Anthropic's IPO Prospectus Lands After Labor Day, Carrying A Bill For Its Own Red Lines

In this storyAnthropic

Anthropic will publicly unveil its IPO prospectus after Labor Day, with a listing targeted for late September or early October, according to The Information. The company submitted a confidential draft S-1 to the SEC on 1 June and announced the step itself, saying the share count and price had not been set.

The scale would be unprecedented. Anthropic closed a $65 billion Series H in late May at a $965 billion post-money valuation, passing OpenAI, which had been valued at $852 billion in March. Backers included Blackstone, Brookfield, D1 Capital Partners, GIC, General Catalyst and Insight Partners. The New York Times has put an eventual public valuation as high as $2 trillion. Anthropic expects the offering to match or exceed SpaceX's $86 billion June listing, the largest on record.

Revenue has moved faster than the valuation. The company ended 2025 at roughly $10 billion in annual revenue. Run-rate crossed $19 billion by early March, $30 billion in April and $47 billion in May, and topped $65 billion by August. Claude Code, its coding agent, passed $2.5 billion in run-rate on its own in February.

The structure of the offering is unusual in two respects. Anthropic is weighing whether to let existing shareholders sell into the IPO, which neither SpaceX nor Cerebras did this year, and considering lockups longer than the standard 180 days. It has also examined requiring rank-and-file employees to sell through preset 10b5-1 plans rather than the usual post-earnings windows.

Citigroup joined the underwriting group in August. Bankers have been running test-the-water meetings in San Francisco since the summer.

The float design is the tell. SpaceX's first lockup expiry this month more than doubled its publicly traded share count, with roughly 12.9 billion more shares scheduled to unlock by mid-2027. Every large listing this year has watched what a supply cliff does to a price. Longer lockups, staggered employee selling and secondary sales inside the offering itself all point the same way: release the shares gradually rather than in one drop.

That is prudent. It also concedes something. A company entirely confident of holding its multiple does not engineer this carefully around the moment insiders are first allowed to sell.

The growth figures cut both ways too. A run-rate that travels from $19 billion to $65 billion inside six months is extraordinary, and it is also close to unpriceable. There is no settled base against which to build a comparable. Public investors will be underwriting a trajectory rather than a business with a known shape, and the first quarter that falls off the curve will be read as a break rather than a fluctuation.

Then there is the cost side, which the prospectus will have to show in full. Anthropic has raised at least $130 billion to fund compute. SpaceX's own prospectus discloses that Anthropic pays it $1.25 billion a month through May 2029, under an agreement either side can end on 90 days' notice. A revolving credit facility above a $10 billion target was being finalised ahead of the public filing. This is not a company listing to bank a win. It is listing because the next round of capital has to come from somewhere larger than private markets can reach.

The genuinely novel disclosure problem sits elsewhere. Anthropic is a public benefit corporation, and it is about to ask public shareholders to fund a company that has demonstrated, on the record and in court, that it will give up revenue over stated limits.

In March the Pentagon designated Anthropic a supply chain risk, the first time an American company had been publicly labelled that way. The dispute was not about security. Contract talks collapsed because Anthropic refused two categories of use: mass surveillance of Americans, and fully autonomous weapons without human control over targeting and firing. Defence Secretary Pete Hegseth's position was that the Pentagon should have unrestricted access for any lawful purpose. Chief financial officer Krishna Rao told the court the government's actions could cut 2026 revenue by multiple billions of dollars.

On 27 August, Judge Rita Lin ruled the designation unlawful, finding it retaliation against protected speech under the First Amendment and a denial of due process under the Fifth, and calling it arbitrary and capricious. She ordered it removed. A second case over a related designation remains pending in Washington.

Anthropic won. It also lost the better part of six months of federal and defence-adjacent business, while OpenAI, Google and xAI held Pentagon contracts of comparable size throughout, having declined to draw the same line in public.

That is a sentence no prospectus has had to contain. Every risk-factor section warns about regulation. This one has to explain that the company's own commitments are a recurring, quantifiable cost, and that management intends to keep incurring it. CNBC reports the filing will also flag public hostility to AI and to data centres as a risk factor.

For Indian founders the number worth holding onto is the $130 billion raised before a single public share trades. That is the entry price at the frontier, and it largely settles the argument running here about whether domestic labs should chase foundation models or build on top of them. The second question is harder and more interesting: whether a company can carry a stated ethical constraint onto a public register and still be rewarded for it. Anthropic is about to run that experiment in front of everyone.

For daily, sharp analysis of the biggest moves in the Indian business and startup ecosystem, follow StartupFox.