Anthropic files to go public and bets safety can survive Wall Street

The company that builds Claude has told regulators it wants to sell shares to the public. Anthropic confirmed it submitted a confidential draft registration statement to the U.S. Securities and Exchange Commission, the first formal step toward an initial public offering. For a lab that has spent its entire existence arguing it would build artificial intelligence more carefully than anyone else, the filing is the moment that argument meets a stock ticker.

For anyone who relies on Claude, competes with it, or simply worries about how fast this technology is moving, the move carries one blunt implication. The most safety-conscious of the major AI labs is preparing to answer to public shareholders, and the promises it has made about caution will soon share a boardroom with investors who can vote on strategy.

The scale behind the decision shows in the numbers. Anthropic raised $65 billion in its most recent funding round, at a valuation near $965 billion, enough to edge past OpenAI as the most valuable company in the field for the first time. Its revenue run-rate has passed $47 billion on an annualized basis, up from roughly $9 billion at the close of 2025. The company expects to book $10.9 billion in revenue for the second quarter alone, more than double the $4.8 billion it reported for the first.

Those figures reflect a specific kind of demand. Much of the growth has come from businesses and developers building on Claude through its programming tools, where the model’s skill at writing and reviewing code has made it a default for many engineering teams. That is a sturdier base than consumer subscriptions alone, and it helps explain why investors have been willing to attach a near-trillion-dollar price to a company that did not exist six years ago.

The filing is confidential, which matters more than it sounds. A confidential draft lets a company begin the regulatory process without publishing the financial statements, risk factors, and internal details that a normal prospectus exposes. Anthropic gets to prepare for Wall Street largely out of view. The people who use its product, and the public it claims to serve, will see little of substance until a later, open version of the document appears. The number of shares and the price remain unset.

Timing is the other half of the story. By filing first, Anthropic positions itself to reach the public market as early as this autumn, ahead of OpenAI, which is said to be preparing its own confidential paperwork in the coming weeks. OpenAI raised $122 billion earlier in the year at an $852 billion valuation, and SpaceX is reportedly moving toward a listing on a far larger scale. The result is a cluster of the most heavily funded private companies in history reaching for public investors at almost the same moment.

Anthropic’s earlier funding came in part from Amazon and Google, both of which committed billions and tied their own cloud and AI plans to its models. A public listing gives those strategic backers, along with the venture funds that led the recent round, a path to eventually sell shares on the open market. It also hands Anthropic something it has lacked, publicly traded stock it can use to raise money and to pay staff in a market where skilled researchers are the scarcest resource.

The company is not a conventional startup in one important respect. It is a public benefit corporation, a structure that legally binds it to a mission alongside profit, in its case the responsible development and maintenance of advanced AI for the long-term benefit of humanity. Founded in 2021 by the siblings Dario and Daniela Amodei, both former OpenAI employees, Anthropic built its identity on the idea that it would slow down where rivals sped up.

That identity is exactly what going public complicates. A listed company answers to shareholders every quarter, and shareholders reward growth, market share, and speed far more reliably than they reward restraint. The benefit-corporation structure gives the mission legal standing, but it does not make it immune to the gravity of a share price. The blunt version of the worry is that Anthropic is preparing to sell a safety mission that Wall Street can vote against. Whether the governance it has built can hold that line under earnings pressure is the question the filing does not answer.

For users, the pressure could cut both ways. Public money can fund faster development, wider access, and cheaper models. It can also push a company toward shipping quickly and pricing aggressively, the opposite of the deliberate pace Anthropic has used to define itself. None of it changes how Claude works today, the product stays available worldwide, and a U.S. listing does not alter what anyone can access. What changes is the incentive behind the lab that makes it.

The timeline is far from settled. A confidential filing is a beginning rather than a commitment, and people close to the process describe the schedule as touch-and-go, a debut that could arrive this summer, this autumn, or not at all. The share count, the price, and the full financial picture will only become public when Anthropic files the open version of its prospectus. Until then, the most closely watched experiment in the industry is whether a company built to be careful can stay that way once the market is watching.

Tags: Anthropic, Claude, IPO, OpenAI, Dario Amodei, Generative AI