Anthropic Public Filing has given the public its first full look at the finances of the company behind Claude, and the picture is striking: revenue surged in the first half of this year while a huge paper loss dominated the bottom line. According to the filing, the artificial intelligence developer reported revenue of about 16.3 billion dollars for the six months ended in June, up sharply from the same period a year earlier, alongside a net loss of roughly 80.7 billion dollars. The contrast between those two lines explains why the document has drawn so much attention from investors, rivals and customers in the days since it landed.

The company filed its registration statement with the United States securities regulator on October 26, seeking a listing for its Class A shares on the Nasdaq exchange under the ticker ANT, according to reporting on the filing. According to reporting on the filing, the share count and price range were left blank, which means no price has been set and no shares can be sold until the regulator declares the registration effective. That blank space is normal for a first public filing, and a later amendment is expected to fill it in before any roadshow in the coming weeks.

The numbers behind the headline

Strip out the accounting noise and the operating story looks very different from the headline loss. The filing shows an operating loss of about 693 million dollars for the first half of the year, a much narrower shortfall than the same period a year earlier, when the operating loss was far larger. On the company's own adjusted measure, which leaves out stock-based pay and other items, the business swung to an operating profit of roughly one and a half billion dollars. Supporters point to that improvement as evidence that the core business is approaching sustainability, while sceptics note that adjusted figures always flatter the company that chooses them, so they should be read beside the audited results rather than instead of them.

Most of the headline loss came from a single non-cash item. According to analysis of the filing, a charge of about 72.4 billion dollars reflected the rising fair value of convertible notes and similar financing instruments as the company's own valuation climbed. No cash left the company for that charge, and the filing describes it as driven primarily by the increase in the fair value of outstanding convertible promissory notes. In other words, the paper loss grew because investors now value the company far more highly than they did when those notes were issued, which is an unusual but logical consequence of a fast-rising private valuation meeting public accounting rules.

Cash is not the constraint here. The filing reports holdings of cash and short-term investments totalling about 131.7 billion dollars at the end of June, a war chest built from successive funding rounds. That stockpile matters because the company's largest cost is computing power: the bill for compute and infrastructure in the first half of the year ran to nearly nine billion dollars, according to the filing's statements of operations. Training and running large models is expensive, and the filing leaves the total for future compute commitments blank, which analysts say is one of the key gaps they want filled in the next amendment.

What buyers and builders should watch

Revenue concentration is the other risk that deserves attention. According to reporting on the filing, more than eleven thousand customers paid the company at least one hundred thousand dollars in the year to September, while a much smaller group of roughly two hundred and fifty customers each paid more than ten million dollars. The filing itself warns that a substantial part of revenue depends on a limited number of customers and distribution channels, and reporting elsewhere suggests the five largest customers supplied about a quarter of revenue in the first nine months of the year. For businesses that build on Claude, that concentration cuts both ways: a few large relationships drive growth, but losing one would hurt.

Governance will also be part of the debate. According to reporting on the filing, a special class of shares would leave the co-founders with just over half of the voting power after listing, even though public buyers would hold the economic majority. Supporters of the structure argue it protects a long-term safety mission from short-term market pressure, while critics counter that it leaves outside shareholders with little say over strategy. The filing also flags the familiar risks of the sector: usage limits, rate limits and model capabilities have changed before and may change again, which matters to any customer planning around a particular model.

Why it matters beyond the markets is straightforward. Anthropic is one of a small handful of companies whose models now sit inside workplaces, classrooms and software products used by millions of people, so its pricing, capacity and safety choices ripple outward. A public listing would force regular disclosure of revenue, costs and risks, giving customers and competitors a clearer view of a business that has so far been valued largely on private rounds and leaked figures. Readers who follow the wider AI build-out can compare this filing with recent infrastructure moves covered here, including Oracle's AI chip expansion and CoreWeave's Denmark expansion, and with a new agent-focused news feed from Axios that shows how AI tools are reshaping even the business of information itself.

What happens next is a sequence, not a single event. The Anthropic Public Filing must be followed by an amendment setting a share count and price range, the regulator must declare the registration effective, and only then can a roadshow test what public investors will pay, according to further reporting on the listing plan. If demand holds, the listing could rank among the largest technology offerings ever attempted; if it falters, the filing will still stand as the most detailed public record yet of what the AI boom costs, earns and risks. Either way, the numbers now on the record will anchor every argument about the company's worth from here on.