Anthropic S-1 shows $16.3 billion revenue, $850 billion in commitments

Anthropic's filing arrives with a tension built in. On one side, the company has been more disciplined than OpenAI or Elon Musk, avoiding consumer-side experiments like AI image generators or TikTok clones and specializing instead in coding and cybersecurity — tasks that are easy for a computer to do and that sell into enterprise budgets. That focus shows up in the customer mix: about three-quarters of first-half 2026 customers spend at least $100,000 a year, and net revenue retention sits at 900 percent. The adjusted income of $1.62 billion in the first half of 2026 suggests that, at least for now, the business can pay for itself.

On the other side, the commitments tell a different story. The S-1 lists $265.6 billion in non-cancelable purchase commitments with terms between one and 20 years, $321.4 billion in lease agreements primarily related to compute and infrastructure capacity that have not yet commenced, about $245 billion in new compute and infrastructure agreements, and an estimated $125 billion to $150 billion in transactions not yet finalized. Together these point to a company that must keep expanding to service what it has already signed.

The filing also flags a risk tied to how Anthropic builds its models. Claude is already building Claude under human direction, and the company warns that using AI to develop more AI may cause capabilities to advance faster than its ability to develop, validate, and implement corresponding safety, alignment, and security measures. In other words, the business plan appears to depend on a pace of improvement that the company itself describes as carrying risk to everyone — a tension that sits at the center of the S-1.