Anthropic, maker of the Claude AI platform, has committed to spend $895bn, almost all of it on computing power and infrastructure, even as its revenue grew twelvefold to about $17bn in the quarter to Sept. 30.

Operating expenses rose 292% to $17.5bn, which tipped the company back into an operating loss of $567m after a $974m operating gain in the prior quarter. The net loss was $23.3bn, against $12.1bn a year earlier.

Of the $895bn, about $448bn falls due by the end of 2030 and roughly $70bn within two years, through September 2027. A further $100bn has been added since Sept. 30. That is the same scale as Meta's roughly $825bn build-out, which came before Meta turned to preferred stock and forward-sale financing. It also lands where credit markets are already uneasy: record levels in credit default swaps, which are contracts that pay out if a borrower fails to repay, and wider corporate bond spreads at Oracle, Meta and Alphabet reflect worry about debt-funded AI spending.

Anthropic filed for its IPO on Oct. 26, and the offering is expected before Thanksgiving. Shares will be sold by the company and some of its backers, in the first IPO from a firm devoted entirely to building frontier AI models. Bloomberg reports that bankers on the deal say investors see fair value at $1.4tn to $1.5tn, with some seeing as much as $1.75tn.

OpenAI has put off its own IPO to next year and is raising tens of billions of dollars privately instead. Its private backers have shown a tolerance for losses in return for growth. At a valuation in that range, Anthropic's offering looks like the first real test of whether public buyers will show the same.

© Raydn · raydn.us