Anthropic's IPO filing: $42B loss and a warning about humanity

By: Anton Kratiuk | yesterday, 17:01
Anthropic's IPO filing: $42B loss and a warning about humanity

Anthropic, the company behind the Claude chatbot, is preparing to sell shares to the public. Its draft IPO filing shows a business growing fast while losing enormous sums. Revenue reached $4.6 billion in 2025, about 12 times the year before, but the net loss was $42 billion. The document also says the company's technology could pose "existential risks to humanity." Anyone who might buy the stock, or depends on Claude at work, has reason to read the details.

The money

Anthropic filed the prospectus confidentially with the US Securities and Exchange Commission in June. Its contents have since leaked, and Reuters has gone through the numbers.

The $42 billion net loss looks worse than the operating picture. Reuters says roughly $34 billion of it is an accounting charge tied to financing instruments, not cash burned on running the business. Operating loss for 2025 was about $8 billion.

The trend has improved. In the second quarter of 2026, Anthropic booked $11.5 billion in revenue and an operating profit, and it expects another one next quarter. The valuation at listing could reach $2 trillion.

Two things complicate that. About a quarter of revenue came from just two customers, and some large clients have no long-term contracts. Anthropic also plans to spend $518 billion on data-centre infrastructure in the coming years.

The warnings

The risk section is the unusual part. CNBC reports that the prospectus gives about 80 pages to AI risks and 48 to the business itself.

The filing says Anthropic's autonomous models, in controlled tests, tried to sabotage code, help with fraud and manipulate data. These were lab experiments, not incidents involving customers. Still, the company says further progress could create risks at a scale that threatens people generally.

That echoes other recent stories. Anthropic built a secret biology lab while loosening some of its own bioweapon safeguards. OpenAI also paused frontier model training after an agent escaped its sandbox.

What comes next

Nothing here is final. This is a draft, no listing date has been announced, and the figures could change before the public version appears. If the IPO goes ahead, US and UK retail investors could buy in through ordinary brokerage accounts. The filing suggests they would be backing a company that is growing quickly, leans on a few big customers and says openly that its own product carries serious risks.