Anthropic, the developer of the Claude models, made its Form S-1 registration public on October 26, 2026, lodging it with the SEC and formally starting the run-up to an initial public offering. It wants its Class A shares to trade on Nasdaq's top Global Select tier under the ticker "ANT". It has not yet set how many shares it will sell or at what price, and the registration statement is not yet effective.
The prospectus gives the first audited look at how quickly the business has grown. Revenue for the half-year to June 30 came to $16.3 billion, against $1.2 billion over the same months of 2025. Its operating loss narrowed to $693 million from about $2.1 billion. The bottom line looks very different: the net loss for the half reached $80.7 billion, which TechCrunch and The Verge both attribute mostly to non-cash accounting charges linked to convertible notes the company issued to raise money. Annual net losses, the filing shows, climbed from $1.5 billion in 2023 to $8.3 billion in 2024 and $42.0 billion in 2025, and warns that profitability may never be reached or sustained.
Enterprise demand is the core of the revenue story. On a trailing twelve-month basis, Anthropic counts more than 11,900 customers generating at least $100,000 each, more than 1,900 at $1 million or above, and roughly 250 above $10 million.
The governance structure is the most unusual part of the offering. A special Class F share, held by a limited liability company whose initial members are the seven co-founders, carries 50.1% of total voting power but no economic rights. A separate class lets the Long-Term Benefit Trust, the financially disinterested body Anthropic created in 2023, elect and remove the directors who will make up a majority of the board. Anthropic presents the design as protection for its public benefit mission; TechCrunch calls the arrangement controversial because outside shareholders will have limited influence over a company that itself describes frontier AI as capable of catastrophic or existential harm.
The filing also quantifies the cost of compute. Anthropic disclosed agreements signed between May and October 2026 with SpaceXAI, a SpaceX subsidiary, for access to NVIDIA GPU capacity worth up to $120.5 billion through 2029, or about $3.1 billion a month once all capacity is delivered, which it expects in December. The Verge, reading the same document, reports further commitments running to 2033 of $44.9 billion with Microsoft and no less than $114.5 billion with Google, plus $110 billion with Amazon over ten years.
Among the risk factors are pending copyright disputes over training data, the possibility of taxes aimed specifically at AI, regulatory consequences if models are seen to have some form of sentience or moral status, and the warning that no single insurer, and perhaps not the insurance market as a whole, could cover a catastrophic AI incident. Goldman Sachs and Morgan Stanley are named as the contacts for the preliminary prospectus.
Several things remain open. The offering size, price range and timing are not yet known, and the SEC review could still change the document. The $80.7 billion loss is dominated by accounting items, so the operating figures are a better guide to the underlying economics, but they still show a company that spends heavily on compute and has yet to turn an operating profit.