Anthropic Files for Nasdaq IPO, Opening Its Books to the Public

Anthropic filed its public S-1 with the U.S. Securities and Exchange Commission on Monday, formally launching what is expected to be the largest AI listing in history. The company behind Claude plans to trade on the Nasdaq Global Select Market under the ticker ANT, putting six years of frontier-lab finances on the public record for the first time.

The registration statement, filed with the SEC on October 26, lists Anthropic, PBC as a Delaware public benefit corporation offering Class A common stock. Price and share count are left blank in the preliminary prospectus, so the size of the raise is not yet public.

What the numbers actually show

Revenue went from $41 million in 2023 to $386 million in 2024 and $4.58 billion in 2025. For the first six months of 2026 alone it hit $16.29 billion, against $1.16 billion a year earlier, according to CNBC's coverage of the filing.

Costs are just as blunt. Compute and infrastructure reached $8.94 billion in the first half of 2026, up from $2.33 billion, with personnel at $3.15 billion and sales and marketing at $3.17 billion. Operating loss for the period narrowed to $693 million from $2.14 billion.

Why the $80 billion net loss is smaller than it looks

The headline net loss for the first half of 2026 is $80.75 billion, which reads like a catastrophe until you open the line items. Of that, $72.39 billion is a non-cash charge from the change in fair value of convertible notes, derivative liabilities and SAFEs - an accounting artifact of a private company's rising valuation, not cash leaving the business.

What matters for public investors is the operating line, and it improved sharply. Anthropic also disclosed $1.65 billion of deferred revenue, $10.8 billion of remaining performance obligations as of June 30, and a customer base where the top five accounts produced roughly 25% of revenue for the first nine months of 2026. No single customer exceeded 10%.

What it means for builders

For anyone shipping agents on Claude, the filing is a map of the business model they depend on.

Consumption pricing is the whole game: Anthropic calls its model "pay-for-intelligence," metered in tokens, and says it is the substantial majority of revenue. The company is also contractually locked into years of compute - the filing warns that commitments include take-or-pay terms tied to AWS Trainium, Google TPUs, Nvidia and AMD silicon, while revenue is largely usage-based and cancellable. And marketplace distribution now carries the business: 48% of first-half 2026 revenue arrived through cloud provider marketplaces, up from 32% in 2024.

The practical read is that pricing, rate limits and model access will be shaped by quarterly earnings pressure from here. For teams building on a single frontier provider, the S-1 is a good moment to price multi-provider fallbacks into the architecture.