🤖 Frontier AI labs keep durable premium pricing (not commoditized)

“What's going to get legislated is the quid pro quo. The quid pro quo is this. The government gets more power, and the duopoly of closed model companies get more regulatory capture or moats around their business. That's going to be the quid pro quo.”

“Why? Because if you constrain the amount of assets that can actually generate compute, and a smaller surface area of it is available to the leading companies, those leading companies will have to pay more, the user experience will decay, and then other products can have time to catch up.”

“Well, in the face of new evidence, you would have to say that they are lowering the prices and the opportunity in the enterprise is so great that they're not being dramatically impacted by it.”

“But I think what Claude is showing is that they have such a high margin. They actually have good elasticity on pricing. So they could push price down, and you could start to see them start to segment the market a little bit, where they could start to be more competitive, perhaps, with cheaper tokens for cheaper, easier tasks, and then premium tasks charge a higher price for intelligence, and start to stratify the value of the intelligence that they're providing by changing the price points.”

“And I think it is a little bit of a refutation to all the people who have been saying over the past year that they're getting commoditized away, that open models are basically are putting them on a hamster wheel where they're basically going to zero on gross margin, or that it's a business that only works because it's heavily subsidized by venture capital. In fact, I think there's a lot of reasons to believe that Anthropic has a pretty big lead. And if you believe that recursive self-improvement has kicked in, as they are claiming it has, then their lead could actually grow.”

“We can only get that with open-source, open-weight models. We can interrogate chain of thoughts. That's kind of one reason we're gravitating towards them, as opposed to getting caught into the closed race.”

“the rose-coloured box is LLM traffic. That's the typical spend you're going to spend on the big frontier labs, closed models effectively, either through things like Bedrock on AWS or what have you. What's really interesting is that the blue box is taking up. The blue box is people buying their own GPUs because they're moving to open source.”

“So they definitely have a duopoly when it comes to revenue and market share. I would argue they have a duopoly when it comes to frontier capability.”

“Thirty times price to sales ratio, Sacks, and this headwind of open source and enterprises moving to that. Most businesses don't need to solve the last seven math problems in the world.”

“Whatever you're holding in the equity markets, whatever you're holding, OpenAI and Anthropic are threats to that. There's almost nothing that you can hold in a portfolio that those two companies don't threaten.”

“Elon announced, hey, Grok Bot, which is a fantastic product, I highly recommend you try it out, is going to pick the best end-to-end model. Does it matter whatever it is? Quote Elon Musk, whatever is the most likely to give you the best outcome is what they're going to use.”

“you will have certain models emerge who may not be nearly as good in frontier capability, but they simplify the moral ambiguity. And that may be the better decision to make from a business perspective and from a user perspective”

“Assuming model convergence, and even if models have slight advantages over others, it seems to be the case that we will maintain pacing.”

“I think that Anthropic and OpenAI right now are a stable duopoly for frontier intelligence because they have a pretty significant lead over all these other companies, and they are able to charge a premium for that.”

“the vertical language models, the small language models are going to do 80 or 90% of the tasks that many corporations need”

“there's a stat that I think in the last 12 weeks alone, token use has flipped from 80-20 closed versus open, to 80-20 open versus closed.”

“It'll force OpenAI and Anth to go up the stack. There is no choice. You can't sit there and serve a token that has diminishing value because then all the subsequent value is getting absorbed and made by people wrapping your intelligence token.”

“as long as we have a closed model duopoly of Frontier AI, and they're reserving the right to get into every vertical application there is, they're declaring in advance they're going to compete with their customers. So how can we trust them with our data?”

“A lot of these AI application companies are now moving off of the frontier models, and I think that's a headwind.”

“if you believe that the information that you have is critically important, you cannot use these services the way that they're currently offered by most people.”

“open source is what drops the cost of AI by 50X and makes it available to everyone so everyone can benefit from AI. No one gets rich. We all get rich with open source.”

“the market structure that we have right now is there's two markets ... And then there's Commodity Intelligence, which is everybody else, including all the open models. And they're not competing on being at the frontier. They're competing on price”

“these capabilities will be matched or close to matched by a handful of closed-source alternatives and open-source alternatives, probably within the next three or four months ... we are seeing the cost of that incremental unit of intelligence being driven further and further down”

“I think Jensen is going all in on open source based on these two acquisitions, Hugging Face and Poolside... He is going for it. He is going to own open source. And that is a really interesting moment in time for us to look at, because we were sitting here six months ago, 12 months ago, open source would never catch up.”

“I happen to believe, having looked at this and thought about it, I think that it is possible. I think it is likely to happen soon. People I've talked to in industry do think it's real. ... which is why I think RSI is probably the future.”

“wrap it in its own harness, it decays in capability. When you take an open source model and you use any other open source model, it improves ... there are all kinds of ways of using open source technologies that are meaningfully more performant and dramatically cheaper than the closed source alternative.”

“can orchestrate these open source 150 IQ intellects. ... 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.”

“I actually think that their pricing power may be quite sustainable, ... a decent subset of the market is willing to pay a premium for true frontier intelligence. ... their lead is growing because they're on the way to”

“You just look at the pricing here of GLM 2 versus Claude Opus, and it is 90% cheaper ... increasingly, corporate America is going to embrace these solutions. ... open-source is the safer bet for us. And that's what's happening in corporate America today and inside of startups.”

“Elon comes out and says, not so fast. We're entering the singularity and the frontier models are way further ahead than people think. I believe that to be true. ... Jensen came out this week and said, closed models are actually cheaper. ... explains why they continue to run away with it on the revenue side”

“the difference between the open source models I'm using ... the Frontier models are not necessary anymore. They're just not necessary. ... I'm still going with open source and Gemini being the leaders in this space.”

“the market for frontier intelligence has become a duopoly. I think it's a very powerful duopoly. I don't think it's being commoditized. ... Anthropic is now over 80 billion of ARR. Started the year at 10, ... people are willing to pay a premium for true frontier intelligence.”

“You're going to see some of the major customers of Anthropic and OpenAI — the eight and nine-figure customers, people spending 50 million, 100 million a year — they're leaving... Eleven Labs, Figma, Lovable... they're all working on their own models... my team has installed Kimi. It is 80% or 90% cheaper already... [and] a friend of ours in the inference space has a customer who just moved nine figures off the Frontier Labs onto GLM 5.2... that's going to be a massive headwind against these companies, massive.”

“Peter Thiel once said that monopolies pretend to be commodities... the market for Frontier AI is already a duopoly... the AI duopoly has a big incentive to promote anything that suggests that they're not actually in complete control of this market, but I think they are... Anthropic breaking into 70 plus billion of ARR [with 80%+ gross margins improving], OpenAI did more net new ARR in July than all of Q2... [per Dwarkesh] demand 10x-ing while compute builds at 3x means compute prices rise — only the most lucrative algorithms can afford to compete for compute. It's a self-reinforcing flywheel... like Apple and Android — Android got the share, Apple got the monetization.”

“These models are getting commoditized much faster than anybody thought... there is no meaningful sustained advantage once a model publishes their performance criteria — within weeks, other models match and in some cases exceed the performance... I've never seen a sector absorb hundreds and hundreds of billions of dollars [where pricing power] effectively evaporates in months... the real business model is not in the foundational model anymore. It's at the application layer above and it's in the infrastructure below... if you're going to build a business model that tries to ascribe this layer as having a lot of terminal value, that is a mathematical mistake.”

“Zucks is about $1.50, Elon is about $1 and the Chinese models are 50 cents. ... on top of the whole data sovereignty, bleeding your alpha away, can you imagine that you're paying 56 bucks as well, per million input tokens for that risk? That is insanity. ... it doesn't take a company that's much, much worse than you to beat you when your cost is 50 to 100x more.”

“the share of wallet, is actually increasing to the frontier labs, while the share of tokens, ... The difference between spending 3 bucks on a cheap model or 15 bucks on an expensive model to replace a $200 an hour consultant, ... The non-consensus argument might be that intelligence is not converging at all, ... as it becomes recursive, you actually extend the lead, ... the distance between the frontier and everybody else doesn't converge, it actually extends,”

“Open source went from 19% last year to 11% this year [of enterprise AI spend]... anyone who's saying that these closed models are going to lose or are somehow losing, you're just not seeing it in the data... this is a case of the spirit is willing, but the flesh is weak... it is just phenomenally convenient to go with the frontier labs... there's only two companies making meaningful revenue... I could see AI easily becoming another tech market that becomes a duopoly.”

“when you use our harness with Claude, it was simultaneously 1.4x cheaper and 1.5x faster than just using Anthropic Opus 48 alone. But if you wrap the open-source model with our software factory, it was 16.4x cheaper. ... if you are a reasonable company, why are you not finding an independent way to access this intelligence in a way that doesn't leak your edge away? To ... continue to make the same decision, I think is really insanely dumb.”

“how good GLM 5.2 is, has challenged some of my beliefs. ... I profoundly believe the future is composable models. ... half the queries are going to go to the open source model, maybe 85%, ... shift economic value from the margins of the frontier labs to the infrastructure. ... To date, frontier tokens are capturing 90% of the economic value. And open source tokens are probably 80% plus of tokens processed.”

“there's currently a natural market force for two things. One is more open source, and second is a fragmentation of the AI stack. ... I'm not sure that what we're looking at today as being some oligopoly of three or four businesses is really where we're going to stand. ... People are also going to run models locally.”

“there is no single best model anymore. At the top of the leaderboard, Opus 47, GPT-55, Sonnet 46, appear almost indistinguishable, separated by less than, ... these things are getting commoditized way too quickly and then you'd say, well, what's the ROI on all this incremental spend?”

“the Chinese models are as competitive as these US models, ... the highest end chips is kind of more of an ego gratification for us. We”

“these revenue ramps are happening while the world is already using open source. This is not frontier only. ... The TAM of intelligence is dramatically larger than any TAM we've ever seen”

“it's possible that if you're the early leader in coding ... you have the most access to code bases, you might get the most training tokens. There is a potential flywheel there, ... the market for coding tokens could be monopolized.”

“I believe open source is going to win the day on the large language models and take 90% of the token usage, and I think the entire frontier model space could be undercut by open source... SLMs, the smaller language models that are verticalized now, that will run on desktops... that is their biggest competitive threat, and I hope it happens.”

“the models and the agents on top of them, ... Anthropic at five or six billion dollars of revenue in a single month, ... the TAM is dramatically bigger than any of us think ... the frontier labs can do well even in the face of it.”

“these things are all kind of becoming the same. Eventually, they're all getting access to enough power, enough compute. They're generating similar results, ... even more important that you have a complexion of models.”