A withdrawn draft
Anthropic is preparing for one of the largest initial public offerings ever attempted by a technology company, but the terms disclosed so far are not final. According to Reuters, a draft sent to banks indicated an intention to raise up to $69.3 billion. The documentation was, however, immediately corrected: a company representative warned the financial institutions that the pricing information was not accurate and that the prospectus would be updated. Anthropic did not specify which elements were incorrect, so the figure should be considered provisional.
The version reviewed by the press envisaged the sale of about 866.7 million Class A shares at an indicative price of between $70 and $80. About 843.3 million would have been issued by the company, while just over 23 million would have been sold by existing shareholders: proceeds from the new shares would have gone into Anthropic's coffers, while those from the secondary tranche would have gone to the sellers. The document also indicated a private investment of $26 billion at the IPO price, with the participation of Google, Nvidia, AMD, Broadcom, Micron, Microsoft, and TSMC. These details belong to the withdrawn version and remain to be confirmed.
Powerful but concentrated partners
The presence of these groups reflects Anthropic's role and, at the same time, its dependence on its partners. Google and Amazon are investors, providers of compute capacity, and commercial channels through which businesses buy Claude: according to Reuters, in 2025, 47% of sales came through the two groups' cloud marketplaces. Such a relationship guarantees access to infrastructure and customers, but entails concentration risks and possible conflicts of interest, given that Google and Amazon are also competitors in the development of artificial intelligence models. Anthropic's growth thus remains tied to the strategies of the technology giants that finance it.
Rapid growth and complex financials
The results show sustained expansion. In the three months ended September 30, 2026, Anthropic recorded revenue of $16.96 billion, up 47% from the previous quarter, but with a slowdown compared with the +144% of the second quarter. At the end of September, the company had more than 11,900 customers with annualized revenue of at least $100,000, versus more than 1,500 at the end of 2025. In the third quarter, the operating loss was $567 million and the net loss was $23.29 billion; the latter was mainly affected by the revaluation of convertible financial instruments and does not equate to a loss generated by operations. In the first nine months of 2026, the reported net loss was $104 billion, largely due to accounting changes.
The most demanding item concerns computing power: in the prospectus filed with the Securities and Exchange Commission, Anthropic reports contractual commitments for infrastructure and compute capacity amounting to about $895.4 billion, of which $447.5 billion by the end of 2030 and $69.8 billion due in the twelve months following September 30, 2026. These are not sums already spent or debts immediately payable, but obligations spread over many years. The company warns that some suppliers depend on external financing to buy chips and build data centers: if credit became more expensive or harder to obtain, part of the promised capacity could be delayed, reduced, or cost more.
The listing will be a test for Anthropic and for the entire AI market. According to Reuters, valuation expectations have been scaled back: after estimates of up to $2 trillion, about $1.5 trillion would have become the upper limit discussed by some investors, though this is not an official valuation. The central question will not only be how much the company manages to raise, but on what terms investors will be willing to finance its growth.
