Anthropic's annualized revenue run rate has surpassed $30 billion, up from roughly $9 billion at the close of 2025, according to an announcement posted to the company's website on April 7. To trace the arc: $4.5 billion in mid-2025, $9 billion at the end of that year, $14 billion in February, and $30 billion now. That trajectory is a hockey stick.
The immediate comparison is OpenAI, which reported approximately $24 billion in annualized run-rate revenue as of late March 2026, based on the company's stated monthly revenue of $2 billion, according to Sherwood News. By this measure, Anthropic has passed OpenAI.
But the more revealing question is why. The wire framed this as a revenue milestone. The driver is Claude Code, Anthropic's agentic coding platform. According to Officechai, which has been tracking Anthropic's growth trajectory, Claude Code is generating over $2.5 billion in run-rate revenue as of February 2026, with weekly active users doubling since January 1. That single product accounts for a material share of the $21 billion jump from $9 billion to $30 billion in roughly three months. The growth story is not primarily about frontier model API pricing — it is about a specific product finding a specific audience at scale.
The customer count reinforces this. Anthropic had more than 500 businesses spending at least $1 million annually as of February. That figure now exceeds 1,000, meaning the number of seven-figure customers doubled in under 60 days.
Claude Code lets developers delegate coding tasks to Claude agents — writing, reviewing, refactoring, and testing, run in loops without human intervention per step. It is a workflow product, not a model product. The revenue comes from seat subscriptions and usage, not from frontier model API calls. That distinction matters: Claude Code's growth signals that AI is being embedded into software development workflows at scale, not just being evaluated in proof-of-concepts.
Anthropic's compute infrastructure is the other half of the story. A Broadcom SEC filing reviewed by TechCrunch shows the Anthropic-Google-Broadcom deal includes 3.5 gigawatts of TPU-based AI compute capacity, an expansion of the October 2025 agreement for more than 1 gigawatt. The new agreement covers multiple gigawatts of next-generation TPU capacity expected to come online starting in 2027. At full buildout, that is enough power for a sustained, large-scale inference operation. Anthropic is not planning for the next model release. It is planning for inference at industrial scale.
The multi-cloud position is harder to replicate. Claude is the only frontier AI model available to customers on all three of the world's largest cloud platforms: Amazon Web Services (Bedrock), Google Cloud (Vertex AI), and Microsoft Azure (Foundry). OpenAI's models are on AWS and Azure but not on Google's cloud — a gap that reflects the fractured competitive relationships between the major labs. Anthropic's presence across all three means it captures enterprise spending regardless of which cloud a company has standardized on. In a market where cloud provider lock-in is a persistent concern for CIOs, being the option that sidesteps that conversation has concrete value.
The share trajectory deserves scrutiny. Anthropic's share of business AI spending relative to OpenAI grew from roughly 10 percent at the start of 2025 to over 65 percent by February 2026, according to Officechai. Directionally consistent with every other data point in this story — but the comparison is Anthropic versus OpenAI alone, not the combined OpenAI-Microsoft enterprise revenue, which would tell a different story. The frame matters.
OpenAI's $24 billion run rate is real. Anthropic's $30 billion is real. Together they suggest the enterprise AI market is large enough for two companies to grow at this rate simultaneously, which points to a total addressable market that outpaces most projections from two years ago.
What comes next is less clear. Anthropic's 2027 compute buildout assumes continued demand growth — 3.5 gigawatts is a multi-year commitment. If Claude Code growth slows, or if a competitor releases a compelling agentic coding product, the infrastructure bet outpaces the revenue. The hockey stick is real. The question is whether the slope holds.
