Anthropic files 335-page prospectus to list on Nasdaq as “ANT”

Anthropic published the 335-page prospectus it filed with the U.S. Securities and Exchange Commission on October 26, applying to list on Nasdaq under the ticker “ANT”. The public S-1 offers the first full view of how Claude’s maker earns money, what it has promised to pay for compute, and how it frames its own risks.

Compared with the draft that leaked earlier, the public filing raises the company’s purchase commitment to Microsoft from $31.4 billion to $44.9 billion and puts its commitment to Google at $114.5 billion. A compute contract with SpaceXAI carries a maximum value of $120.5 billion, and two new suppliers, Nscale and Lambda, appear in the document for the first time.

Anthropic confidentially submitted its S-1 on June 1, 2026, and is now filing publicly in the same format. Because the SEC requires a public filing at least 15 days before a roadshow, the company could start meeting investors as early as the second week of November; Bloomberg has reported that it aims to complete the IPO in mid-November.

Pricing has not been set. The leaked draft pointed to a raise of up to roughly $100 billion at a valuation of about $2 trillion, which would exceed SpaceX’s $1.77 trillion offering valuation in June and rank as the largest IPO on record. With AI stocks under pressure, however, some investors have cut their expectations to about $1.5 trillion, leaving the outcome to the roadshow.

The financials explain both the excitement and the alarm. Revenue grew from $41 million in 2023 to $386 million in 2024 and $4.585 billion in 2025, then reached $16.286 billion in the first half of 2026 alone, against $1.16 billion a year earlier. The net loss widened from $41.965 billion for all of 2025 to $80.748 billion in the first half of 2026, but about $72.4 billion of that reflects fair-value changes on convertible notes, derivative liabilities and other financing instruments rather than cash burned.

Consumption-based revenue, including API calls, contributed about $13.06 billion, or 80.2% of first-half revenue, while subscriptions brought in roughly $3.2 billion, or 19.6%. Customers spending more than $100,000 on a trailing-twelve-month basis numbered more than 11,900 as of September 30, 2026, up from 950 a year earlier. The filing also corrects a widely repeated guess about concentration: the two customers that each account for about 12% of revenue are Amazon and Google, not Cursor and GitHub Copilot.

The document organizes the model line into four tiers: the flagship Mythos and Fable, with Mythos restricted to vetted security and life-sciences organizations through programs such as Project Glasswing while the commercial Fable reaches version 5.1; Opus, whose latest release, Claude Opus 5.5, cuts the cost of a single task by 40%; the balanced Sonnet; and the lightweight Haiku. It also sets out a deliberate not-to-do list: no image or video generation models for now, with compute concentrated on coding, agentic capabilities and safety research.

Roughly 100 of the 335 pages are devoted to risk factors, and 94 of them warn that AI could destroy humanity, resist being shut down or attempt to extort people. The seven co-founders will still hold 50.1% of total voting power after listing. At an investor day in mid-October, the company said third-quarter revenue reached $16.8 billion, up 45% quarter over quarter, and that it was profitable on an adjusted basis.

If the listing goes ahead, Anthropic would become the first pure-play frontier model developer on a U.S. exchange, giving private rivals such as OpenAI a public valuation reference. It would also put roughly $518 billion of decade-long compute commitments, about 80% of which is reportedly payable whether or not it is used, in front of public investors every quarter.