The most important line in Anthropic’s S-1 is on page 188.
In April 2026, Anthropic’s internal use of Claude, priced at what a customer would pay, passed its monthly cash payroll. Over 90% of the code added to its codebase in September was written by Claude.
Three days after the filing, METR published an outside assessment based on eleven weeks inside the company. It found Anthropic is likely improving its models about twice as fast as it did through 2024 and most of 2025. It credits about 80% of that speedup to Anthropic’s own use of AI, and it saw no bottleneck coming in the next year.
A few things I took from it:
Revenue was $17.0B in Q3 alone, up 1,127% from a year earlier. The $104B net loss is mostly an accounting charge that grew because the company became worth more.
Fortune 500 companies spend less than 0.1% of their expenses on Anthropic. If that rises to a few percent, most of the money will come out of payroll.
The lab is building its own apps, industry-specific products, a services business and robotics tools. Startups that exist to patch a weakness in today’s models last about six months.
Startups get cheaper to build, so there will be more of them. That’s good for small-check funds and hard on big-check funds.
In the same year, the U.S. government tried to cut Anthropic off, restricted exports of two of its models, and saw bills proposed that would allow nationalization.
The singularity used to be something people argued about at conferences. Now it’s a risk factor in a securities filing, attached to a ticker.
Full piece: https://teamignite.vc/blog/the-singularity-is-a-line-item