Anthropic has moved from second place to first in enterprise AI, and the financial signals explain why

Dario Amodei describes the shift plainly: Anthropic has gone from second place to first in the AI race. That claim, coming from a founder, would ordinarily be discounted as competitive posturing. The spending and adoption data circulating among enterprise observers makes it harder to dismiss.

Harry Stebbings puts the sharpest number on the table: Claude’s enterprise market share climbed from 21% to 48%, while Gemini moved from 27% to 40% over a comparable period. Public reporting from Enterprise Technology Research, surfaced via The Wall Street Journal, corroborates the broad direction of that shift, describing Claude’s enterprise adoption as having nearly doubled in a span of roughly twelve months. Whatever the precise window, the trajectory is not disputed.

The financial signals underneath the share data are, in some ways, more revealing. Krishna Rao reports that Anthropic’s net dollar retention rate exceeds 500% on an annualized basis. That figure describes not just acquisition but expansion: customers are not holding steady, they are spending dramatically more over time. Nico Laqua offers a concrete illustration of what that retention looks like at the company level, disclosing that his firm spends approximately $400,000 per month on Anthropic and nothing on OpenAI. Andrew Lee adds a different angle on the same dynamic: when users churn from his product, 80% of them migrate to an Anthropic offering, most often because they already hold a Max subscription. Churn that consistently lands at one competitor is not random attrition. It is a signal about where enterprise users have decided to consolidate.

Our net dollar retention rate is over 500% on an annualized basis. Krishna Rao

The product story behind these numbers is less mysterious than it might appear. Martin Casado attributes Anthropic’s momentum to a deliberate bet made under resource constraints: Anthropic, having raised less capital than its rivals, chose to concentrate on coding. That focus held, and the bet paid off. The pattern is now extending beyond its original beachhead. Krishna Rao notes that co-work, a product built on Claude, is being adopted faster than Claude Code was at the equivalent point after its own launch, suggesting the coding foothold is being used to expand into adjacent enterprise workflows rather than treated as a ceiling.

The competitive landscape at the model level reflects the same shift. Gavin Baker describes the Pareto frontier of AI model intelligence versus cost as now being dominated by Anthropic and OpenAI, with Grok joining them, whereas Google held that position nine months ago. Yasser Elsaid, describing usage patterns at Chatbase, reports that roughly half of model usage runs on OpenAI and half on everything else, with Anthropic and Google making up the bulk of that second half. The market has moved from a single dominant provider toward something closer to a two-provider structure at the frontier, with Anthropic occupying one of those two poles.

What is less clear is whether the institutional advantages OpenAI built through its Microsoft distribution relationship can be offset by model quality and retention economics alone. Amodei’s claim of a first-place position is directionally supported by the spending data, but the competitive picture Stebbings and Baker describe is one of rapid movement rather than settled hierarchy. The 500% net dollar retention figure Krishna Rao cites is the kind of number that, if it holds at scale, forecloses the question. Retention at that rate means the customer base is effectively self-compounding. Whether Anthropic can maintain that rate as the installed base grows beyond its current concentrated, technically sophisticated user profile is the question the current evidence cannot answer. What it can support is the simpler observation: the enterprise AI market is no longer OpenAI’s to lose. It is now a contest, and Anthropic is, by the available measures, winning it.