
Anthropic circulated IPO terms to banks on Monday evening and withdrew them within hours, leaving the market without a confirmed price range a day before the company was due to begin its roadshow.
The document sent to banks showed about 866.7 million Class A shares offered at $70 to $80 each. At the top of that range the offering would raise $69.3 billion. Roughly 843.3 million of the shares were to come from the company and about 23.3 million from existing shareholders.
A second email followed, describing the pricing details as inaccurate and saying the filing would be updated later. It did not say which figures were wrong, and Anthropic declined to comment.
The withdrawn draft also described a $26 billion private placement alongside the offering. Google, Nvidia, AMD, Broadcom, Micron, Microsoft and TSMC were named as buyers, with their shares priced at the final IPO price and the transaction closing immediately after the listing.
No price range has been publicly confirmed. The company’s public registration was filed on Oct. 26, and the 15-day waiting period before marketing can start ends on Nov. 10.
The episode lands as AI-linked assets are under pressure. Nvidia has been hit by a board probe into chief executive Jensen Huang, and SoftBank, widely seen as a proxy for sentiment toward OpenAI and the wider AI sector, reported a quarterly loss of about $7 billion on Tuesday after a US data center writedown.
The retracted figures sit below some of the valuation talk that preceded the filing. Investor talk of a $2 trillion valuation has been scaled back, with $1.5 trillion now the upper limit under discussion.
Revenue for the quarter ended Sept. 30 was nearly $17 billion, a figure disclosed earlier this month in an amended filing. Quarterly growth has slowed significantly.
The placement terms were part of the document that was withdrawn and have not been reissued.
Sources: techzine.eu, Reuters, Bloomberg.









