From $9 Billion to $30 Billion in Four Months
Anthropic disclosed on April 6 that its annualized revenue run rate has surpassed $30 billion, more than tripling the approximately $9 billion figure the company reported at the close of 2025. The growth has been driven by accelerating enterprise demand for its Claude AI models, which are now deployed across industries ranging from financial services to software development. When Anthropic closed its $30 billion Series G funding round in February at a $380 billion post-money valuation, the company reported that more than 500 business customers were each spending over $1 million per year. That figure has now exceeded 1,000, effectively doubling in less than two months.
To put the pace in context, OpenAI — long considered the industry's commercial leader — disclosed in late March that it was generating roughly $2 billion per month, or about $24 billion on an annualized basis. Anthropic's $30 billion run rate now exceeds that figure, though OpenAI maintains a vastly larger consumer user base with 900 million weekly active users and 50 million paid subscribers.
"We are making our most significant compute commitment to date to keep pace with our unprecedented growth."
A 3.5-Gigawatt Chip Deal Built on Google TPUs
Alongside the revenue disclosure, Anthropic announced what it called its largest infrastructure commitment to date — an agreement with Google and Broadcom to secure approximately 3.5 gigawatts of next-generation tensor processing unit capacity, expected to come online starting in 2027. The deal builds on the 1 gigawatt of TPU compute that Broadcom CEO Hock Tan confirmed the company is already supplying to Anthropic in 2026. The relationship between the three companies has been escalating in public view since September 2025, when Tan disclosed during an earnings call that a mystery customer had placed a $10 billion order for custom TPU racks. By December, he confirmed the customer was Anthropic and noted an additional $11 billion order had followed.
A Broadcom securities filing accompanying the announcement included a notable caveat: the consumption of this expanded compute capacity is dependent on Anthropic's continued commercial success, and the parties are in discussions with operational and financial partners to support the deployment. Mizuho analysts estimated that Broadcom's AI revenue from Anthropic alone could reach $21 billion in 2026 and $42 billion in 2027. Broadcom shares gained as much as 3.6 percent in after-hours trading following the filing.
A gigawatt is roughly the output of a large nuclear power plant. Anthropic is now committing to consume 3.5 gigawatts of compute capacity just for its TPU infrastructure — not counting its separate AWS Trainium clusters and NVIDIA GPU allocations. For comparison, the entire city of San Francisco draws about 1 gigawatt at peak demand. The sheer energy requirement underscores how AI infrastructure is becoming one of the largest single categories of electricity consumption on the planet, rivaling the power needs of small nations.
The Multi-Chip Playbook
One of the distinguishing features of Anthropic's infrastructure strategy is its deliberate diversification across chip platforms. Claude is trained and served on three hardware families: Amazon's custom Trainium processors, Google's TPUs, and NVIDIA's GPUs. Amazon Web Services remains the primary cloud provider and training partner, with the companies continuing to develop Project Rainier — a massive compute cluster purpose-built for Anthropic. Meanwhile, the new Google-Broadcom agreement deepens a parallel supply chain. The result is that Claude is now the only frontier AI model available on all three of the world's largest cloud platforms: AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry.
The multi-vendor approach gives Anthropic both resilience and leverage. If supply constraints hit one chip family, workloads can shift to another. It also allows the company to match specific tasks to the silicon best suited for them — training on one architecture, inference on another — creating efficiency advantages that single-platform competitors may lack.
The Hottest Trade in Private Markets
The revenue disclosure arrived against a backdrop of surging investor demand for Anthropic equity. Glen Anderson, president of the investment bank Rainmaker Securities, described Anthropic as the single hardest stock to source in his marketplace, noting that prospective buyers have billions of dollars lined up with virtually no sellers willing to part with their shares. Bloomberg reported earlier in the week that buyers had indicated to one firm that they had $2 billion in cash ready to deploy into Anthropic, even as roughly $600 million worth of OpenAI shares sat unsold.
Anderson attributed part of the frenzy to an unlikely catalyst: Anthropic's public standoff with the Department of Defense, which classified the company as a supply-chain risk after a dispute over AI safety guardrails. Rather than damaging the brand, the confrontation appeared to galvanize public support. Consumer usage of Claude surged, and investors began to view Anthropic as meaningfully differentiated from OpenAI. Goldman Sachs is now charging its customary 15 to 20 percent carry on Anthropic positions — a fee structure it does not impose on OpenAI shares — suggesting the bank sees higher expected returns in Anthropic paper.
"The hardest stock to source in our marketplace is Anthropic. There's just no sellers."
Growing Pains at the Subscription Layer
Not all of the weekend's Anthropic news was celebratory. On Friday, the company blocked Claude subscriptions from being used to power third-party autonomous agent tools such as OpenClaw, an open-source framework that allowed users to run AI agents continuously. Going forward, those users must pay through the API or a new pay-as-you-go extra-usage system rather than relying on flat-rate subscription plans. The move sparked immediate backlash in the developer community, with one prominent product manager characterizing it as the end of an all-you-can-eat buffet.
Anthropic framed the decision as a capacity management issue. Boris Cherny, the head of Claude Code, explained that the company's subscriptions were never designed for the sustained, high-compute usage patterns that agent harnesses demand. The company had already introduced five-hour session caps during peak periods, but heavy agent usage was reducing availability for everyday users. The tension points to a broader industry challenge: agentic AI — systems that run autonomously for hours, taking actions across applications — is far more expensive to serve than conversational chatbots, and someone has to absorb the cost.
Anthropic continues to navigate fallout from its designation as a "supply-chain risk" by the Department of Defense. The company has warned that the label could cost it billions in lost enterprise revenue from customers who do business with the U.S. government. Despite this, the commercial trajectory has only accelerated — the tripling of run-rate revenue occurred almost entirely after the dispute became public. Anthropic's legal challenge to the designation remains ongoing, and the confrontation has become one of the most closely watched regulatory disputes in the AI industry.
What This Means for the AI Industry
The combined force of these announcements — a revenue trajectory that now leads the industry, a multi-gigawatt infrastructure deal with two of the world's most important chip companies, and a secondary-market valuation that has investors lining up with cash in hand — marks a genuine inflection point for Anthropic and for the competitive landscape of AI at large. The company's willingness to commit to 3.5 gigawatts of compute in a single agreement reflects a bet that demand for frontier AI models will continue to accelerate at a pace that would have been considered implausible even a year ago.
Both Anthropic and OpenAI are reportedly exploring initial public offerings this year, but SpaceX's decision to file first may complicate the timing for both. Rainmaker's Anderson warned that SpaceX could absorb significant IPO liquidity, leaving less capital available for whichever AI company follows. For now, though, the numbers speak for themselves: in four months, Anthropic went from $9 billion to $30 billion — and it is building the infrastructure to sustain the next order of magnitude.