Chinese do not want to be reliant on Anthropic and OpenAI to run the next-generation economy. It's very smart. And whatever investment the government has quietly made to subsidize all of these providers, it's a drop in the bucket, isn't it? At the sovereign level, what, $10 billion, $20 billion, even $50 billion. I mean, this is nothing compared to what does an aircraft carrier cost? A lot. It's easier just to subsidize an open-source model and pretend the training costs are $10 million. What did DeepSeek claim their training model was back when it launched? $15 million to train an OpenAI competitor. Like, of course it wasn't true, right?
The round itself is pretty wild in a couple of different ways, in that it's $7.4 billion at a $50 billion, give or take, price. The thing that's crazy is the founder is committing 20 billion yuan himself, which is like 40% of the round. I mean, it's like $3 billion the founder is committing himself. Wait for it. And there's fewer than 10 investors, including JD.com. I saw this, I was like, what? And then none of them are getting any rights. The only people getting rights is the Chinese state, which retains governance control. It's wild.
Well, it's just a different world, right? These are not companies that, you know, it goes back to the old hallway drama. They're not independent of the government in any way, shape, or form, right? And so you either do it the right way or you end up thrown out of your own company or, you know, having to recycle and give back your $2 billion that Meta took from you. I mean, you got to play by their rules, right? And so, you know, this is the rule for DeepSeek to not be taken from you.
You know, a couple of comments here. First of all, your comment on, oh, only the Chinese government's getting voting rights. I felt like saying, so the only people getting voting rights are the only people who don't need them, because the Chinese government doesn't need voting rights, right? Because they have sovereignty, right? They have an army, right? As we've seen by Meituan, they can make you disgorge back your money right after you've got it, and just by leaning heavily on you. So it's not like, I mean, so in one sense the round's not surprising at all. And, you know, so zooming out two levels, two other things on that. One is, yeah, it's a $50 billion, but we have two companies, the leading two American companies, closed-source models, which by definition have more economic upside, are trading plus or minus a trillion. So $50 billion for the open-source competitor feels roughly right, right? You know, one-twentieth of the price. And, you know, there is a couple of, I think, and Zhipu, I checked, Z.AI, which is the Americanized version of it, is actually public now in China, and that's trading, it's trading at $100 billion or something like that, a thousand times revenues. So again, plus or minus, if you think of these as the sovereign alternative and the open-source alternative, they're trading at numbers that aren't crazy at all compared to the US. And then the other thing is just to say it, you know, like we're doing a little bit of the, oh, you know, you have a lot of state interference in China, and, you know, you're seeing it over and over again, and your property rights are weak. We saw it with Alibaba back in the day with Jack Ma. We're seeing it here and now. But, you know, at some point we'll have to talk about the fact that you're seeing more of that in the US as well, right? We're seeing Anthropic unable to ship the most recent model until they satisfy the concerns of the US government. And I got to give, you know, Leopold in his 2024 Situational Awareness credit, as like he called it that about now national governments are getting involved, and we can argue that a little bit overwrought, but in both, yeah, you can't give China grief for this kind of thing when we're doing the same thing ourselves. Both governments are feeling this technology is pretty existential, pretty impactful, and trying to figure out, do you regulate it? Do you take over it? Do you stop it from being used by other people? There's a whole bunch of, you know, policy questions going well beyond economics that are being raised here. And yeah, China's solving it their way, and we're solving it ours, and Europe is doing theirs.
Rory, you mentioned Zhipu, or I hope I pronounced it right. Z.AI's GLM 5.2 beats GPT 5.5 on coding benchmarks. How significant is this? This was deemed one of the most consequential open-source AI releases.
I mean, look, to me, all it says is these guys are cranking. I mean, there's probably six total open-source Chinese models. Three of them, you know, top at or close to US performance, three more just behind it. The aha for me is it's comp, to Jason's point, it's a compelling competitive alternative. You know, many US companies are starting to build their own model based on the open-source alternative. So it's providing some kind of, it's providing a ceiling on profitability for some of the US closed-source vendors. It's really economically significant. And I think the most compelling fact is that there's six of these companies pounding it out there. And there's just a lot of competition on the open-source, you know, state-of-the-art alternative. I mean, we can talk later about distillation and how much of what they do is a function of, you know, the ability to learn from what the frontier models are doing, but they're there. They're relevant and they're providing a competitive, umbrella is not the right word, a competitive drag on, you know, what Anthropic and OpenAI can charge.
Behind all of these, there is actually the infrastructure, and a core part of that infrastructure is seeing increases in price. Memory being one of the biggest. Price of memory has gone up four to five times in certain cases. Tim Cook told the Wall Street Journal that Apple faces a 100-year flood in memory costs driven by AI infrastructure demand. DRAM contract prices rose 90 to 95% in Q1 alone. How significant are these cost increases? Who feels them? How will we see this? What should we take from this?
I think other than a profound regret on not buying Sandisk and Micron Technology a year ago and making a 20x. I mean, look, I think it's funny. I was thinking about this and going back to the kind of discussion on the researchers at the start. Very different topics, but actually they're all about the same thing. The investment in AI is commanding resources, right? And then via the price mechanism, everyone else is getting impacted by that. I mean, you know, and the impact of this, it's going to manifest itself, as you say, I can take four DRAM pricing, right? It's going to manifest itself in the price of your iPhone. It's going to manifest itself in the price of your electricity. It's going to manifest itself in the price of your house in San Francisco, right? It's going to manifest itself in terms of 20% of you losing your jobs if you're one of the companies that wants to kind of go all in on capex for AI, like Oracle, right? So what you're seeing is the positive side, and again, this is not a, oh, AI is bad, as much as economics just sends its signal. It doesn't have morality. It just says, oh, you want to devote more resources and more DRAM to a data center in Tennessee or Mississippi. That means you're going to have to have less DRAM for Rory's iPhone. And the only way to make that happen is to raise the price of the iPhone, right? And that's just what's going to happen. I mean, this is price filtering. And the amazing thing about the AI capex explosion is it's so huge that it's literally impacting everything, right? And this is just one of the examples. It's just sucking in the money. It's sucking in the resources. And specifically, what it's going to mean is Apple has clearly decided wisely not to swallow the loss and lower their margins. They're just going to raise prices. And they'll sell a few less iPhones because of price. You know, there's some price elasticity, and everyone will have to pay more for their...
iPhone. Goldman Sachs projected there'd be 7.6 trillion in cumulative AI capex from 2026 to 2031, but it was fascinating and it just brought up now the $725 billion question. I remember David Kahn from Sequoia did the 50 billion, I think it started...
600 billion, I remember well. It was a good piece.
Yeah. Are we going to have a trillion dollar question next year that we're going to be discussing?
I mean, explain for what you're basically saying is because it's worth getting outside the inside baseball part of it. What you're saying is what Goldman Sachs is saying is the hyperscalers are now spending $700 billion in capex a year, which times five or six years is $3.5 trillion, some astronomical sum. And you only do that if you expect revenue to be greater than expenses because that's how American capitalism is meant to work. So that implies that at some point someone has to spend $700 billion in revenue for you to make a buck, right? And right now they were nowhere near that. We're probably well under a hundred billion. So AI as a whole is getting a hundred billion in revenue and spending 700 billion a year. That's not a great business. And what you're basically saying is how does that end? And with the increase in capex, does that increase the revenue requirements?
Absolutely. No, people are going the other way, and that's the funny thing to say. I would have thought that David Kahn piece a year ago was really good and it made me think, and I was wrestling with the same things. And then you step back and you say since then all that's happened is people have doubled their conviction and doubled their willingness to spend. At that point it was, oh my god, capex is like 60% of free cash flow for the Mag 7. Now it's 120% of free cash flow for the Mag 7 and they're borrowing to fund it. Right? So even though a year ago, and this is the classic thing about bull markets, is that yeah, you can be intellectually right but the narrative can keep going a long time, right? People's willingness to invest, you know, it's very hard to know how that ends, right? When, not how it ends, when it ends.
I listen, I'm only maybe Rory can again share some of history where there's an analog. I just think in my lifetime in tech, being aware of things, I can't think of another time where this level of demand was infinite. And so how we deal with it, how we deploy capital, but if you're sitting on the other side of infinite demand, you can choose not to embrace it, right? That's like shorting everything, but where does it get you, right? I mean, you've got to, the demand for AI is an order of magnitude more than it is today if it could be served cost-effectively, right? We would all be consuming tokens 24/7 if we could. So the demand is so much higher than is delivered today. And do you play the game or are you a Debbie Downer, right? I don't know what the right answer is, but certainly in venture, you don't make money not deploying your fund. Like it's hard to, I mean, fees are nice and all, but like you got to deploy into demand cycles and just hope to God you get liquidity before it crashes down on you.
Diving one level deeper because you're right, Jason, you can't just say, 'Oh, I think this is irrational and go home.' Right? First of all, that would clearly have been the wrong decision for the last two years. And B, it's not the business we're all in. We're in business investing the future. But you do have to think about, and the comment you said was interesting, which is that you've never seen this level of infinite demand and then you caveat it in a very important way, price, because you said price adjusted. And I think that's the difference between this and prior generations of software technology because with something like Salesforce, if you're buying a SaaS product, there was really only two states. If you didn't need it, you didn't buy it, right? And if you needed it and you're a big enough organization, you paid the hundred grand and you said it's just a cost of doing business, right? Because it was kind of a fixed price. What's interesting now about tokens and intelligence is there's an infinite demand for intelligence if it's free, right? But it's not free and now everyone's just wrestling with some version of so how do we allocate that given that it's not free? Like no one said I'm going to buy 100 seats of Salesforce if it's $1,000 a year a seat, but if it was only $500 a seat I'd buy 2,000 seats, right? I bought seats for my people if I needed them and then I didn't, right? Whereas intelligence is this thing as you say where the real question is not do you want it, the real question is how do you, you now have to have this new skill that didn't have to exist in kind of the prior world. If you're a CIO, how do you decide how much to spend and where the cutoff is, right? Because I think, and I've always felt that the slowdown here doesn't come from some kind of technical barrier around, oh, the models, because the people who articulate the models can do everything are correct. The models are going to keep on doing everything. They're going to get better, blah, blah, blah. Right? The real question is price. Price is going to be the arbiter of how much you can do. And I think you're right on that. To make it in practical terms, even the last month and a half, there was suddenly this recognition that maybe token maxing wasn't a good idea, and now you're seeing people try to be more efficient. The question is, do you think that shows up in actual slower revenue growth for Anthropic and OpenAI? Because that's the rubber hits the road question. If they're sending you an email, Jason, trying to effectively cut your bill by saying if you're more efficient, you won't be spending as much with us, do you think that reduces the growth from 10x to 5x? How does all that shape out in terms of revenue growth? Do you have an opinion?
I don't know. I mean, I don't know if you know, it's interesting. I got that email. You guys got that email, but also I'm on the max plan where for $200 I get $10,000 of inference a month if I can use it properly. Right? So for me, they may have more of an incentive to send me that email than someone on an enterprise plan where for $10,000 I spend $10,000 a month, right? Anthropic and OpenAI are running an A/B test where they have a free segment of their base which they are subsidizing, right? For sure. It's just bigger at OpenAI. They have an enterprise base which has between 40 and 70% in gross margin just on inference and then there's this prosumer one, the max guys, right? That some of us they're making a profit on but some of them is massively subsidized, right? The prosumer, the kid vibe coding $10,000 of tokens a month paying $100 or $200 for a max plan is massively subsidized and that's not a joke, right? That's the whole OpenClaw drama. It's a real issue. And so we have this weird spectrum of free, massively subsidized, and actually if you just look at inference and not training, fairly profitable enterprise customers, right? Fairly profitable. I don't know what the term they use is, right? It's an unblended gross margin, but the inference margins are attractive, right? And that's what open source is attacking, that enterprise customer, the lucrative customer.
Yes. Because it's lucratively profitable for the vendor, right? And the question is, is it profitable for the customer?
In other words, have people said, assuming you're not on some kind of cap plan, and I think those are increasingly harder to get for the enterprise, assuming you're on the API, can you really afford to give everyone free untapped intelligence all the time, and does the math of that work? And as I say, the question is what's happening? I mean, I'd love to know what's happening in real time now on usage and revenue across those companies because that's the be all and end all question.
Yeah, I think the story, you know, it's interesting, token maxing isn't really the story. I think to what Rory said, I think the big story of 2027 in AI in the enterprise and all the margins in the enterprise, right, is show me the ROI next year. Right now, it's token maxing because 2025 into early 2026 is, guys, just go do it. I don't want to be behind, right? We can make fun of token maxing, but it was the best way to get teams AI fluent. Just go build it, guys. Here's a hundred million, 50 million, 10, 5 million of whatever, guys, just go build it. We've even seen in our portfolio companies then the reaction was, guys, it's actually not a joke, you spent too much, right? You blew through the IT budget is bounded, it turns out, right? Then the review, you're already starting to see it, but I think going into 2027, CIOs and others are going to be, okay, just show me the ROI. We're not just going to ration tokens based on who we like the most in the company and who makes the best PowerPoint pitch. Show me the ROI. And if you have ROI, if you were able to lay off 20% of your department or you have the highest growing division in our company, we will give you more tokens. And this group that can't ship or can't get anything done or is in decline, we're just not going to give you the tokens. And so for the first time, ROI is really going to have to connect to a lot of this spend.
I agree. You're right. I mean, it's sometimes I think saying the obvious is really helpful. I think you're exactly right, Jason. And if you're running a big company and you said, 'Okay, I wrote off the first half of 2026. We spent way more than we thought, but at least my people now know how this works.' Write it off as a one-time thing. You're exact. And the question now is ROI. And then bringing it back to you, the $750 billion question, the Goldman Capex question. I mean, I did this math a while back on the fly on the podcast. I'll kind of revisit it again. Let's round up to a trillion because if you're spending 750 on capex, you got to pay for electricity too. You need a trillion dollars of revenue, right? And if companies are going to give you a trillion dollars, they got to be getting more value than that from that spend, right? So trillion and a half plus or minus, you know, right? The total spend on labor in the US across everything is, you know, GDP's 30 trillion, you know, it's sub 20 trillion. So you're really talking about 7-8% of the labor force being replaced by tokens for the math to work. It's, I mean, when you look at that, you go, it's a dauntingly high bar, right? There's a little part of me that says I think this is amazing, but I don't know if that last dollar in capex is going to earn a return, right? And if it does, it's going to be because there's a huge amount of labor force disruption. Let's call labor force disruption the negative spin. The positive spin is of course productivity improvements. If the trillion dollars is going to earn a buck, you're going to have to have huge productivity improvements from AI. And huge productivity improvements result in labor displacement and hopefully eventually new jobs for that labor that's displaced. But if one's going to happen, the other's got to happen. The problem with productivity calculations is there's also a parity tax, right? Like, you know, if you're the only one using a tool, you can achieve certain types of productivity even if it works. Even a lot of software productivity was BS, but let's assume it even works. But then all my competitors deploy Salesforce and Anthropic and all of a sudden it doesn't show up because we have parity. So you can't go back. But we may reach a situation in 2027 where we just cannot prove any of these productivity gains. And so going to your math, everyone may just need to lay off 10% of their company in addition to all other layoffs, right? Maybe the Oracle layoffs and others are just early views of where it's going and Robinhood and everyone's just going to have to say, 'Listen, we have to fund this. We have no choice.' Right? This isn't really, we can talk about ROI and productivity, but if we don't do it, we're going to lose to our competition. So our team has to be 15% leaner next year, guys. It's just this simple. It has to be 15% leaner.
I'm going to, and now I'm going to do the professor thing just for that. Productivity, you're right. You can have an improvement in productivity and a massive improvement in productivity in industry by virtue of the advent of an enabling technology like AI and no improvement in profitability for that industry if everyone adopts the technology. You're right. So if every bank adopted ATMs at roughly the same time, everyone could in theory save on tellers. In fact, they ended up taking more tellers because the business expanded, but we'll skip that for a second. But the problem didn't change. And the same could happen here. Everyone adopts AI and everyone's cost function reduces by the same amount and you're right. But the people who don't adopt it are dead. That's your point, right? Which is why you have to lean in. And I think that's probably likely in a bunch of these industries. If management goes, segueing to, if consulting or any of these companies, any of these white collar jobs are 20% more efficient with AI, then everyone should adopt it and then everyone will be, then the cost of those white collar jobs like audit should reduce by 20% all other things being equal. My question is, do we see the same acceleration in model progression capability that we saw in coding in legal in accounting? Because if we do, where Andrej Karpathy says I move from 20% to 80% in a six-month period, if you're able to see that same advancement in legal...
Whoa. I think why wouldn't, listen, obviously coding is uniquely well suited, but we just built an AI VP of finance while we were in China and it's already better than any of the humans on our team.
It creates the quote. It builds the contract. It ships the contract. It gets it signed. It updates the opportunity in Salesforce. It logs into bill.com where Aurora is an investor. It sends the invoice. It follows up on the invoice. It gets it paid. It interacts with Brex and it closes out the entire transaction. Then it logs into QuickBooks and makes sure for the first time in 10 years our books are accurate. They've never been accurate before and the revenue is properly characterized. The agent does all that. This is what the models today, and which are not tuned for this workflow.
It's so good. We just got, we had a human that for SaaStr Annual this year where Rory was at forgot to invoice $80,000 of revenue we had to write off. Now listen, I can afford it, but it was so frustrating that this person just didn't do it and didn't even explain any reason why. Just didn't do it. So Amelia's like, I'm going to build an agent to do this in China when we're there for disaster and then boom. Like it's just, you know, it's better. It's better than the humans.
Did you have a full human being doing all that? Did you contract labor as a result of this or is that person doing different stuff?
Oh, I mean we use contractors and for us it's more that we will use less hours of the contractor on accident, but this is an issue too, right? Variable cost human labor which does exist, right? You may even use less of it incidentally due to agents and intentionally this is not replacing for humans on the team. This is for us, this is the agentic story. We replace things that humans are unwilling to do on our team. They're unwilling to follow up, they're unwilling to update the sales job. They're unwilling to send a proper invoice. They're unwilling to do these. And rather than fight the fact that as humans, we really only want to do like 5 or 10% of our jobs, including investing. I only want to do 5-10% of that job too. We just have agents do the other parts. But if it's a variable price, you could see what you pay to these humans fall by 50 to 60% without even intentionally trying to save money. Brandon from Record posted nine hours ago, which I sent to the whole team here. Training agents will be the largest job category in five years' time. Largest job category.
Yeah. Training agents will be the largest job category in 5 years. Well, look, I'm sure he's right. He's much smarter than me. But didn't we say this about prompt engineers when the show started?
How many prompt engineers have you hired on the 20VC team?
You know, the fact that we were able to build a director, we're going to call him a VP, a director of finance remotely in China that is better than any human on the team in a single-digit number of hours is kind of this point, right? And then Rory asked, why isn't everybody doing it? And my response is they can. They just don't have the mindset to do it today. They just haven't spent a year vibe coding. So they don't know what's possible and they haven't scaled a certain amount of human learning to have the comfort. You know, it's like doing your first venture investment. It's a little scary, right? But when you've been doing it for a while, the next one doesn't seem so complicated, even if you know how to put the pieces together. So it's kind of like that. So I do think it's true. I think this is the number one skill is being a master of agents. But I think what that means is going to get redefined each year, right? So again, when we started the show, people were still hiring prompt engineers because prompts were complicated to craft. They were really hard to, if you didn't, when we started the show, whatever 60 weeks ago, if you didn't craft the right prompt, the software that came out was unusable. Okay. Today I can go in, I can vibe code something. I can say build me an AI VP of finance, connect it to QuickBooks, Salesforce and our other app, reuse what we have and just automate all of our billing process. That can be my prompt. I think most people can write that prompt, can't they? So you don't need a prompt engineer, but you do need a master of agents to understand what you're going to get out of that, where the limitations are, where it's going to break, what it's not going to see, where it's going to get lazy. Our AI VP of finance last night admitted it didn't fully read a contract. We asked it why. It's like, I don't have a good answer for you.
Okay. So being a master of agent is not throwing your monitor out the window when you hear that or calling it. It's like, okay, I get what happened. Okay. This is running on Sonnet. Sonnet rapidly goal seeks. It tries not to finish complicated behaviors. I need to work with the agent to change how we do it and make clear all contracts must be read from beginning to end. And then I have to be comfortable with it will still not do it sometimes. Right? So I know I'm rambling. That's a type of, but in a year maybe we won't, like the models and the harnesses will get so good you won't have to do that anymore. Right? So what that means is going to change. And even this whole idea of folks talking about loops and agents looping is like an early view of where everything's going to go because if your agent is constantly looping and improving itself in the background, which is already happening, it just fundamentally changes the way we build agents. They're not static. They're looping. They're constantly improving themselves in the background. Again, when we started this show, you needed to be a prompt. What they would joke, this was the highest paying job out of college was prompt engineers were making $150,000 out of college because they knew how to write a prompt. That skill's worthless today. So I think Brandon is right for sure. It just maybe it'll be really interesting what skills it takes to be an agentic expert in three or four years. It's just the rate of change here. It's just so crazy that with a very mediocre prompt, we could build an AI VP director of finance from China.
So we have a lot of early stage founders that listen. I saw this brilliant tweet that I did actually want to discuss with you guys and it was from Nicholas Desen at Y Combinator, formerly founder of... well, and he said, a most common reason I see good companies fail to raise their Series A/B right now isn't growth, it's margin. I keep meeting founders doing real revenue and growing fast but once you remove delivery costs there's almost nothing left. Investors don't fund revenue, they fund the margin on it. If that's you, fix the unit economics first. Fast growth on revenue you don't keep is a trap. That was the tweet. I like that. We've looked at Alolia back there. I think it's awesome. I disagree with him on this.
Look, I'll go further than that. The objective reality is that's not what was happening. In other words, companies with tough gross margin profiles and hyper growth have been getting funded and frankly have been able to improve their gross margins and pull it off and you can, the foundation, that describes the foundation models. It describes the inference providers and it describes the coding agents. So historically what he's saying is not correct. Now he may be picking on something in real time which is that ability to build a company with a tough gross margin profile and then fix it over time probably makes the most sense in that kind of big bang stage of AI which I think was the last 3 years where it went from nothing to something and you had 10x growth and it just paid you to grab the ground like it paid Cursor to grab the ground. It may be that investors now are saying, hm, as a first generation coding environment, you can be Cursor, you can have negative gross margins, you can still be worth 60 billion because you just grabbed the space. It may be that next generation they're seeing, because they see a lot, obviously given the volume, they're seeing a little more focus on gross margins now which is plausible but there's no doubt that to date the bet that my gross margins are [__] but my growth will cover it and I will figure it out even though it sounds stupid when you say it has in fact been 100% true.
I think he was synthesizing all the learnings across the Y Combinator portfolio and this change, right, is it okay for startups to have negative gross margins or no clear path to positive margins because they'll figure it out. That's Rory's point and I think you could argue both sides to Rory's point, but maybe that age is ending partially, right? Maybe it should end, right? You know, we're sitting here and Menlo just raised a $3 billion fund after 50 years on the back of a crazy bet on Anthropic that went very well, right, when the margins were crazy. But maybe that era doesn't last forever. And I'm sure all three of us are sitting on a portfolio company investment we made in the last 12 to 15 months where inference was the marketing strategy and the gross margins were negative and we're sitting here today and we're like, company's doing okay but I'm not sure we're going to get right side up on that investment. We're all sitting with a couple investments like that and then we see a few others that are wildly efficient. They took advantage of AI in other ways that's wildly efficient and we're wondering in that two by two, do we really want to be in the bottom, is it the bottom left, right, highly inefficient and not better growth, not top 0.1% of growth? I don't think we want to be in that two by two. Do we? Everyone will have some deals where you go, we thought we'd earn our way out of the gross margin problem and for whatever reason we couldn't either because we didn't grow quickly enough to get to scale, fail, or because the foundation model company started grinding us or competition bundled it in and now you're just in a shitty gross margin profile business which will go bankrupt.
Because when things slow down, no one says, 'Hey, let's do an acquisition of an adjacent gross margin negative thing and it's going to be fun,' right? So...
Yeah, that's all these startups will just fail.
Right. If they don't achieve massive growth, they will all fail. I think that's the good simplification, right? And so it's just for an early investor, that's not that fun. That outcome is not that fun.
Jason, you mentioned Menlo. Obviously one of the biggest winners from Anthropic raised $3 billion. I'm sure we've all got friends at Menlo. Awesome, good people. Happy for them. They're also in Lovable. They've been some of the best AI ambassadors of this wave. I think categorically the interesting thing, honestly, for me, and you're going to kill me for this, guy's child of the boom, whatever, why is it not more? They legitimately are one of the best. They'll deliver billions and billions of dollars back in the wake of Thrive and GC and Lightspeed. This three billion's pretty conservative. I thought that, I actually don't have an answer.
The only answer I could come up with, right? I do think, look, not only is Menlo wildly successful, but they're omnivorous. They'll invest at every stage. They will take lead positions, but they will also do smaller checks and small positions, right? And so they'll, you know, if they lose the round, they'll do 2% into the round or maybe they'll invest more later. Maybe they won't. Like if it's a great one, they'll just do it, right? Kind of like how Felic's got going. It's, you know, any size like 20%, 1%. So my first answer is that that's all they could raise. Poor guys could only raise three billion. Okay, this was my first read. Rory's shaking his head, the professor. And listen, I don't know. I just shouldn't have gone first with the answer. But I think my secondary read is, look, they'll just raise another 27 billion in other vehicles like they did for Anthropic. Like this is just one or two funds and they'll raise another 10 or 20 billion in SPVs or sidecars or other funds. The headline number of the fund size is not always correlated to the amount they'll end up deploying over the life cycle of those investments, right? I think they were smart, right? I think the interesting...
Yeah, to raise that amount because yeah, given their amazing performance, they'll always have access to more and by not starting the clock on a much larger growth fund, you know, with the fees and the drag that entails, you just set up for success. I mean, remember all other things being equal, provided you've got a big enough check size to play in any round, as a GP you actually quote unquote have higher risk adjusted return if you have smaller funds because then you have less cross deal aggregation. In other words, two separate $1 billion funds versus a single $2 billion fund. All other things being equal, you have more probability of winning on at least one of the two funds than on the single two billion fund. So that's not the reason by the way they did it. But just to point that out, I think the zoom out question is this. Two is a lot of money. Like how many deals are there going to be like Anthropic or OpenAI or Anduril where you can put one or two billion dollars to work? And the question is do you create your main vehicle for that or do you accept that there are anomalies and you know you have access to that capital via SPV. So you run your business on what you think you can deploy across normal cycles on normal deals including wildly successful, you know, 10 billion, 20 billion outcomes and then you accept that, you know, two or three times a decade there's going to be a trillion dollar outcome and you want to have access to that capital but you can get that via an SPV. I think it's a reasonably rational structure.
