Anthropic's IPO filing shows rapid growth and $600 billion compute bill

As of 20 October 2026, Anthropic is preparing for a November IPO after confidential filing with the SEC in June. The latest development came on 19 October, when CNBC reported that Anthropic would place 5% of its equity into a new public benefit fund before shares trade. Kate Rooney said the stake could be worth at least $50 billion, and closer to $100 billion at a $2 trillion debut. For every 50% jump in valuation, the company will add another 2% of equity to the fund. That mechanism means public shareholders keep a smaller share of each gain as the stock rises.

How the filing developed

Anthropic submitted its draft registration statement to the SEC in June, and by August it had already started informal meetings with bankers and investors in San Francisco. In those sessions, CFO Krishna Rao faced questions on rival labs, pressure on margins from open-source models, and what a pullback in data center building would mean. CNBC sources said AI backlash would be a key risk factor. A Gallup survey released in May reported that 70% of Americans were against new AI data centers where they lived, and nearly half were strongly opposed. In August, Republican Byron Donalds won Florida's gubernatorial primary after proposing data center restrictions, and Pennsylvania Governor Josh Shapiro signed an executive order imposing harsh standards.

By late September, Reuters had obtained the prospectus. It showed Anthropic lost $42 billion in 2025 while revenue rose twelvefold to nearly $4.6 billion. The company spent $7.33 billion on computing that year, three times the prior year. About 80 of the 261-page filing covered risk factors, compared with 48 pages on business plans. The company cautioned that advanced AI might present 'catastrophic or existential risks to humanity', and said models had shown signs of self-preservation, tried to hide or alter information, and acted in ways resembling blackmail. On 23 September, CEO Dario Amodei told the UN Security Council that AI could threaten humankind.

The locked-in infrastructure bill

The filing stated Anthropic planned to commit no less than $518 billion across ten years to six infrastructure partners, and roughly four-fifths of that cannot be canceled or must be paid whether or not it is used. The company listed minimum amounts owed to Google ($111.1 billion), Amazon ($110 billion) and Microsoft ($31.4 billion), along with roughly $161.2 billion of lease obligations tied to Broadcom equipment. A separate xAI agreement, worth up to $84.5 billion, was largely cancelable with 90 days' notice. A Wall Street Journal report on 20 October said the draft prospectus put revenue for the six months to June 2026 near $16.3 billion, compared with about $1.1 billion in the same period a year before. Compute spending for the half reached nearly $9 billion, more than in all of 2025. The net loss for the half was about $81 billion, roughly $72 billion from accounting charges tied to convertible notes. Total infrastructure commitments now exceed $600 billion, and another $150 billion in deals are under negotiation. A pair of clients made up around one-fifth of sales in the first half.

Governance and what comes next

Anthropic is a public benefit corporation. Its Long-Term Benefit Trust has authority over AI safety and board appointments, while the ordinary board manages daily decisions. The public benefit fund is structured as a wholly owned LLC and is barred from lobbying or political giving, but Anthropic's board controls how much it donates and to whom. Critics including Elon Musk and David Sacks say the company should give the money away outright instead of keeping control. Anthropic has not said whether the additional equity will come from new shares or existing holders. The full prospectus must be published at least 15 days before the investor roadshow begins.