A difference of $600bn would be a rounding error in most stories. In Anthropic's, it is the gap between the valuation the market was told to expect and the one the company has now filed. According to a US regulatory document published on Tuesday, the developer of the Claude chatbot is seeking a valuation of up to $1.4tn in its stock market flotation, well below the $2tn that had been widely reported.
The mechanics are enormous even at the lower figure. Anthropic plans to sell 825m shares at up to $78 each, raising as much as $64.4bn. A separate private placement worth $26bn is lined up with cornerstone investors, including Alphabet, Google's parent company, and the chipmaker Nvidia.
If demand is strong, a so-called greenshoe option could add $9.7bn, bringing the total to roughly $100bn. Trading is expected to begin next week. Why the retreat?
Analysts point to a change of mood rather than a change in the company's fortunes. Last week's US midterm elections produced victories for opponents of datacentres, sharpening an older worry: that firms are spending vast sums on AI infrastructure with no clear timetable for earning that money back. Brian Hopkins, a principal analyst at Forrester Research, argues that investors were looking for a reason to correct valuations they already found uncomfortable, and the election supplied one.