🚨 The first existential IPO

Anthropic looks set to be the first of these AI platforms to hit the public markets, and Reuters got hold of its IPO prospectus, the S-1, yesterday. That filing turns the usual startup story on its head: not a tidy growth tale, but a company spending hard for scale and showing the first signs that the math may eventually work.

The headline numbers are wild. For 2025, the filing points to an operating loss of $8bn and a net loss of $42bn, with the bigger figure pushed up by an accounting charge. But the more interesting bit is what sits underneath: revenues are rising much faster than costs, and that gap is expected to keep widening into 2026.

That matters because it suggests this is not just a money furnace. The company is heading toward profitability next year, according to the analysis of the filing. That would be a major shift for a business that is still deep in the red, and it explains why investors are likely to stare past the losses and focus on the growth curve.

Then there’s the scale of the ambition. Anthropic has $518bn in compute commitments stretching over 7 to 10 years. About 80% of that, roughly $410bn, cannot be cancelled. If Anthropic wants a decent margin on top, the math points to about a trillion dollars in sales over that period. And by the end of 2026, annualised revenue is expected to be above $100bn.

The rumour mill says the IPO could happen in November. If that happens, this won’t just be another listing. It will be a very public test of whether investors are willing to finance AI companies that look absurd on paper before they start to look like businesses.