Anthropic is generating $19 billion in annualized revenue. Bloomberg’s Mark Gurman reported that Apple, despite partnering with Google for Siri, internally “runs its business on Anthropic” for product development. Those are two significant data points that say a lot about where enterprise AI spending is actually going.
The Revenue Trajectory
The growth rate here is the more interesting story than the number itself. Anthropic closed 2024 at roughly $1 billion ARR. By end of 2025, that had grown to around $9-10 billion. One month before this report, at the close of their $380 billion valuation funding round, it sat at $14 billion. Now it’s at $19 billion, a 20x increase in approximately 15 months. Bank of America pegged the surge at 110%.
A major driver of that acceleration is Claude Code. It now accounts for roughly 4% of all public GitHub commits, with projections putting it at 20% by end of 2026. Anthropic’s own engineering team uses it for 70-90% of their code, and the head of product reportedly uses it for 100% of theirs. That internal adoption pattern is worth noting given what Apple is doing.
What Apple Is Actually Doing With Claude
The public-facing story is that Apple uses Google Gemini for Siri. That deal is real. But Gurman’s reporting makes clear that Claude is what Apple’s internal teams are working with day to day. Apple hosts Claude instances on its own servers for tasks like prototype testing, code generation, design reviews, UI copy, and security automation. They explored a public Claude integration for Siri but talks collapsed in mid-2025, reportedly over pricing demands in the range of several billion dollars per year with escalators built in.
The self-hosted approach is consistent with how Apple operates. Proprietary code doesn’t leave Apple’s infrastructure, and Claude was specifically chosen for complex reasoning, safety, and controllability in those internal workflows. Anthropic also launched a Claude integration with Xcode for AI‑assisted coding, which fits the pattern of how Apple engineers are using it.
This is a good illustration of how enterprise AI adoption often works in practice. The public partnership and the internal toolchain are two different decisions made by different teams for different reasons. Apple’s deal with Google covers the consumer‑facing product. Claude covers the work that builds that product.
Enterprise Concentration vs. Consumer Volume
The business model comparison between Anthropic and OpenAI is worth understanding clearly. Anthropic captures roughly 5% of ChatGPT’s consumer user base but generates around 40% of OpenAI’s revenue. The per‑user economics reflect that gap: Anthropic’s enterprise customers average around $211 per month versus OpenAI’s consumer tier at roughly $25 per week. Eight of the ten largest Fortune 10 companies are Anthropic customers, and over 500 companies are spending more than a million dollars a year on the platform.
That structure means Anthropic’s revenue is mostly predictable enterprise contract spend rather than consumer subscription churn. The trade‑off is limited upside from consumer‑scale distribution, but the stability is real. OpenAI is building toward profitability with a much larger and more expensive consumer base. Anthropic is doing it with a smaller, higher‑margin enterprise book. Neither approach is wrong, but they are genuinely different bets on how AI businesses should be built.
For more context on how AI model costs and pricing structures have shifted recently, the post on cost creep in 2026 covers some of that ground.
What Drives the Stability Perception
Anthropic’s reputation as the more stable AI business isn’t only about revenue mix. It also comes from the way their model lineup has held together. Claude Opus 4.5 remains one of the stronger options for deep research and long‑document reasoning, and Claude Sonnet 4.5 covers most production workloads efficiently. The enterprise tooling built around these models, particularly Claude Code, has become embedded in how large engineering teams work.
Stability also benefits from not being in the middle of a restructuring. OpenAI has had significant internal changes and public drama over the past year. Anthropic has been quiet by comparison, and that quietness carries weight with enterprise procurement teams making multi‑year commitments. A Pentagon‑related controversy that briefly hit Anthropic earlier this year ended up boosting consumer adoption, with Claude reaching the top spot in the Apple App Store free apps chart and adding 1.5 million sign‑ups during that period. That is an unusual outcome from what should have been a PR problem.
The $19B number will attract skepticism, and some of that skepticism is warranted given how quickly this has moved. But the underlying mechanics of Anthropic’s business, the enterprise concentration, the Claude Code adoption curve, and the self‑hosted deployments at companies like Apple, are not speculative. Those are the actual drivers of what you’re seeing in the revenue line. The number reflects a real shift in where large organizations are placing their AI spend, and Anthropic has positioned itself to capture the portion of that market that prioritizes reliability and controllability over consumer reach.