Most people will read the Anthropic S-1 for the headline numbers. Operators shou
By Dan Ashburn · November 2, 2026 · Curated by George's Blog
Most people will read the Anthropic S-1 for the headline numbers. Operators should start with the risk factors.
That section tells you where the real friction sits. Compute, regulation, how the metrics are calculated, who controls the votes. It is the most honest part of any filing, and it starts on page 26 of this one.
Then read the customer numbers, because they tell a story about adoption.
Customers spending $100,000 or more a year with Anthropic went from more than 950 to more than 11,900 in twelve months (September 2025 to September 2026, per the filing).
That is not a pilot curve. The filing itself says customers usually start with a single use case and then expand across departments and workflows.
That matches what I see. The businesses getting results have stopped testing AI in a corner. Our team agents run real work every day, owned by real people, with clear jobs and clear checks.
Three things I would take from this filing as an operator.
1. Read the risks before the growth. Your AI plan has its own risk factors. Write them down.
2. Pick one workflow and take it all the way to production. Then expand.
3. Track cost per task, not cost per seat. The filing says one model update cut customer cost per task by about 40%.
Pilots were 2024. 2026 is about who has AI doing the actual work.