Anthropic reports $16.3 billion revenue and $80.7 billion loss in IPO draft

Anthropic, founded in 2021 by Dario Amodei and six co-founders after splitting from OpenAI, built its Claude models into a business that grew revenue from $1.2 billion a year earlier to $16.3 billion in the first half of 2026. The S-1 filed with the SEC sets up a Nasdaq listing under 'ANT', following a wave of AI firms testing public markets after years of private mega-rounds that valued the company near $2 trillion.

Investors buying ANT shares inherit a company locked into paying roughly $265.6 billion for computing power whether or not clients use that capacity, a sum about 16 times first-half revenue. That obligation falls on cloud partners like Amazon and Google as much as on Anthropic, and it means future cash flow is already spoken for. Shareholders gain exposure to Claude's growth but also to a cost structure that can't easily shrink if demand slows.

Dario Amodei and six others retain 50.1% of voting power after the listing, so public buyers get one vote per share but no path to outvote insiders on anything that matters. That is the same founder-control model Meta and Alphabet popularized, and it raises a fair question about what 'going public' actually transfers: capital flows to the company, but governance stays exactly where it was before the IPO.

The S-1 made public this week still lacks a price range or share count, so the real test comes when Anthropic sets terms and prices the deal in the weeks ahead. A valuation that holds near $2 trillion despite an $80.7 billion first-half loss would validate the bet that revenue growth outruns the compute bill; a markdown would signal that investors are pricing the $265.6 billion obligation as the bigger risk.