When your AI company grows 80 times faster than planned, you need more than a bigger server rack. Anthropic just signed a deal with Elon Musk’s SpaceX to take over an entire data center — and it still might not be enough.
Dario Amodei didn’t mince words. “That is the reason we have had difficulties with compute,” the Anthropic CEO told an audience this week, explaining why his company had been struggling to keep up with demand for its Claude AI products. The reason, it turned out, was almost absurdly simple: Anthropic had planned for its revenue and usage to grow roughly tenfold in the first quarter of 2026. Instead, it grew eighty times over.
“Just crazy,” Amodei called it. “Too hard to handle.”
In the span of a few weeks, that staggering admission has given way to an equally staggering announcement: Anthropic has struck a deal with SpaceX to take over all of the available compute capacity at the company’s Colossus 1 data center in Memphis, Tennessee — a facility packed with more than 220,000 NVIDIA GPUs and capable of delivering over 300 megawatts of computing power. The agreement, confirmed by both companies on May 6, is one of the largest single-facility compute deals in the history of the AI industry. It also carries an almost novelistic irony: the fastest-growing AI company in the world is now renting its infrastructure from the company owned by the man suing its closest competitor.
From $9 Billion to $30 Billion in Four Months
To understand why a deal this large was necessary, you have to look at the numbers behind Anthropic’s growth — numbers that have stunned even the company’s own leadership.
At the end of 2025, Anthropic’s annualized revenue run rate stood at roughly $9 billion. By April 2026, that figure had surged past $30 billion, overtaking OpenAI’s $25 billion ARR and making Anthropic — for the first time — the leading AI company by revenue. The jump happened in under four months. It is, by almost any measure, one of the fastest revenue ascents in tech history.
The engine behind that growth is Claude, Anthropic’s family of AI models, and in particular the explosive adoption of Claude among enterprise customers. Today, more than 1,000 companies spend over $1 million per year on Claude — a figure that itself doubled in under two months, from 500 to more than 1,000, between Anthropic’s Series G funding announcement in February and early May. Roughly 80% of Anthropic’s revenue now comes from business customers rather than individual consumers, a composition that distinguishes it from rivals like OpenAI, whose revenue base skews more toward subscription products like ChatGPT Plus.
Fueling much of this enterprise boom is Claude Code, Anthropic’s AI coding assistant, which has seen particularly dramatic adoption among software development teams. The product has become a daily tool at thousands of companies, with developers using it to write, debug, and review code at a pace that has repeatedly strained Anthropic’s infrastructure.
A Compute Crisis — and a Surprising Partner
Explosive growth at this scale creates a very specific kind of problem: you need chips, and you need them now. AI model inference — the process of actually running a model in response to a user’s query — is extraordinarily compute-intensive, and the gap between demand and available capacity has real consequences. Users experience slower responses, rate limits, and in some cases outright service disruptions.
To close that gap, Anthropic has been aggressively building out its compute partnerships. It previously announced an agreement with Amazon Web Services for up to five gigawatts of capacity, with nearly one gigawatt expected to come online by the end of 2026. Google, which has committed up to $40 billion in investment to Anthropic, is also a major infrastructure partner. But with usage surging at eighty times the projected rate, even those arrangements proved insufficient.
Enter SpaceX.
The partnership, which was announced alongside an upgrade to usage limits for Claude Pro and Claude Max subscribers, gives Anthropic immediate access to the full compute capacity of Colossus 1 — a data center that SpaceX originally built to support its own AI initiatives under the xAI brand. The facility’s 220,000-plus NVIDIA GPUs represent a massive pool of inference capacity, and the 300-plus megawatts of power it draws is roughly equivalent to the electricity consumption of a small city.
The deal also comes with an intriguing forward-looking element: Anthropic has “expressed interest” in working with SpaceX to explore developing multiple gigawatts of AI compute capacity in space — a concept that would have seemed like science fiction just a few years ago but increasingly appears on the roadmaps of companies trying to build infrastructure unconstrained by terrestrial land, power, and cooling limitations.
The Revenue Race — and What It Means
The significance of Anthropic surpassing OpenAI in annualized revenue is hard to overstate. For most of the past several years, OpenAI has been the defining company of the AI era — the creator of ChatGPT, the most-used AI product in history, and the company most closely associated in the public mind with the AI revolution. For a rival to overtake it on revenue is not just a business milestone; it is a signal that the competitive landscape has fundamentally shifted.
Anthropic’s trajectory looks different from OpenAI’s in other ways too. While OpenAI is projected to spend $125 billion per year on model training by 2030, Anthropic’s equivalent projection for the same period is roughly $30 billion — a figure that suggests the company has found a way to develop frontier AI models at dramatically lower cost. That efficiency, combined with its enterprise-heavy revenue mix and the surging demand for Claude Code and Claude’s API, has given Anthropic a financial profile that investors and analysts are increasingly viewing as a more durable business model than consumer subscription products.
What Comes Next
Amodei’s public acknowledgment that his company’s growth is “too hard to handle” was, in part, a joke — a CEO’s wry admission that success can create its own problems. But it was also a window into the genuine operational challenge Anthropic faces. More compute means more capacity, but it also means more cost, more complexity, and more pressure to keep expanding. The SpaceX deal buys time. It does not solve the underlying equation.
For Claude users, the more immediate takeaway is straightforward: usage limits for Claude Pro and Max subscribers are going up, and the infrastructure driving those products is about to get significantly more powerful. For the industry, the story is larger — a reminder that in the current AI moment, the ability to acquire and deploy compute at scale may matter as much as the models themselves.
And for anyone watching the race between OpenAI, Anthropic, and Google, the scoreboard just changed. The company that built Claude — quietly, methodically, with a stated commitment to AI safety as its core mission — is now, by one important measure, the biggest AI company in the world.
Whether it can stay there may depend on how many more data centers it can find.
Sources: CNBC, Bloomberg, Axios, Inc., Al Jazeera, Anthropic
