What Anthropic’s IPO Filing Reveals

Anthropic’s roughly 400-page S-1 depicts a company scaling at extraordinary speed but uncertain whether growth can support its long-term commitments. It reported $16.3 billion in revenue in the first half of the year, compared with $4.6 billion for all of 2025. The filing leaves third-quarter revenue blank, so it does not confirm the reported $16.8 billion figure Anthropic gave investors; it does disclose about 45% quarter-over-quarter growth in Q3 and cautions that roughly 1,000% year-over-year growth will not continue. More than 90 pages of the prospectus address risks.

Profitability depends on the measure. Anthropic reported a $974 million operating profit in Q2, its first profitable quarter under DAP accounting standards, and an 85% gross operating margin in the first half, up from 40% in 2025. Yet its operating margin was negative 4.3% as research, sales, and administrative costs increased. The reported first-half net loss was $80 billion, including $72 billion attributed to a convertible-debt accounting change rather than a cash loss. Anthropic spent $19.2 billion on compute since 2023, including $8.9 billion in the first half, and said about 70% of its compute budget went to internal work. Claude now writes 90% of its code; in April, internal Claude use valued at API prices cost more than payroll.

The largest investor concern is the scale and duration of Anthropic’s infrastructure obligations. The filing discloses $417 billion to $431 billion in non-cancelable future compute commitments, with total signed commitments above $600 billion and estimates potentially reaching $800 billion. They extend over more than a decade, with some contracts lasting 20 years; spread over time, the host characterized them as roughly $20 billion to $40 billion a year. The founders called the spending “bold but sane,” while recognizing that it could initially look irresponsibly large. Anthropic warns it may not secure enough compute, that costs could rise, and that Google and AWS might not remain aligned with its interests or honor their contracts. It is beginning to pursue its own data centers to reduce dependence on partners that could become ruthless competitors.

Anthropic reported $132 billion in cash and cash equivalents against $141 billion raised over its history, and $1.1 billion in first-half interest income. The transcript said cash was expected to carry the company through year-end, but operating income alone was unlikely to fund its infrastructure plans; the S-1 warns that financing may be needed sooner than anticipated. Funding arrangements include $6.2 billion in convertible notes owed to Amazon, the ability to borrow another $77 billion from Amazon, Broadcom, and Google through promissory notes, and a $15 billion Morgan Stanley revolving credit facility described as emergency borrowing. Anthropic also established a $35 billion Broadcom financing vehicle in June and signed a new $63 billion vehicle with Broadcom, both structured with Apollo and Blackstone to finance TPU purchases. The filing additionally warns that no single insurer, and possibly not the insurance industry as a whole, has enough capacity or expertise to cover its exposures.

Customer growth offers a counterweight to the commitments. As of September, more than 11,900 customers spent at least $100,000 annually, over 1,900 spent more than $1 million, and about 250 spent over $10 million. Net revenue retention was 900%, meaning the average customer had increased spending ninefold year over year; the count of customers spending at least $100,000 had grown from about 1,500 the previous December. The top five customers accounted for 21% of revenue, down from about 36% at the end of Q1, although 48% of revenue flowed through distributors such as Google and AWS. Anthropic said its services represented less than 0.1% of a Fortune 500 customer’s expenses. Aaron Levy suggested Fortune 500 spending could eventually reach 10 to 50 times its current level, or 1% to 5% of spend; Semi-analysis called the expansion “likely the most efficient scaling of a business in the history of capitalism.”

The filing estimates global knowledge work at $40 trillion in 2025 and cites an IDC estimate of $22.5 trillion in cumulative global economic impact from AI investment by 2031, but offers no clear total addressable market figure. It names OpenAI and Google as its few direct competitors. The founders pledged to keep Claude free of advertising, arguing that ad targets could pressure the company to optimize for engagement, a position the discussion linked to its reliance on enterprise adoption. Founders’ Class F shares give them 50.1% control over key decisions, including board selection, despite their small share of overall equity. The S-1 describes AI as a potential driver of scientific innovation and economic growth, including cures for major diseases, while emphasizing risks that must be navigated and prevented. The central IPO question remains whether future growth will justify the obligations already signed, many of which are not yet due.