Anthropic IPO Filing Puts CEO Dario Amodei’s ‘No Hedge on Earth’ Rule to the Test

When Dario Amodei explained this winter why Anthropic would not buy as much computing power as its growth curve seemed to justify, he reached for the language of ruin. Guess the timing of AI revenue wrong by a single year, the Anthropic chief executive said, and no hedge could save the company. The Claude maker’s IPO filing is the first document in which investors can check how closely he followed his own rule.

Quoted in coverage of the filing, the line reads like a confession, proof that the company is wagering everything on one curve. Read in full, it says something narrower and more useful. Amodei was describing what would happen if he bought too much, too early. The prospectus shows what he bought instead, and which parts of it he can still walk away from.

The quote Amodei actually gave

He said it on Dwarkesh Patel’s podcast in February, when asked why Anthropic was not spending more aggressively. “If my revenue is not $1 trillion, if it’s even $800 billion, there’s no force on Earth, there’s no hedge on Earth that could stop me from going bankrupt if I buy that much compute,” he said, according to Fortune’s account of the interview. The last clause carries the meaning. Business Today’s analysis of the IPO quoted the sentence without it, which turns a reason for restraint into a description of the bet. His argument was that he could not sign up for a trillion dollars of compute a year in 2027, because growth of five times a year instead of ten would sink the company.

What Anthropic signed instead

The S-1, made public on Monday for a Nasdaq listing under the ticker ANT, puts numbers on that restraint. Compute and data-centre deals signed or amended since the end of 2025 carry roughly $417 billion in minimum payment obligations, on terms of up to 20 years. Capacity ordered for delivery between 2028 and 2031 adds an estimated $125 billion to $150 billion. Revenue for the first half of 2026 was $16.3 billion, up from $1.2 billion a year earlier.

The structure of those deals is where Amodei’s bankruptcy line now lives. The largest blocks sit with Anthropic’s cloud partners, and they do not bend: at least $110 billion with Amazon through April 2036, about $114.5 billion with Google through July 2033 and at least $44.9 billion with Microsoft through May 2033, each committed “regardless of actual usage”. If spending with Amazon or Google falls short, Anthropic pays the difference. All three are non-cancelable unless the provider commits an uncured material breach.

The GPU capacity rented from SpaceX’s AI unit works differently. That deal runs to as much as $120.5 billion, around $3.1 billion a month once fully delivered, but after initial periods ending in early 2027 either side can cancel on 90 days’ notice. It is a hedge, of a kind, and the smaller one. The commitments Anthropic cannot exit are the ones that run for most of a decade.

The filing concedes the exact risk Amodei described. It calls compute forecasts “inherently uncertain”, admits that “in the past, we have significantly underestimated our compute requirements”, and warns that capacity arriving out of step with demand could mean “committed cash outflows that do not produce immediate revenue”.

The loss that is not the problem

The figure that dominated the headlines, an $80.7 billion net loss for the half, is mostly accounting. Of it, $72.4 billion came from changes in the fair value of convertible notes, derivatives and SAFEs. The operating loss was $693 million, down from $2.1 billion a year earlier. CNBC’s reading of the filing puts the implied gross margin at 85%, against 12% in the same period of 2025, and net revenue retention stood at roughly 900% in September: existing customers are multiplying what they spend ninefold. By the measures Anthropic chose to lead with, the business works.

The margin arrives with its own warning. Operating margins, the prospectus says, will “fluctuate over time, particularly in periods when substantial incremental compute is delivered to us”. Put plainly, the profitability on show belongs to a company whose biggest bills have not fully arrived.

Who carries the risk after the listing

Anthropic says the proceeds will go to general corporate purposes, explicitly including “investments in compute and infrastructure”. Public shareholders will fund part of that build-out without steering it. A single Class F share, held by a company owned by the seven co-founders, will carry 50.1% of the vote, and a Long-Term Benefit Trust generally holds the power to elect a majority of the board. The filing leaves the price range and share count blank; market reports cited by Business Today put the hoped-for valuation between $1.5 trillion and $2 trillion.

The timing is not kind. CNBC notes that the neoclouds CoreWeave and Nebius each lost more than 30% of their value in October as credit worries spread through the data-centre trade, the same build-out Anthropic’s long commitments rely on.

In February, Amodei said no hedge on Earth could save a lab that guessed the timing wrong. With ANT, he is about to sell one.

Tags: Anthropic, AI infrastructure, Dario Amodei, Nasdaq, IPOs