Anthropic has overtaken OpenAI in revenue for the first time since the two labs began competing head-to-head for enterprise AI budgets, marking a reversal in a rivalry that OpenAI had led by most public estimates for years.

According to The Decoder, the shift caps a period in which Anthropic leaned almost entirely on enterprise and developer-facing products — the Claude API, Claude Code, and direct enterprise contracts — while OpenAI's business has stayed anchored to ChatGPT's mass-market subscription base.

For a company that launched years after OpenAI and without a consumer app anywhere near ChatGPT's scale, closing that revenue gap says less about brand recognition and more about where the money in applied AI is actually concentrating right now: inside company workflows, not chat windows.

Two different businesses, same AI-lab reputation

OpenAI and Anthropic have never really been competing in the same market segment, even though headlines usually lump them together as rival foundation-model labs. OpenAI built ChatGPT into a subscription product used by hundreds of millions of people, plus a developer API layered on top. Anthropic, by contrast, has consistently positioned Claude as an API-first, enterprise-first product, with Claude.ai as a secondary consumer surface rather than the main engine.

That structural difference matters for how each company's revenue behaves:

• Consumer subscriptions scale with user count but are price-sensitive and easy to churn.
• Enterprise API contracts scale with usage volume and are stickier once a workflow is built on top of a specific model.
• Developer tools like coding assistants generate usage that compounds — more code written means more tokens billed, independent of subscriber growth.

Anthropic's revenue passing OpenAI's suggests the enterprise-and-developer lane has grown faster than the consumer-subscription lane over the period The Decoder is describing — even without OpenAI's audience.

Coding is doing the heavy lifting

The clearest driver behind Anthropic's climb has been coding. Claude's models, particularly the Sonnet and Opus lines, built a reputation among developers for reliability in agentic coding tasks, and Claude Code turned that reputation into a standalone product that bills by usage rather than by seat.

That matters because coding workloads are unusually token-hungry: an agent reading a codebase, running tests, and iterating on a fix burns far more tokens than a single chat reply. A relatively small base of professional developers and engineering teams can generate outsized API revenue simply because of how much they run through the model, which is a very different growth curve than adding another consumer subscriber.

What this means for teams building on these models

For anyone building products, agents, or internal tools on top of frontier models, this shift is a signal worth factoring into vendor decisions, not just a scoreboard update:

• Pricing pressure could go either way. A model provider generating strong revenue has less incentive to cut prices aggressively — but also more capital to invest in inference efficiency that eventually lowers costs.
• Coding-specific tooling is now a real business line, not a side feature. Expect continued investment from both labs — and competitors — in agentic coding products specifically, since that's where usage-based revenue is proving out.
• Enterprise-first vendors may prioritize different roadmaps than consumer-first ones. Anthropic's incentives point toward deeper API capabilities, longer context, and tool-use reliability; OpenAI's consumer base gives it reasons to keep investing in ChatGPT-facing features that don't always translate to API improvements.
• Multi-model strategies remain the safer bet. Revenue leadership can change quickly in this market, and teams that hard-wire a single provider into their stack take on switching-cost risk if pricing or capabilities shift.

None of this changes what model to pick for a given task today — that should still come down to benchmarks, latency, and cost for the specific workload. But it does explain why Anthropic has been able to fund aggressive coding-model releases and enterprise sales pushes at a pace that, a year or two ago, would have looked out of reach for a company its size.

AiiN's takeaway

The headline framing — 'Anthropic beats OpenAI' — is less interesting than the mechanism behind it. This isn't Claude out-competing ChatGPT for consumer mindshare; it's enterprise and developer spend outgrowing consumer subscription spend inside the AI market as a whole. In our estimation, that trend likely keeps favoring whichever lab ships the most reliable agentic and coding tools over the next year, regardless of which one currently leads on revenue. For builders, the practical takeaway isn't to switch providers because of a revenue headline — it's to keep watching where usage-based, developer-facing products are winning, because that's where the next round of capability investment will land first.