Open models and the frontier premium

All of these open source models, we started running them, picking up about 85% of our tokens right now. And every startup I know is saying, we are standing up our local models and we only use the top models, the paid ones, when we have jobs we can't do. So I'm just curious your thoughts on that.

I believe that this is going to derail their IPOs. I'm taking it from the top. It's going to derail their IPOs. It's going to be headwinds against it because I think that they're going to have massive margin compression. I believe they're spending so much money that I think they're going to get caught in a trap. I think this could be a trap for them. They overspend. They don't have the same profitability and it doesn't pencil out. And startups are the future. The startups are what eventually the enterprise copies.

And the time and the compute. So if an AI agent is replacing a $200 an hour consultant, right? Take that as an example. So three consulting firms, they're competing. They need the smartest consultant. They're charging 200 bucks an hour. The difference between spending 3 bucks on a cheap model or 15 bucks on an expensive model to replace a $200 an hour consultant, it's just irrelevance. That inference cost difference is irrelevance if you're getting something that's bulletproof for 15 bucks. And so I think that's what we're seeing play out.

I'm talking about what is Anthropic's revenue growth compared to OpenAI, compared to the open-source models. Millions of independent actors are choosing every single day. The open-source companies are growing, right? But they're growing selling something that is really, really cheap. And there's room in every single market for premium products, for mid-tier products, and for commodity products. And I think we see a lot of this token growth. People are speculating that the intelligence gap between that commodity stuff and the frontier stuff is going to collapse to the point that people won't pay for the frontier stuff. There is no evidence of that on the field today. It may develop over the course of the next couple of years, but it's not on the field today.

Moreover, Chamath, to your point, it's not only about the model, it's also about the harness, it's about the connectors, it's about the enterprise agreements. There's a lot of things here that you need in order to grow a business like this.

I mean, first I would say that I am very enthusiastic for open source. We see it in widespread use everywhere. But here's the interesting thing, for the advanced companies, they're doing some planning with the Frontier Labs and then they're kind of doing the execution, if you will, with the open source models. So they're running an ensemble of model strategy. But here's what I think is more impressive. We have incredible open source models nearly on the frontier. And notwithstanding that, we're seeing companies like Anthropic at five or six billion dollars of revenue in a single month, which is extraordinary. We've never seen anything like it in technology. And that's in the face, Jason, of all this open source usage. So what does it tell me? It tells me that the TAM is dramatically bigger than any of us think that it is. And that, you know, when we when we look back on this period, you know, that will be the big takeaway. It's a takeaway with Uber, the takeaway with Google, the takeaway with Amazon. The TAM was way bigger. We've crossed an important threshold. Open source will be a part of it. But clearly, the frontier labs can do well even in the face of it.

There's two points that I think are really important. Open source makes frontier tokens more valuable. So like, let's just say, let's use an example. Let's say that, you know, Anthropic, OpenAI and Grok, they're at like a 250 IQ or 200 IQ. Well, if you have a 200 or 250 IQ intellect, they can orchestrate these open source 150 IQ intellects. Like that, that's actually more, that makes the 250 IQ intellect more valuable, is because it can farm out tasks. And just in the same way, you know, if we're going to use the Manhattan Project analogy, you know, everybody, you know, talks, you know, talks about Richard Feynman. And you know, who's the guy who ran it all? And you know, there's these 10 really smart physicists. Who's the guy who ran it? You know, who said, I am become death? Who's the guy who ran the Manhattan Project? Come on, Oppenheimer, Oppenheimer. You see at Oppenheimer and even you had 10, you know, effectively Nobel laureates, maybe 30 of them. But they needed like the 2000, you know, merely, you know, top 10,000 in the world physicists. So having more having more affordable intelligence, it actually makes the frontier even more valuable. And I think a very plausible outcome is one which frontier tokens are 65 to 85% of the economic value. Open source tokens are like 80% of the volume. And that's good for everyone. And everyone wins.

It's also okay for Anthropic to say, you know what, I have this next generation class of model. I'm going to instantiate, I don't know, a life sciences program, a cybersecurity business. That's where they can capture, as Sacks was saying before, over time, trillions and trillions of enterprise value, because I do think they have excellent models and they have excellent engineers and they have momentum. But if you're going to build a business model that tries to ascribe this layer as having a lot of terminal value, I think that that is a mathematical mistake. That's it, you can't do it.

And it does provide a bunch of advantages over closed source, right? You get customization, you get control, you can run on your own hardware. You don't have to worry about the data problem, you know, your alpha getting leaked to these companies that might compete with you, all those types of things. And the market is so big that I'm sure we will see some success with open source. It will take a meaningful chunk of the market. But if you're looking at where the revenue is right now, it's these two companies. It might end up being a situation like Apple and Android, where Android got a lot of market share, but Apple is where all the monetization was. Yeah. And I think that Dwarc has raises a really good point that as the demand is 10x-ing year over year, but the compute can only be built out at, say, 3x year over year, because just all the friction of all the things in the real world that get in the way, permitting, regulation, bans on new data centers, all that kind of stuff, I think that the price of the compute is going to go up, and that will provide an advantage to the models that have the most lucrative algorithms, that are able to produce the most intelligence per watt, or the most intelligence per token, or per GPU. And right now, that is those two companies. And in a way, you could say they have a self-reinforcing loop, because if you have all the revenue, and right now, like I said, it's just two companies of all the revenue, you can then plow that money back into the next training run. Right? So that's the flywheel here.