Claude's maker chases a record float despite billions in losses

By Zara Kincaid · 2026-10-06

The filing

Anthropic PBC confidentially submitted a draft registration statement on Form S-1 to the U.S. Securities and Exchange Commission on 1 June 2026, flagging its intention to pursue what analysts say could become the largest initial public offering in history.^[1] The company has not yet set a final price or confirmed a listing date, and the document remains subject to SEC review.

Anthropic recorded revenue of nearly US$4.6 billion in 2025, a twelvefold increase from 2024.^[2] A twelvefold revenue jump in a single year is what summary pages are built to show off. What sits two lines below it is harder to absorb.

The numbers behind the numbers

The prospectus discloses an operating loss of more than US$8 billion in 2025, excluding writedowns tied to previous fundraising rounds.^[2] Compute, talent and infrastructure costs at the frontier of large language model development produce that kind of gap between revenue and profitability, and the absolute size of the loss is a number investors will need to model carefully against any US$2 trillion valuation target.

Analysts have drawn comparisons to SpaceX's US$1.77 trillion market debut, noting that Anthropic's float could set fresh valuation benchmarks for companies offering pure-play AI exposure.^[2] Whether the market will absorb a loss-making company at that multiple is the operative question.

One third of the prospectus is a warning

The draft dedicates roughly 80 of its 261 pages to risk factors, including warnings that Anthropic's own AI models could pose catastrophic or existential risks to humanity.^[3] Most S-1 risk sections cover market competition, regulatory exposure and key-person dependencies; spending a third of the document on existential danger has no real precedent in technology listings.

Anthropic was established in 2021 by former OpenAI researchers specifically to build AI more safely, using what it calls a constitutional AI approach to improve model behaviour. Flagging civilisation-scale risk in the prospectus is consistent with that founding posture, though a pension fund allocator will read those pages differently to an AI safety researcher.

The Pentagon problem

The more immediate risk sits inside a U.S. Court of Appeals filing dated 8 April 2026. On 3 March 2026, then-Secretary of War Pete Hegseth designated Anthropic a supply-chain risk under 41 U.S.C. § 4713.^[4] The Department of Defense subsequently cancelled its contracts with Anthropic, began removing Claude from its systems, and barred contractors from using Anthropic products for any work connected to the department.

The designation appears to stem from Anthropic's refusal to remove usage constraints from its models for defence applications. CEO Dario Amodei said the vast majority of Anthropic's customers would be unaffected, since the designation plainly applies only to the use of Claude by customers as a direct part of contracts with the Department of War, not all use of Claude by customers who have such contracts.^[4] Amodei described the underlying characterisation of Anthropic's conduct as "completely false and just straight up lies."^[4]

Regardless of how the legal dispute resolves, the designation is now a material item in the prospectus and a variable investors must price before the float. The SEC review process and final pricing timeline had not been publicly confirmed as of the date of this article.

Zara Kincaid

Zara Kincaid writes about artificial intelligence and search. Her focus is what happens to businesses when the front page of the internet stops being a list of links and starts being an answer.