Anthropic's quiet engine: getting API buyers to adopt its apps

Photo by Conny Schneider on Unsplash

The number everyone will quote from Anthropic's S-1 is $16.3 billion: revenue for the first half of 2026, against $1.2 billion in the same period a year earlier. That is more than 13 times higher in twelve months.

A smaller number deeper in the filing says more about where the business is heading. More than half of customers spending at least $100,000 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, the share was about 30%.

What is growing

The enterprise base is large and getting top-heavy. As of September 30, Anthropic counted more than 11,900 customers spending $100,000 or more annually, more than 1,900 spending $1 million or more, and about 250 spending $10 million or more.

The company puts net revenue retention at about 900% as of the end of the third quarter. It credits customers expanding their use as they roll Claude out across their organizations. If that figure holds, existing accounts are spending many times what they did before. That is a different thing from simply signing new logos.

Why the pairing matters

The pattern is a familiar one. Amazon pulled customers into AWS with cloud computing, then extended the relationship into storage and databases. Uber began with rides and added food delivery and freight. Each step deepens the tie without a fresh sales campaign.

Anthropic can run the same play. A company that starts with the API can hand Claude Code to its engineers, then Cowork to the wider workforce. The buyer, the contract and the trust are already in place.

Every app a customer adopts is another reason not to switch models.

The filing offers one data point on what that is worth. In the third quarter, the top 5% of Claude Enterprise customers who used Claude Code spent about 20x as much per month as the average enterprise user who did not. That compares a heavy-user slice with an average, so it is not a forecast for every account. It does suggest how much headroom per-user spending may have.

Anthropic is also building the machinery to push this along. It runs an internal Applied AI team that helps place Claude inside customers' existing systems. It recently launched a venture with Blackstone and other private equity firms, called Ode with Anthropic, that embeds forward-deployed engineers in businesses to build custom tools.

The cloud channel works differently from the usual reseller model. Anthropic keeps the direct contract with the customer and pays a fee to AWS, Google Cloud or Microsoft Azure. The filing says that arrangement lets it set pricing and avoid conflict between sales channels.

Where the stickiness could slip

The same hyperscalers are a pressure point. The filing says they increasingly view Anthropic as a competitor. That could make it harder to secure compute from them, since Amazon, Google and Microsoft may put their own needs, or those of rivals they have backed, first.

The dependence is large. The S-1 lists commitments of $110 billion to Amazon over ten years, at least $114.5 billion to Google through 2033 and $44.9 billion to Microsoft through 2033. It also includes about $3.1 billion a month to SpaceX for Nvidia GPU access once planned capacity is delivered, totaling as much as $120.5 billion by 2029.

The financial picture is mixed. The first-half operating loss was $693 million, down from $2.1 billion a year earlier. The net loss was $80.7 billion, which reporting on the filing attributes largely to noncash accounting charges tied to convertible debt. The company itself says it "may not achieve or sustain profitability in the future."

The risk factors also warn that losing one or more of its largest customers could hurt the business if other accounts do not make up the difference. With spending concentrated in a few thousand big accounts, cross-selling is a hedge against exactly that.

What it signals

The pitch to public investors is that Claude is a stack of models, a developer platform and applications, and that progress at any layer lifts the others. The cross-product numbers are the evidence offered for that claim.

That suggests the contest among AI labs is moving from who has the best model to who owns the daily workflow around it. Every app a customer adopts raises the cost of switching providers.

There is a tension in the filing worth watching. Anthropic says it will spend less effort than a typical company on entrenching current products and more on creating new ones. Stickiness is the strategy right now, but the company says it does not want to rely on it.

Anthropic plans to list on the Nasdaq under the ticker ANT. It has not yet said how much it hopes to raise or at what price. When it does, investors will be pricing both the growth and the dependence on cloud partners who are also competitors.