Anthropic S-1: 19% operating margin, but $808 billion at risk

Anthropic's S-1 arrives as a test of whether a leading AI lab can pair rapid revenue growth with something closer to sustainable unit economics. The filing shows revenue per dollar of compute cost improved 50% between the first and second quarters of 2026, and the company reported $3.9 billion in first-half free cash flow. Those figures contrast with the scale of its commitments: the S-1 lists at least $602 to $627 billion in data center and infrastructure spending, rising to as much as $808 billion when chip deals are included. That gap between current profitability and future obligations is the central tension the document exposes.

The filing also carries governance and risk disclosures that sit uneasily beside the financial picture. Control of the board rests with the Long-Term Benefit Trust and the seven cofounders, while regular investors and backers Amazon and Google hold shares with little practical voting power. The prospectus warns of risks ranging from authoritarian misuse of AI to the potential for AI to pose existential risks to humanity.

What the S-1 does not resolve is whether revenue growth can keep pace with the coming spending. Third-quarter results were omitted even though Anthropic briefed investors on them on October 14, and the figures that can be imputed suggest growth slowed: revenue for the three months ended September 30 was about $16.8 billion, roughly 45% above the second quarter, well below the 147% sequential growth earlier in the year. How convertible notes and non-cancelable compute contracts affect the true economics is another open question investors will likely press before shares begin trading.