Anthropic has filed to go public, and its first detailed set of numbers tells two stories at once: revenue of $16.3bn in the first half of 2026, up from $1.2bn a year earlier, and a net loss of $80.7bn over the same six months that is almost entirely a paper charge rather than cash the company spent.
Anthropic, the San Francisco AI company behind the Claude models, founded in 2021 by former OpenAI staff led by Dario and Daniela Amodei, has filed a draft S-1 registration statement with the US Securities and Exchange Commission. It plans to list on the Nasdaq Global Select Market under the ticker "ANT", with Goldman Sachs and Morgan Stanley leading the offering.
That revenue is about 14 times what it was a year earlier, and the growth is still steepening: for the three months to June the filing puts the year-on-year increase at 1,368%.
The offering price, the number of shares and the valuation are all still blank in this draft. Reporting on the filing points to a listing in the coming weeks.
How much did Anthropic lose in the first half of 2026?
The filing reports a net loss of $80.7bn for the six months to June. Almost none of that is money the company spent.
$72.4bn of the loss is a non-cash charge for the change in fair value of convertible notes, derivatives and SAFEs, the instruments start-ups use to raise money before an IPO. As a company's value rises, the accounting cost of converting those instruments into shares rises with it. It lands on the income statement as a loss even though no cash leaves the business. The Wall Street Journal first reported the $80bn figure from the draft.
Take that charge out and the picture changes. Anthropic's operating loss was $693m, on total running costs of about $17bn that only just ran ahead of revenue. On the company's own adjusted measures it was in profit, with adjusted operating income of $1.6bn and adjusted net income of $2.7bn for the half.
That is a sharp turn from 2025, when the full-year adjusted operating loss was $4.9bn.
Revenue is now bigger than the compute bill
The line many will study is compute. Anthropic spent $8.9bn on compute and infrastructure in the half, the biggest single cost of running a frontier AI model. Revenue of $16.3bn came in well above it, which is rare for a model-builder at this stage and sets it apart from the AI startups that raised millions before collapsing on costs they could not cover.
The customer base is weighted to the top. As of 30 September 2026 more than 11,900 customers spent $100,000 or more with Anthropic over the trailing year, and about 250 spent $10m or more. Net revenue retention was around 900%, meaning existing customers alone spent roughly nine times more than a year earlier. Coverage of the S-1 notes that about a quarter of revenue comes from its two largest customers, a concentration the filing lists among its risks.
What the filing says about cash and control
The balance sheet is unusual for a company still posting losses. Anthropic held more than $130bn in cash and short-term investments at the end of June, having raised $141.1bn in total to date.
Control stays with the founders and a trust. Anthropic is a Delaware public benefit corporation, and its stated purpose is "the responsible development and maintenance of advanced AI for the long-term benefit of humanity". A Long-Term Benefit Trust elects most of the board, and the seven co-founders hold a class of shares carrying 50.1% of the voting power, so the company will list as a Nasdaq "controlled company". Amazon's convertible notes turn into a separate share class when the offering closes.
Two things are worth taking from it.
First, read the loss correctly. A headline net loss of $80.7bn sounds like a company burning cash at an impossible rate. The real operating loss was $693m, and the rest is an accounting effect of a rising valuation.
Founders raising on SAFEs or convertibles will see the same mechanic, smaller, in their own accounts, and investors reading UK filings should know the difference before they react to a number.
Second, it resets the benchmark. An AI company can now show software-scale revenue against a compute bill that, for once, is the smaller number.
That is the case UK AI founders will be measured against when they raise, and the reason public money keeps moving toward the sector, from the British Business Bank's £180m into AI and deeptech funds to the Sovereign AI Fund.
It also concentrates attention and capital at the top, in a market where London already takes 78% of what UK startups raise.
Anthropic has shown the growth. The IPO will test what investors will pay for it.
