Anthropic just released its S-1 form, which contains the official financial information it must report to the SEC in order to go public. From the S-1, we now know that Anthropic's revenue grew more than 100x (specifically, 112x) from 2023 to 2025. Revenue was:
- $41 million in 2023
- $386 million in 2024
- $4.59 billion in 2025
A few people in the EA community apparently believe this revenue growth rate will continue for the next several years at least. However, I am confident we won't see Anthropic bring in $514 billion in revenue in 2027, which is what extrapolation would imply.
Indeed, I'm confident that Anthropic won't hit half of what extrapolation would imply, which is $257 billion in 2027.
I'm willing to bet as much and would be delighted to wager $20 on it, to be donated to the charity of the winner's choice.
I also have an open offer for a more general AI bubble bet.
Update (Friday, October 30th at 5:45pm Eastern):
Here's a good article on this topic. Two economists look at Anthropic's S-1 and ask whether a $2 trillion valuation is reasonable: https://econjared.substack.com/p/the-fantastical-promises-of-a-2-trillion


After much ado six months ago in and under a featured post about Anthropic's revenue growth rate, lo and behold, Anthropic's growth rate has slowed significantly, at least in the most recent quarter:
Source: Reuters.
Six months ago when I said Anthropic's growth would eventually slow down, a few people seemed pretty dang certain I was wrong. When they tried to articulate their reasoning, it did not make sense to me. (Nor do I think it would make sense to the vast majority of financial economists or equity research analysts.)
If someone wants to double down and say this is just a temporary blip, I'd be happy to bet that Anthropic's revenue growth in future years (e.g., 2027, 2028) will be slower than the growth from 2023 to 2025.
Saying that Anthropic will have a revenue growth rate under 10X per year is not an interesting statement. To maintain that growth rate given its current revenue of more 33 billion in 2026 so far, it would need to make 330 billion in 2027 and 3.3 trillion in 2028. That would be triple the highest revenue company in the world right now. If people believe that Anthropic will continue to have that kind of growth rate, they can & should just buy ANT rather than make a bet with you.
Now, I think there is a possibility that ANT has a 10X revenue growth over the next year. But in those worlds, my concerns of AI being a massive societal deal will be obvious, and a bet with you would prove nothing.
In the worlds that ANT has a 4X revenue growth over the next year, AI is 'still' a massive deal.
If you're bearish on ANT, please go ahead and short it. Here's a link on how to do so. https://graniteshares.com/research/how-to-short-anthropic-stock-ant/
The purpose of the bet is not the money. Bets are not legally enforceable (if they're even legal in the first place), and people who lose can just decide not to pay up if they don't feel like it. Obviously, bets are inferior to securities markets. The purpose of a bet like this is to get people on the record with their predictions. All I want to do is bet people $20 for charity. If people want to bet for zero money, that's almost just as good.
You're veering off-topic into different, adjacent debates that are not directly about the disagreement in and under the linked post. Please read that discussion if you don't know what I'm talking about.
There are endless different disagreements people have about AI; this bet about Anthropic's revenue growth would not attempt to resolve all of them. It would only attempt to resolve one specific, narrow, quirky argument made in that featured post and in the comments underneath it.
For instance, Josh You from Epoch AI wrote:
What's wrong with this argument? Fortunately, I can answer in the form of a cartoon:
The sort of blind extrapolation of graphs that people like Stefan Shubert, Josh You, and others argued for in and around that post just doesn't make any sense to me. It makes as little sense to the vast majority of financial economists and equity research analysts, who find blind extrapolation just as dubious as I do.
If you agree that blind extrapolation of graphs makes no sense, then we agree at least on that point, even if we might disagree about other things.
I did not endorse forecasting frontier AI revenue via naive extrapolation, as I clarified in email correspondence around that time. My Forum comment at the time pointed out that the robustness of frontier AI revenue growth since 2023 was good evidence that this revenue would continue to grow rapidly, but not necessarily at a constant rate, such that a multi-trillion revenue run rate (implying significant social/economic impact of AI) seems likely by the early 2030s. E.g. 5-7 doublings from the mid-2026 run rate would be around 2 to 8 trillion dollars, and this doesn't require naive extrapolating the roughly two-doublings-per-year annual growth of frontier AI revenue since 2023.
I currently believe, very roughly, that Anthropic will ~double its revenue run rate over the next 12 months if model progress does not accelerate from its current pace of progress, though my unconditional forecast is higher than that (not going to put a number on it) since I think model progress is more likely than not to accelerate.
Hi Josh. Unless I'm misunderstanding you, it sounds like you're still reiterating the same sort of argument I had a problem with originally, which is satirized in the XKCD cartoon above, and which I believe most financial economists and equity research analysts would reject.
It's one thing if this forecast is based on an underlying causal explanation of Anthropic's revenue growth. For instance, maybe you believe — based on other evidence like the METR graph or AI 2027 or Situational Awareness or Epoch AI's work or other things of that nature — that LLMs are on a trajectory toward AGI, and it's mainly a matter of scaling data/compute or figuring out recursive self-improvement. That's a theory, an explanation, that stipulates a cause — rapidly increasing general AI capabilities — for an effect — rapidly increasing revenue. I think you do have an underlying causal explanation, since you talk about AI model progress and whether you think progress will accelerate or not.
But what is Anthropic's revenue growth actually itself evidence for? Is Anthropic's past revenue growth "good evidence" that "a multi-trillion revenue run rate" will be "likely by the early 2030s"? No, Anthropic's revenue growth is not good evidence of that, not by itself. The past rate of revenue growth is not good evidence for the future rate of revenue growth. By itself, it tells us very little. We need a causal explanation of what's happening to forecast what's next.
Another XKCD cartoon called "Extrapolating" satirizes this same reasoning error:
If you want a good indication that equity research analysts, institutional investors, and so on disagree with your reasoning, just look at Anthropic's current market cap: about $1.2 trillion. You're predicting at least $2 trillion in revenue by 2033. At the current revenue multiple for software companies of about 8-9, that would imply a valuation in 2033 of about $17.5 trillion.
If we apply an aggressive discount rate of 12% over the next 7 years, that gets us to a current valuation of about $7.9 trillion. Yet Anthropic's market cap is only around $1.2 trillion right now. You're off by about $6.7 trillion, or roughly 5x.
Why are the equity research analysts and institutional investors not convinced of your reasoning?
I asked an equity research analyst I happen to know and they agreed with me that this is just not a reasonable way to forecast revenue. That's the answer I expected.
You might respond: ah, ah, ah, but the world of finance knows nothing about AI! And maybe so. But we're not asking here what certain causal, explanatory theories or models about AI — based on non-financial evidence — would lead us to predict about AI companies' revenue. We're asking what the revenue itself is evidence of. And the answer is: by itself, not much.
Maybe you think it's strange that I'm stuck on this one narrow point. The reason is that there just aren't that many arguments or that much evidence offered to support predictions of AGI or transformative AGI — or anything close — within the next 5-10 years. This odd argument about extrapolating revenue was one of just two arguments offered in a post that the EA Forum staff decided to feature (or curate) as a high-quality post everyone should read. A lot of people seemed to agree with the post, and no one else pointed out that the argument about the revenue didn't make sense.
If this is supposed to be an example of high-quality argumentation endorsed by the EA community for near-term AGI or transformative AI (or something close), it's worth pointing out that it doesn't make sense, and in fact it's an example of such a well-known reasoning error that it's satirized in cartoons. I've caught subtler errors before that some people missed and some even initially denied, before the people who made the errors ultimately acknowledged them. This is a case where it would be helpful to talk to outside experts like financial economists or equity research analysts, who might hold more credibility for you than I do, and who might stand a better chance of convincing you.
Source: Reuters.