Anthropic's S-1 Shows 11 900 Customers Paying $100 Thousand or More a Year

Anthropic has filed a Form S-1 with the SEC, and the most interesting part is not just the growth but how the company holds on to it. More than half of customers spending at least $100 thousand a year now use both the Claude platform and a first-party app such as Claude Code or Cowork. At the end of last year, that share was about 30%. Net revenue retention at the end of the third quarter was roughly 900%, and revenue for the first half of 2026 reached $16.3 billion, up from $1.2 billion in the same period a year earlier.

The customer base in numbers from the S-1

As of September 30, the company counted:

• more than 11 900 customers spending $100 thousand or more a year; • more than 1 900 customers spending $1 million or more; • about 250 customers paying $10 million or more a year; • more than 50% of customers in the first group use both the Claude platform and a first-party app (Claude Code or Cowork), versus about 30% at the end of 2025; • net revenue retention of about 900% for the third quarter; • revenue for the first half of 2026 of $16.3 billion, versus $1.2 billion a year earlier.

Net revenue retention of 900% means existing customers are spending several times more, on average, than they did a year ago. Anthropic itself attributes this to customers rolling Claude out across their entire organizations once they see the value.

The Amazon and Uber playbook

The pattern is familiar. Uber started with rides, then added food delivery and freight. AWS hooked customers with cloud computing, then expanded the relationship to storage, databases and other services, and leaving that ecosystem became harder and harder.

Anthropic follows the same logic. The company doesn't have to sell each product from scratch to a new buyer: a business arrives for the API, then adds Claude Code for its engineers, then Cowork for everyone else. The S-1 calls this an "integrated approach": an improvement at any of the three layers (models, developer platform, applications) strengthens the other two. The filing also takes a jab at the entire classic SaaS industry, arguing that Claude is not a "static product for a single use case" but frontier AI with rapidly growing capabilities.

The main argument for this strategy is the spending gap. In the third quarter, the top 5% of Claude Enterprise customers by spend who used Claude Code spent roughly 20 times more per month than the average enterprise user without that tool. For Anthropic, this is a measure of headroom: the more accounts that arrive through the API and then adopt the apps, the higher the revenue per user.

Applied AI, Blackstone and cloud partners

The company gives organic adoption a manual push. An internal Applied AI team helps startups and corporations embed Claude into their existing technology stacks. Separately, Anthropic recently launched Ode with Anthropic together with Blackstone and other private equity firms: forward-deployed engineers go into companies and build custom AI tools on top of Claude.

Distribution also runs through the hyperscalers: AWS, Google Cloud and Microsoft Azure. Under the "joint go-to-market" model, Anthropic keeps the direct contract with the customer and pays the cloud provider a commission. According to the company, this lets it "avoid channel conflict" and "set prices" itself.

The risks the company doesn't hide

Growth looks impressive against this backdrop, but the risk section is just as interesting to read. The cloud partners that help sell Claude increasingly see Anthropic as a competitor. By the company's own admission, this may complicate its access to compute: Amazon, Google and Microsoft could prioritize their own needs or those of competitors in which they hold stakes.

The second risk is concentration. A significant cut in spending, or the loss of one or more of its largest customers, could hurt the business, financial condition and results of operations if new customers or expanded use by existing ones do not make up the difference.

At the same time, Anthropic says it will spend less effort than a typical company on reinforcing its current products and business models, and more on identifying new ones. The goal is to ship "new classes of products" that will eclipse the economic value of today's.

For the market, the takeaway is practical. Enterprise AI is shifting from a choice of model to a choice of ecosystem: whoever enters through the API gets a free sales channel for apps inside accounts that are already paid for. Competitors will have to respond not only with model quality but with a bundle of platform, applications and implementation services. Buyers should weigh the downside: the deeper Claude is embedded in their processes, the costlier it becomes to switch vendors, and the terms on pricing and compute depend more and more on Anthropic's relationships with the cloud giants.