For most of the generative-AI era, the scoreboard everyone watched was users. ChatGPT’s weekly active count, the app-store rankings, the viral screenshots — that was the story. In 2026, a quieter number has started to matter more. Anthropic, the maker of Claude, has reportedly moved ahead of OpenAI on annualized revenue, a milestone that says less about who has the flashiest chatbot and more about where the money actually is.
The important caveats first: these are reported and estimated figures, not audited results, and the specific metric matters enormously. What has reportedly crossed over is annualized revenue run-rate — a snapshot of the latest month or quarter multiplied out to a year — driven overwhelmingly by enterprise contracts and API usage. That is a genuinely different thing from saying Anthropic is “bigger than OpenAI.” It isn’t, on most measures that consumers would recognise. But as a signal about which business model is monetising fastest, it is worth reading carefully and neutrally.
The milestone
The clearest way to describe what happened: several outlets and analysts reported in the first half of 2026 that Anthropic’s annualized revenue run-rate had overtaken OpenAI’s. By some accounts the crossover occurred around April 2026, when Anthropic reached roughly a $30 billion annualized run-rate against OpenAI’s roughly $24 billion. Later estimates from the research firm Sacra put Anthropic near a $45–47 billion run-rate by May 2026, versus an OpenAI figure widely cited in the $25–33 billion range.
It is worth stressing that these numbers do not all agree, and they should not be treated as precise. The research group Epoch AI, modelling the two companies’ trajectories, framed the crossover as a projection — estimating a mid-to-late-2026 tipping point around $43 billion, with a wide confidence band running from early 2026 into 2027. Epoch is explicit that its fit relies on media reports “which may have timing or accuracy uncertainties.” In other words: the direction of travel is well established; the exact date and dollar figure are not.
What is not in dispute is the shape of the growth. Anthropic’s annualized revenue reportedly climbed from single-digit billions at the end of 2025 to tens of billions within months — one of the steepest revenue ramps the enterprise-software industry has seen. The milestone is measured in revenue, not in user counts, and that distinction is the whole point.
Why enterprise revenue differs
The reason Anthropic can lead on revenue while trailing badly on consumer mindshare comes down to how each company makes money. Anthropic was built enterprise-first. The bulk of its revenue reportedly comes from pay-per-token API calls and large business contracts rather than $20-a-month consumer subscriptions. Sacra’s breakdown puts something on the order of 70–80% of Anthropic’s revenue in the API and business bucket. The company has said it counts hundreds of enterprise customers each spending more than $1 million a year, with several of the largest companies in the world among them.
OpenAI’s mix is the mirror image, though it is shifting fast. ChatGPT gave OpenAI an enormous consumer base and brand, and consumer subscriptions have historically been the centre of gravity. But enterprise is now reportedly north of 40% of OpenAI’s revenue and closing on parity — so this is a story about two go-to-market strategies converging from opposite ends, not one company doing “enterprise” and the other ignoring it.
Why does the enterprise/API weighting matter for the revenue league table?
- Revenue density. A single large customer running production workloads through an API can generate more revenue than tens of thousands of consumer subscribers. A smaller user base can produce a larger top line.
- Durability. Business contracts and embedded API integrations tend to be stickier and more predictable than consumer subscriptions, which churn on novelty and price.
- Workload gravity. A large share of Anthropic’s growth has reportedly come from coding — Claude Code alone was cited at a multi-billion-dollar annualized run-rate within a year of launch. Developer and agentic workloads consume tokens continuously, not occasionally.
None of this makes one model “better.” Consumer scale builds brand, data and distribution; enterprise revenue builds a durable, higher-margin book of business. They are simply different bets, and in 2026 the enterprise bet is the one producing the bigger reported revenue number.
The caveats
This is where discipline matters, because a revenue-run-rate headline is easy to over-read.
Revenue is not profit. Both companies are burning enormous amounts of cash to train frontier models and serve inference at scale. OpenAI has reportedly projected losses in the region of $14 billion for 2026 and does not expect positive free cash flow for years; Anthropic’s burn is understood to be large as well, higher revenue notwithstanding. Leading on run-rate says nothing about who is closer to sustainable economics.
The figures are reported, not audited. Both are private companies. The numbers circulating come from media reports, analyst estimates, investor materials and selective company disclosures — not audited financial statements. There are also methodology wrinkles: revenue booked through cloud resellers such as AWS, Google Cloud and Microsoft can be reported on a gross basis, which can inflate a headline figure relative to a net-reporting peer. Comparing two run-rates built on different conventions is not apples-to-apples.
The race is close and fast-moving. The gap, on the estimates available, is measured in single-digit or low-double-digit billions between two companies growing at extraordinary rates. Both have signalled slower growth ahead. A crossover in one quarter does not lock in a lead; the ordering could plausibly flip again, and a run-rate snapshot is exactly that — a snapshot.
The honest summary: Anthropic has reportedly taken the lead on annualized enterprise- and API-driven revenue run-rate. That is a specific, meaningful claim — and it is not the same as “Anthropic has overtaken OpenAI,” full stop.
The India read
For founders, marketers and operators in India, the useful signal here is not the leaderboard — it is the reframing. The AI race is increasingly being scored on durable revenue, not consumer buzz, and that maps closely to how Indian businesses are actually adopting these tools.
Enterprise-first adoption is the norm here too. Indian IT services firms, banks, SaaS companies and fast-growing startups are wiring model APIs into support desks, code review, document processing and internal copilots — production workloads billed by usage, not viral consumer apps. That means the vendor questions that matter locally are the same ones this revenue story implies:
- Reliability and economics over hype. The right question is not “which model has the best demo,” but “which vendor gives us predictable latency, uptime, data-handling terms and a token price that survives contact with real volume.”
- Portability. With two credible frontier providers trading the lead, teams that keep their prompts, evaluations and integrations reasonably model-agnostic preserve leverage — and avoid betting a roadmap on whoever happens to be ahead this quarter.
- Read past the consumer scoreboard. If you are evaluating AI vendors, user counts and app-store rankings are the wrong lens. Enterprise revenue, API pricing trajectory and customer concentration tell you far more about whether a provider will still be a sensible partner in two years.
The Anthropic-versus-OpenAI revenue story, read neutrally, is less a coronation than a correction to how the AI race is scored. The flashiest chatbot and the fastest-growing revenue base need not be the same company — and for anyone deploying these models in production, the second number is the one worth watching.
