Anthropic has moved from private-market rumour to SEC machinery. On Monday, the company said it had confidentially submitted a draft Form S-1 for a proposed IPO of common stock. The offering has no share count yet, no price, and no public prospectus. It is still conditional on SEC review, market conditions, and the usual list of factors that mean, in plainer English, this can be paused if the weather turns.

However, the signal is not cautious. Anthropic has put its hand on the door first. CNBC reports that OpenAI is preparing its own confidential filing, while Axios frames the year as a race between SpaceX, Anthropic, and OpenAI for public-market attention at trillion-dollar scale. That sounds absurd until you look at the numbers Anthropic is already asking investors to swallow.

Last week the company announced a $65 billion Series H at a $965 billion post-money valuation, led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital. Its own Series H note says run-rate revenue crossed $47 billion earlier in May. In February, according to the Guardian's summary of the funding history, Anthropic was valued at $380 billion. The market has not so much repriced the company as yanked it through a trapdoor into another category.

I wrote in April about Anthropic moving past OpenAI in the private valuation story. At the time, the IPO talk sounded aggressive because the company was still being priced in whispers, board meetings, and preemptive term sheets. Now the sequence is visible: raise at almost a trillion, file confidentially, let the public S-1 follow if the regulator's comments and the market both behave. The filing does not guarantee a float, but it changes the tempo.

The public S-1, if Anthropic proceeds, is where the romance gets audited. Axios makes the useful procedural point: a public filing would include financial data, risk factors, and the details that private-market storytelling can blur. Revenue run rate is one thing. Gross margin, compute obligations, customer concentration, stock-based compensation, related-party cloud deals, and the real cost of serving Claude Code are another. A frontier lab can be a religion in private. In public markets it becomes a spreadsheet with a risk section.

This is why the timing matters more than the ceremonial language around going public. Anthropic has spent the spring making itself legible to capital. The company raised at a near-trillion valuation, talked about compute expansion and product partnerships, and kept pushing the enterprise story that makes Claude look less like a chatbot and more like operating infrastructure. Its Wall Street embedding strategy already pointed this way. The model was not merely a product; it was a reason for banks, private equity firms, and portfolio companies to reorganise how work gets done.

Public investors will be less patient with mythology than late-stage private money, at least in theory. They will want to know whether $47 billion of run-rate revenue can hold its shape when discounts, compute costs, and implementation work are all counted honestly. They will also want to know how much of the AI boom is now a queue for the same finite capital. NPR quoted Wedbush calling the moment an opening of the IPO floodgates. Floodgates are a good image, but queues may be better. SpaceX, Anthropic, and OpenAI cannot all be the single impossible listing that defines the year.

The funny part is that confidential filing is meant to reduce drama. It gives the company a private conversation with the SEC before the theatre begins. In this case the privacy is almost beside the point. Anthropic has told everyone where it wants to go, and now the rest of the market has to decide whether a frontier-model lab is a software company, an infrastructure company, a consultancy, a defence-adjacent contractor, or some expensive composite of all four.

Maybe the S-1 will make that cleaner. More likely it will make the argument more interesting.

Sources: