Anthropic's $81bn listing puts AI's losses on the public record

- Anthropic priced shares at $83 and raised $81bn, with trading due to begin in New York on 19 November.
- The flotation values the five-year-old company at roughly $1.5tn, above Walmart and Samsung.
- The IPO document reports $33bn of revenue in the nine months to September against $104bn of net losses.
- Total proceeds could reach $107bn once investments from Google, Microsoft and Nvidia are counted.
Anthropic, the company behind the Claude chatbot, priced its shares at $83 and raised $81bn in a flotation that the BBC’s report on the AI firm’s market debut describes as a record, with trading due to begin in New York on 19 November. Demand outstripped the proposed $70 to $78 range. More stock was sold than originally planned, and the five-year-old business now carries a valuation of roughly $1.5tn, above Walmart and Samsung and close to Tesla and Meta.
The company was set up by seven former OpenAI employees, led by Dario and Daniela Amodei, and is the first purely AI-focused American firm to sell shares to the public, with OpenAI expected to follow next year. Once investments from Google, Microsoft and Nvidia are included, total proceeds could reach $107bn.
The disclosure attached to the listing is more informative than the headline figure. Revenue reached $33bn in the nine months to September, while net losses climbed to $104bn, a gap of roughly three to one. Private funding rounds can absorb that arithmetic quietly: backers agree a price among themselves and publish little. A public listing cannot. Filed accounts, quarterly reporting and securities law convert projections into statements that named people must stand behind.
Analysts quoted alongside the flotation point to heavy financial obligations, chip shortages and regulatory pressure following the US midterm elections as reasons to doubt future profitability. Those are not incidental risks. They describe a business whose costs are set by hardware supply and whose operating environment is set by legislators. It controls neither.
For institutions weighing whether to build a chatbot into teaching and assessment, the relevant detail is the cost structure rather than the valuation. A vendor that spends more than it earns depends on continued access to capital. Procurement decisions that treat such a tool as permanent infrastructure inherit that dependence without usually saying so.
The filing now places some of this on the public record, which is unusual for a sector that has grown mostly on private money. Whether that scrutiny changes how the technology is bought, taught with and governed, or simply runs alongside it, is a question the documents cannot answer on their own.