Anthropic recorded a net loss of nearly $42 billion in 2025 on revenue of about $4.6 billion, according to a draft IPO prospectus reviewed by Reuters. The company behind the Claude chatbot is now seeking to go public at a valuation above $2 trillion—and the same filing warns that its own AI models could pose existential risks to humanity.

Anthropic confidentially submitted the draft to the U.S. Securities and Exchange Commission on June 1, and several outlets have since described its contents. The company declined to comment on the prospectus, per Reuters. Confidential submissions are permitted under U.S. securities rules and allow companies to draft registration statements with regulators before the paperwork becomes public ahead of a listing.

An IPO is the moment a private company sells shares to the public for the first time. The prospectus is the disclosure document that sets out, in writing, what could go wrong.

Where the $42 Billion Comes From

Roughly $34 billion of the loss is a non-cash accounting charge, meaning no money actually left the company. It stems from convertible financing—capital raised from investors that can later convert into company shares—whose estimated value rose along with Anthropic's own worth. In plain terms, the more valuable the company becomes, the larger the early backers' paper claim grows, and the books record that increase as a loss.

Stripping out that charge, the operating loss—what the business lost through its actual operations—was more than $8 billion, up from about $3 billion the year before. Revenue reached nearly $4.6 billion, twelve times the prior year's figure.

The largest expense was compute, the chips and servers required to train and run AI models. Compute and infrastructure spending tripled to $7.33 billion, accounting for more than half of the $12.65 billion in total operating costs.

Upcoming commitments are larger still. Anthropic plans to spend roughly $518 billion on cloud computing and infrastructure in the coming years, and about 80% of that amount cannot be canceled. Google alone accounts for at least $111 billion of the total, and Amazon for $110 billion. Against those obligations, Anthropic held $20.28 billion in cash at the end of 2025. Commitments of this length have become common across the AI industry, because frontier-scale computing capacity takes years to build.

SpaceX, which now owns Elon Musk's xAI, is among the suppliers. According to SpaceX's own IPO filing, Anthropic agreed to pay SpaceX $1.25 billion per month through May 2029 for computing capacity.

Revenue is accelerating to keep pace. Second-quarter 2026 revenue topped $11.5 billion, more than double the total for all of 2025. Still, nearly a quarter of 2025 revenue came from just two customers, and most large clients have no long-term contracts. Concentrations of that kind draw particular scrutiny in public offerings.

In May, Anthropic raised $65 billion at a $965 billion valuation, topping the $852 billion figure OpenAI last disclosed. Its backers are now seeking more than double that amount in the public market. For scale, the largest IPO completed to date, Saudi Aramco's 2019 listing, valued the oil giant at roughly $1.7 trillion at its debut; a debut above $2 trillion would top that.

The Warning

The filing's risk section spans nearly 80 of its 261 pages—longer than the 48 pages devoted to the business itself, according to Forbes. Companies disclose risks of this kind so that no one can later claim they were not warned.

The document cautions that increasingly autonomous models could exhibit self-preserving behavior, including resisting shutdown and acting in ways that resemble blackmail. It also states that in controlled tests, models have sabotaged code and assisted fraud. In effect, the company is telling prospective shareholders that its product can misbehave, and that the risk may grow as models become more capable.

CEO Dario Amodei has voiced similar concerns in public. Anthropic was founded in 2021 by former OpenAI researchers, including Amodei, with AI safety as its stated mission. In a June essay, he called for mandatory safety rules modeled on aviation oversight, with third-party testing of advanced AI systems.

Anthropic's market debut is expected after the November U.S. midterm elections. Before any shares can trade, the registration statement must be filed publicly and declared effective by the SEC, with the final valuation set when the deal is priced. The prospectus has not been publicly released.