Anthropic Passed OpenAI in Revenue Months Early, and the Enterprise-Consumer Split Explains Why
Anthropic hit a $47 billion annualized revenue run rate in May 2026, up from $9 billion at the end of 2025, passing OpenAI's roughly $25 billion annualized run rate months ahead of the crossover analysts had projected for August. Anthropic's May 28, 2026 Series H valued the company at $965 billion post-money, ahead of OpenAI's $852 billion.
The split matters more than the headline
The revenue mix behind these numbers explains the crossover better than any single product decision does. Roughly 85% of Anthropic's revenue comes from enterprise sources, the API, Claude Code, agentic tooling embedded into other companies' products. Roughly 85% of OpenAI's revenue comes from consumer ChatGPT subscriptions. These are two companies that have effectively won two different markets, not one company beating the other at the same game. Anthropic optimized for the workload technical buyers actually pay the most for, developer-facing and agentic infrastructure, while OpenAI optimized for the largest possible individual subscriber base.
Why the split should shape how you weight these two vendors
Revenue mix is a reasonable proxy for where a company's product investment actually goes. A lab earning 85% of its revenue from enterprise and developer tooling has a direct financial incentive to keep investing in exactly the surface area a technical buyer cares about: API reliability, agentic tooling, developer experience. A lab earning 85% from consumer subscriptions has the opposite incentive, keep improving the product that drives individual subscriptions, which doesn't always overlap with what an enterprise API customer needs most.
Neither mix is objectively better, they're built for different customers. But if you're picking a model provider specifically for enterprise or developer-facing work, the revenue mix is a legitimate signal about which vendor's roadmap is actually built around your use case, independent of which one currently tops a capability benchmark.