Anthropic has filed the most revealing document of the AI boom so far, and it reads like a company trying to warn investors about itself. Amendment No. 1 to the Claude maker’s IPO prospectus, filed with the US Securities and Exchange Commission on November 2, fills in the blank the original filing left empty: Anthropic has committed about $895.4 billion in spending, nearly all of it for computing power and infrastructure. That is not a typo, and it is not a forecast. It is money the company has signed up to spend.
The revenue side is just as startling. Anthropic took in $17.0 billion in the third quarter of 2026, up 47 percent on the second quarter, bringing nine-month revenue to $33.2 billion — a 1,206 percent increase on the same period a year earlier. The company also reported a third-quarter net loss of $23.3 billion, driven mostly by accounting changes on the value of convertible notes rather than cash going out the door, and a $567 million operating loss after a $974 million operating profit in the second quarter. Growth is real, and so is the cost of sustaining it: compute and infrastructure costs climbed 77 percent in a single quarter.
Break the $895.4 billion down and the shape of the AI economy comes into focus. At least $276.2 billion of it is minimum spending promised to just three cloud providers — $121.3 billion to Google, $110.0 billion to Amazon and $44.9 billion to Microsoft — owed whether Anthropic uses the capacity or not. About $447.5 billion falls due before the end of 2030, with most of the rest spread between 2031 and 2045. Around $213.8 billion of the total is tied to arrangements that can still be terminated or depend on conditions like a partner securing financing. The filing also discloses developer guarantees of up to $4.7 billion.
Then there is the sentence that has nothing to do with money. Anthropic writes that its internal estimates suggest future models “may imminently” meet its own definition of dramatically accelerated progress attributable to AI — the point at which AI starts speeding up research on AI itself. The company is not saying it has crossed that line. It is saying the line is close enough to name in a legal filing, alongside a note that more than 90 percent of contributions to its codebase in September 2026 were written by Claude.
Here is the honest read. Anthropic’s original IPO filing a week ago was already the largest AI listing ever attempted; this amendment makes the bet explicit. The company is borrowing against a future in which demand for its models keeps growing fast enough to justify nearly a trillion dollars of infrastructure — while telling investors, in the same document, that the technology may soon start improving itself in ways nobody can fully predict. Both things can be true. That is precisely why the filing matters: it prices the boom and names its strangest risk in the same breath.
What is still missing is almost as telling. The filing leaves the share price, the number of shares and the listing date blank — the cover page still reads “between $ and $” — and gives founders’ company majority voting power, meaning Class A shareholders generally will not be able to change the outcome of a vote. A roadshow could begin as early as November 10, with a listing the week of November 16 reported but unconfirmed. Until those blanks fill in, the $895 billion figure is the story: the AI industry’s biggest company-to-be has just told the world what its ambitions actually cost.
