The most striking figure in Anthropic’s revised IPO filing isn’t the revenue. It’s the $895.4 billion in purchase commitments and lease agreements, almost all of it for compute. That is a bet on the future that dwarfs anything a software company has put in a prospectus.
According to CNBC, the amended S-1 adds third-quarter numbers that the first filing lacked. Anthropic made about $17 billion in Q3 and $33.24 billion over nine months, up roughly 1,200% year over year. The nine-month net loss was $104 billion, though most of that comes from changes in the fair value of convertible notes and derivative liabilities. Q3 alone showed a $23.29 billion loss. Those are accounting-heavy figures, so read them with care.
The number I’d watch is growth. Q3 revenue rose about 47% from the prior quarter, down from 144% in Q2 and 132% in Q1. That is still enormous for a business this size. But it is a clear deceleration, and a roadshow audience will ask whether the curve is flattening just as the spending curve goes vertical. Compute and infrastructure costs rose $14.4 billion over nine months, up 329%.
The timing is awkward. Markets are already nervous about the AI buildout. Rising rates threaten the debt that data center developers and neoclouds depend on, and last week’s earnings from Microsoft, Alphabet and Amazon added to the worry that their heavy AI spending is squeezing cash flow. Anthropic gets much of its capacity from Amazon, Google and SpaceX, with neoclouds filling gaps. If those suppliers can’t build fast enough, Anthropic’s commitments become a constraint rather than an advantage.
For developers and founders building on Claude, this matters beyond the stock ticker. A company carrying this much committed spend needs sustained demand, which could mean pricing pressure, aggressive enterprise sales and tighter rate limits as capacity gets rationed. It also means Anthropic is now far more exposed to infrastructure partners than a typical model lab.
The competitive picture is simple. OpenAI, founded by the people Anthropic’s co-founders left, filed confidentially in June but is signaling a 2027 debut. That gives Anthropic the chance to set the public-market template for AI labs. It also means Anthropic takes the first round of scrutiny.
The filing also runs more than 90 pages of risk factors, covering political headwinds, infrastructure complexity and the tension between safety and commercial goals. Anthropic’s co-founders wrote that they consider it their “personal responsibility” to help the world understand these risks. That fits the brand, and it is also good legal cover.
So is this noise? No. It’s the first audited look at the economics of a frontier lab, and the picture is fast growth, slowing momentum and a gigantic bill. Investors will decide whether that is a moat or a trap.
