A fast-growing business with very large fixed bills. Anthropic filed for its IPO in October 2026. The filing shows seven computing deals, most running to 2033 or later. By Amar Harolikar's rough estimate, the deals it must pay for whether used or not need about twice today's revenue.
- Growth: revenue grew 45% in July-September on the quarter before, after 144% the quarter before that. If growth slows to 30%, 25% and 20% a quarter, revenue reaches the must-pay bill by April-June 2027, while the bills rise as new capacity arrives.
- Cash: at 30 June it held $132 billion of the $141 billion it has raised, which gives it time.
- Profit: April-June had an operating profit of about $1 billion, before $26.6 billion of staff share-award costs that start once it lists.
- Staff: about 5,300 people on 30 September, up from about 940 at the end of 2024. Anthropic puts revenue per employee over the last 12 months above $6.6 million, and says over 90% of the code added in September was written by Claude.
How the estimate works. The filing gives no yearly split, so each deal is spread evenly over its term, with SpaceXAI at its full $3.1 billion a month. That gives roughly $115 billion a year while all seven deals overlap, around 2027-29. SpaceXAI can be cancelled on 90 days' notice from early 2027, others end after 2029, and more TPU capacity is already committed for 2028-31.
The other risk is rivals catching up. NIST puts GLM-5.3, from China's Z.ai, about four months behind the best US models on its cyber tests. Anthropic's own filing says others have tried to copy and distill its models, and that open-weight models could take parts of the market and push prices down. It also says that if strong models become easy to build, competition could depend more on access to computing on good terms. With most deals running to 2033 or later, the lead has to hold for years. If it slips, prices fall while the bills stay fixed.
Amar Harolikar's read: Claude Code is his bread and butter, and he sees a very hard climb ahead. Covering the must-pay deals looks within reach. Covering all of them, and holding the lead for years, is the hard part. The filing's results stop at 30 June, and July-September results are expected in an updated filing before the shares are priced. Analysis of public sources, for information only. Not investment advice.
Full post with the deck, the working and the caveats: https://www.tigzig.com/post/anthropic-s1-must-pay-deals-rivals-oct2026. Related: can a freely downloadable AI model build cyber exploits?, how AI has changed work inside Anthropic, how the hyperscalers are paying for the build and who is warning about AI debt and returns.
Building something like this? How I work covers the rates, the availability and what I take on.