Hey there!

On Monday, 26 October, Anthropic, the company behind Claude, publicly filed its S-1 with the US Securities and Exchange Commission. An S-1 is the long document a company has to publish before an IPO (initial public offering), the point where its shares start trading on a public stock exchange. Anthropic plans to list on the Nasdaq under the ticker ANT, and the filing shot up Hacker News within an hour of going live.

It's the first time Anthropic has had to publish its full financial statements, so for once we get real numbers on how a frontier AI lab makes and spends money. In this post I'll go through what the Anthropic S-1 says about revenue, the huge loss figure in the headlines, and who will actually control the company once it's public.

One quick note: this is an explainer, not financial advice, so none of it is a recommendation on whether to buy the stock.

What an S-1 filing is, and what's still missing

The S-1 is a registration statement. Until the SEC declares it effective, Anthropic can't sell a single share. In its announcement of the S-1 filing, Anthropic says the number of shares and the price range haven't been decided yet, and it names Goldman Sachs and Morgan Stanley as the banks handling the offering. So the filing tells us a lot about the business and nothing yet about what a share will cost.

On timing, Fox Business reported, citing unnamed sources, that the Anthropic IPO could happen as soon as 19 or 20 November. Anthropic hasn't confirmed a date, so treat that as a report rather than a schedule.

Anthropic's revenue: from $386 million to $16 billion

In 2024 Anthropic brought in $386 million. In 2025 that grew to about $4.6 billion, roughly 12 times more. In the first six months of 2026 alone it made about $16.3 billion, which is already more than three and a half times its revenue for the whole of 2025.

The filing also shows how many large customers are behind that. As of 30 September 2026, more than 11,900 customers had paid Anthropic at least $100,000 over the previous twelve months, more than 1,900 had paid at least $1 million, and about 250 had paid at least $10 million. Anthropic also reports revenue per employee of over $6.6 million for that twelve-month period, up from $1.4 million a year earlier.

How that money arrives matters too. The filing says revenue is mostly consumption-based, meaning customers pay for what they use, and much of it isn't tied to long-term contracts. It also warns that a substantial portion comes from a limited number of customers and distribution channels; Yahoo Finance's reading of the filing puts roughly a fifth of revenue at just two customers. Usage-based revenue can climb quickly, and it can drop quickly if a big customer moves its spending elsewhere.

That $80.7 billion loss, explained

The number getting the most attention is the net loss: $80.7 billion for the first half of 2026, after a $42 billion loss in 2025. Next to $16 billion of revenue, that looks like a company burning money at an impossible rate.

Most of it isn't the cost of running the business. The operating loss, which covers things like compute, salaries and sales, was about $693 million in those six months, or around 4% of revenue. A year earlier the picture was much worse: an $8.1 billion operating loss in 2025 on $4.6 billion of revenue.

The bulk of the net loss, $72.4 billion, is an accounting charge linked to convertible notes, SAFEs and similar instruments. These are deals where investors put money in early in exchange for the right to receive shares later. As Anthropic's value rises, those rights become worth more, and accounting rules make the company record that increase as a loss. The filing says the change was mainly driven by a rise in the fair value of its convertible notes. Put simply, a large part of the scary headline number is a side effect of the company becoming more valuable.

The spending is still enormous. Yahoo Finance reports that Anthropic spent $9 billion on AI infrastructure in the first half of 2026, more than the $7.3 billion it spent in all of 2025, and that it plans to spend more than $450 billion on cloud computing in the coming years through deals with partners including Amazon and Google. Whether revenue keeps growing faster than those compute bills is the question the numbers can't answer yet.

Who controls Anthropic after the IPO

This is what makes the Anthropic IPO unusual. Anthropic is a public benefit corporation (PBC), a type of company that's legally allowed to weigh a stated mission alongside profit. Its stated public benefit is the responsible development and maintenance of advanced AI for the long-term benefit of humanity. The S-1 backs that up with five classes of shares:

• Class A: the shares the public can buy, with one vote each. This is the only class that will be listed.
• Class N: held mainly by strategic partners, with no votes except where the law requires them.
• Class T: held by the Long-Term Benefit Trust, an independent group of five trustees with no financial stake in the company. It elects the directors who make up a majority of the board, and has no economic rights.
• Class F: a single share held by a company whose initial members are Anthropic's seven co-founders. It carries 50.1% of the total voting power on matters that go to a shareholder vote, and also has no economic rights.
• Class E: planned for employees after the offering, with limited voting.

In practice, anyone who buys ANT shares gets a slice of the economics and very little say. The trust picks most of the board, and the founders' single share outweighs every other shareholder in a vote. Because the trust controls the board majority, Anthropic will count as a "controlled company" under Nasdaq rules, although it says it only expects to use the related exemption for its nominating and corporate governance committee. The filing's risk factors say this concentration of voting power will last for the foreseeable future.

Founder-controlled share classes aren't new; Alphabet and Meta both have them. Pairing founder control with an independent trust that elects the board is the unusual part.

The risks Anthropic lists

Every S-1 has a risk factors section, and Anthropic's includes some that only an AI company would write. It warns that its models may produce inaccurate, misleading or harmful output. It also points out that some of the cloud providers it buys large amounts of compute from build competing AI models of their own, and that their interests may not always line up with Anthropic's.

Yahoo Finance's coverage of the filing also points to two run-ins with the US government this year: the Defense Department's designation of Anthropic as a supply chain risk after it refused certain military uses, which Anthropic is challenging in court, and an export control directive that forced it to suspend access to some models for foreign nationals.

The takeaway

The headline loss and the underlying business tell very different stories. Revenue grew roughly 12 times in 2025 and has kept climbing in 2026, the operating loss has shrunk to about 4% of revenue, and most of the $80.7 billion net loss is an accounting effect of rising valuations. The open questions are compute costs and reliance on a few big customers, both flagged in the filing itself, plus the government disputes covered in the press.

Anyone who buys in is also buying into a company where the public holds the economics while the founders and a safety-focused trust hold control. Before trading starts, expect amended filings that add the share count and a price range. If you want to dig into the details yourself, the full Anthropic S-1 is on the SEC website.