Anthropic’s original IPO prospectus framed the company as both a commercial AI contender and a public-benefit project, asking investors to look beyond an ordinary transaction toward what its founders called “something greater.”¹ That lofty appeal came with unusually blunt disclosures: advanced AI could bring major gains, the filing said, but also catastrophic or existential risks.

On Monday, an amended S-1 supplied the hard numbers. Anthropic reported roughly $17 billion in third-quarter revenue and $33.24 billion for the first nine months through September, about 1,200% above the year-earlier period. It also reported a $104 billion net loss, largely driven by changes in the value of convertible notes and derivative liabilities.²

The filing’s central tension lies in the price of that growth. Anthropic disclosed $895.4 billion in aggregate purchase and lease commitments, “substantially all” tied to compute and infrastructure.³ About $447.5 billion is expected by the end of 2030, with nearly $70 billion due in the 12 months after September 30, 2026; the remainder stretches through 2045. The company held $131.8 billion in cash and short-term investments, a formidable cushion but a fraction of its long-dated obligations.⁴

Supporters can argue that the business has earned the right to spend: it targets enterprise customers, and revenue jumped 47% quarter on quarter. But the growth rate itself has slowed sharply from the previous quarter, while investors still cannot see how much future compute is destined for serving customers versus training the next models.⁵

That uncertainty fuels the skeptical reading. One analysis of the filing put it simply: “Growth is not optional for this company. The contracts assume it.”⁶ Anthropic’s bet is that stronger models, falling unit costs and durable corporate demand will outrun an infrastructure bill unlike anything a newly public company has carried before.