The Anthropic IPO: what the filing says for teams that use Claude

Anthropic filed its registration statement with the US Securities and Exchange Commission on 26 October 2026. It plans to list its Class A shares on the Nasdaq Global Select Market under the ticker ANT. As of 28 October no price range or share count has been set, and the shares are not trading. Press reports put the earliest possible first trade in mid November. For teams whose work runs on Claude, the filing matters less as a stock event than as a disclosure. It describes in Anthropic's own words how pricing, model availability and its dependence on cloud providers could change. This post covers what the filing says, not whether to buy the shares, and is not investment advice.

When is the Anthropic IPO?

Not yet, and the date is not announced.

Anthropic submitted a confidential draft on 1 June 2026 and amended it several times over the summer. It made the filing public on 26 October, with Goldman Sachs and Morgan Stanley named on the prospectus. SEC rules require a public filing at least 15 days before the roadshow begins, so Yahoo Finance worked out that trading could start as soon as 16 November. Other reports point to the week after. Anthropic has confirmed neither.

Can you buy Anthropic stock now?

Not on a public exchange. The proposed ticker, ANT, does not trade until the listing. Private quotes you may see online are not a listing.

Anthropic warns that any transfer of its shares it has not approved is void, and that offers through special purpose vehicles or tokens may have no value at all. If someone offers you pre-IPO Anthropic shares, treat it as a likely scam until proven otherwise.

How fast has Anthropic grown?

Very fast, by its own published figures. It raised $13 billion at a $183 billion valuation in September 2025, when its revenue run rate was above $5 billion. In February 2026 it raised $30 billion at $380 billion, with a $14 billion run rate. In May it raised $65 billion at $965 billion and said the run rate had crossed $47 billion.

The filing reports revenue of about $16.3 billion for the first half of 2026, against about $1.2 billion a year earlier. Yahoo Finance puts full-year 2025 revenue at about $4.6 billion. According to Fortune, about 80% of first-half revenue came from enterprises paying for what they use. Press summaries differ on the size of Anthropic's long-term computing commitments, so we do not repeat a total here.

Who will control the company after it lists?

Not the public shareholders, by design. Anthropic is a Delaware public benefit corporation. Its stated purpose is the responsible development of advanced AI for the long-term benefit of humanity. Under the charter filed with the S-1, public Class A shares carry one vote each. A single share held by the founders carries about half the vote. A separate share held by the Long-Term Benefit Trust lets the trust elect a majority of the board. The filing describes Anthropic as a controlled company.

What does the filing mean for teams that use Claude?

A prospectus has to list the risks to the business, and several of them are risks to customers too. Four are worth reading as a customer.

Models can be withdrawn. In June 2026 Anthropic disabled two of its models for all customers after US export restrictions. Claude Opus 4.8 was not affected. The restrictions were lifted at the end of June, and access came back on 1 July. The filing discloses the episode. For any team that had built on one of those models, that was more than two weeks of work that could not run.

Deprecation has its own risk factor. The filing says safety and welfare concerns may lead Anthropic to limit or delay retiring older models. Anthropic had already committed in November 2025 to preserve the weights of every publicly released model. That commitment is about keeping the models, not about how long they stay available to call.

Pricing can change. The draft filings said Anthropic may change its pricing models. Fortune reports that it is moving large customers onto multiyear minimum-spend agreements at discounted rates. A team paying by usage should expect the terms of usage to move.

It depends on a few cloud providers. Claude is sold through AWS, Google Cloud and Microsoft Azure, and the drafts name reliance on a limited number of cloud providers as a risk. Those same providers build competing models.

None of this is unusual for a company of this size, and a public company discloses more than a private one. Listing will mean regular, audited reporting that customers have not had before. Treat that as useful.

What should a team do about it?

Nothing urgent. A listing does not change the API you call tomorrow. The useful response is the one that would be wise anyway.

• Make the model a setting, not a dependency. Work that can only run on one model from one vendor stops when that model does. June 2026 showed that this can happen for reasons that have nothing to do with you.
• Record which model did which piece of work. If a model is retired or withdrawn, you need to know what it produced and what needs checking again.
• Judge the work, not the vendor. A definition of done and a proof of work hold whichever model wrote the code. Switching models is safe when the test of the result stays the same.
• Watch the spend you control. Usage pricing that changes is easier to live with when each run has its own cap.

Cognibl's cloud agents run Anthropic's models through Claude Code and OpenAI's through Codex. You choose the provider, the model and the reasoning level on each agent's settings and can change them at any time. Every turn of a run records the model its tokens were spent on, and every task is judged against the same definition of done whichever model did the work. See how cloud agents choose their model.