The warning just days before the listing
Just days before what could become the largest initial public offering in history, Anthropic CEO Dario Amodei identified the main fragility in the company’s business model. Speaking on Friday at an investor event hosted by JPMorgan Chase in New York, he summed up the situation with a line reported by The Decoder: “There is enormous uncertainty about what will happen next year.” It is not the first time Amodei has issued a warning of this kind.
The computing-capacity question and the deadlines
According to Amodei, computing capacity represents “the deepest dilemma” for the company. AI technology is advancing so rapidly, he explained, that no one yet knows how this progress will translate into economic value, while hardware must be purchased one or two years in advance, before it is known whether revenue will cover the costs. Buying too little leaves the company without resources for customers and new technologies; buying too much exposes it to the opposite risk, Amodei said: “You can become financially overextended.”
According to Bloomberg, cited by the source, Anthropic has already made commitments of nearly $901 billion, largely for computing capacity and chips. About $71 billion would mature by September 2027, an amount that exceeds the market capitalization of about two-thirds of the companies in the S&P 500. The source also recalls that in February 2026 Amodei had warned, in blunter terms, that a misjudgment of just one year could mean failure, and that in his view some competitors had not fully grasped the risks they had taken on.
The countermeasures and the offering figures
Amodei described two tools for mitigating the dilemma: gross margins and balance-sheet liquidity. He called the company’s gross margins “pretty good,” especially compared with competitors, and spoke of substantial cash reserves, with the intention of raising more. Anthropic also buys a relatively small share of capacity in advance and can turn to the spot market for additional capacity: an approach Amodei likened to insurance. The listing itself, he added, serves mainly as a financial buffer: more cash means a wider margin for error between a shortage and an excess of computing capacity.
According to Bloomberg, the offering aims to raise up to $64.4 billion through the IPO and another $26 billion through a private placement; together, they would make it the largest such transaction ever carried out. Demand is said to be already about six times the offering. Company documents indicate annualized revenue of $81 billion at the end of October, up about $10 billion from the end of September, supported by the popularity of AI products, including coding models.
Amodei said he was not focused on a specific month or on particular market conditions: Anthropic has gone from zero to its current size in six years. The most convincing advice he said he received before the IPO was not to let the company be worth a dollar more than it is truly worth, in line with the internal approach of promising little and delivering much. At the same meeting, however, he expressed hope that AI would cure cancer and Alzheimer’s within 5–15 years.
The central point, then, is that the IPO does not resolve the crux Amodei identified but covers it: revenue will have to prove that it justifies commitments of that size. It remains to be verified, in formal offering documents, how the maturities are distributed and whether revenue is sustainable against costs.
