Anthropic Says It Made 11.5 Billion Dollars in Three Months, and Turned a Profit
Preliminary Q2 figures shown to investors put revenue at more than fourteen times the same quarter last year, with positive adjusted operating income. An IPO may follow in the autumn.
Anthropic has told investors that preliminary revenue for the second quarter of 2026 came in above 11.5 billion dollars, according to Bloomberg. For scale: the same quarter in 2025 brought in 787 million, and the first quarter of this year brought in 4.73 billion. That is a fourteenfold jump year on year and more than a doubling quarter on quarter. The company also reported positive adjusted operating income, which would make this its first quarter in the black.
Two caveats belong in the same breath. These are preliminary figures shared with investors, not audited results, and “adjusted operating income” is a number companies define themselves. It typically strips out things like share-based compensation and one-off costs. It is not the same as profit in the sense your accountant would use. It is still a meaningful signal, because for the past three years the standard assumption about frontier AI labs was that they burn cash indefinitely.
The timing is not accidental. Morgan Stanley, Goldman Sachs and JPMorgan Chase are reportedly preparing a possible autumn listing, and Anthropic crossed a 47 billion dollar annualised run rate back in May. Numbers like these are exactly what you publish when you are walking into IPO meetings.
What is actually driving it? Overwhelmingly, business customers rather than consumer subscriptions. Coding agents in particular have turned into a real line item for software companies, and Anthropic’s own risk report noted last week that Claude now writes a majority of the code merged into its production systems. When a tool moves from “nice to have” to “part of how the work gets done”, the revenue stops looking like a fad. The open question is durability. Model quality gaps close in months, not years, and a customer who switched to Claude because it was best at code can switch away just as easily. Growth this steep also invites the obvious comparison to the last few technology cycles, where enormous revenue and enormous spending arrived together and only one of them turned out to be permanent.
What this means for you: as a user, expect the paid tiers to keep getting more capable and the free tiers to keep getting squeezed, because that is the shape of a company optimising for business revenue ahead of a listing. If you run a small business and you have been treating AI assistants as an experiment, note that the spending is coming from companies that stopped experimenting. And if you are watching the AI bubble question, this is one real data point in a field full of projections: at least one lab is now covering its operating costs. Whether the rest can is a different question, and nobody has answered it yet.
Sources
Source: https://finance.yahoo.com/technology/ai/articles/anthropic-revenue-surges-over-11-210857853.html
Debian Is Voting on Whether AI May Touch Its Code
One of the oldest Linux projects has eight options on the ballot, from a full ban on LLM-assisted contributions to a permissive framework. Voting runs until 28 August.
