The S-1 Anthropic filed with the SEC on Monday, October 26, 2026, puts the company's enterprise customer tiers on the public record for the first time. TechCrunch AI's desk report on the filing breaks out the cross-product adoption curve and the revenue trajectory the company is using to justify its listing. The filing is a regulatory document, not a vendor page, and the desk report is the only page in this packet that states the figures below.

The table carries the numbers. More than 11,900 customers clear the $100,000 annual threshold; 1,900 clear $1 million; roughly 250 clear $10 million. First-half 2026 revenue sits at $16.3 billion against $1.2 billion in the same period a year earlier. Net revenue retention, as the filing states it, is about 900 percent. The cross-product figure is the one that changes the buyer's calculus: more than half of the $100,000-plus cohort now runs both the Claude platform and a first-party app such as Claude Code or Cowork, up from about 30 percent at the end of the prior year.

The 20x multiplier is the number a procurement team should circle. The filing says the highest-spending five percent of Claude Enterprise customers who use Claude Code spend roughly twenty times the monthly per-user amount of the average enterprise user who does not. That is not a pricing tier. It is a usage pattern the company is engineering into the account. The API entry point becomes a funnel, and the first-party apps are where per-seat spend compounds. Anthropic is not selling a model; it is selling a deployment path that widens as the customer's organization adopts more layers.

What the table does not prove is whether the cloud partnerships that distribute Claude will keep distributing it. The filing names Amazon, Google, and Microsoft as joint go-to-market partners, then flags in its risk section that those same providers may prioritize their own models or the rivals they have invested in. The company holds the direct contract and pays each cloud a fee, which it calls a way to "avoid channel conflict." That structure works while the cloud providers are selling compute. It becomes a negotiation when the cloud providers are selling a competing model on the same rack.

The filing does not state a per-seat price, a gross margin, or a compute allocation from any of the three hyperscalers. The 900 percent retention figure is a single-quarter snapshot, and the customer-concentration risk paragraph acknowledges that the loss of one or two of the largest accounts would move the P&L in a way the revenue growth line does not yet show. The no-ads pledge that other desks have attached to this filing does not appear in the text TechCrunch AI reported from.

The 20x multiplier turns the API entry point into a funnel rather than a product, and the buyer's per-seat cost compounds as the organization adopts more first-party layers. The cloud partnerships that distribute Claude are the same providers selling competing models, and the filing's own risk section concedes the channel can flip.

The filing does not state a per-seat price, a gross margin, or a compute allocation from AWS, Google Cloud, or Microsoft Azure.

The cross-product adoption curve (30% to >50% year-over-year) means the API is a top-of-funnel, not a destination. The 20x spend gap on Claude Code users is a usage pattern, not a price tier, and it depends on the hyperscaler partners continuing to route compute to Anthropic rather than to their own models.