Anthropic's 80x Growth Sends It Scrambling for SpaceX Compute as Musk Becomes AI Landlord

Anthropic CEO Dario Amodei told attendees at the company's San Francisco developer conference on Wednesday that his team planned for 10x growth this year. What they got was 80x — a surge so extreme it broke their infrastructure planning. Revenue and customer usage both jumped 80‑fold in Q1 2026 on an annualized basis, according to CNBC.

"That is the reason we have had difficulties with compute," Amodei said on stage. "We're working as quickly as possible to provide more" capacity, promising to "pass that compute on to you as soon as we can."

The company's annual revenue run rate has reportedly hit $4 billion, roughly quadruple where it stood at the start of 2025, according to Sherwood News. Amodei described the current trajectory as "just crazy" and "too hard to handle," adding he hoped for "more normal" expansion going forward. Anthropic is now reportedly in talks to raise capital at a $900 billion valuation — potentially surpassing OpenAI.

Three months ago, Elon Musk posted on X that Anthropic was "evil," "misanthropic," and "hated Western civilization." On Wednesday, he leased them one of his most valuable assets: the Colossus 1 supercomputer in Memphis, Tennessee.

The deal gives Anthropic access to 300 megawatts of compute capacity and more than 220,000 Nvidia GPUs — including H100, H200, and GB200 accelerators — all expected to come online within the month, The Wall Street Journal reported. Financial terms weren't disclosed, but Fortune analyst estimates peg the agreement at $3 billion to $4 billion annually, with more than $2.5 billion in profit — the data center is already built, so margins are unusually high.

Musk himself explained the pivot bluntly: "After that, I was ok leasing Colossus 1 to Anthropic, as SpaceXAI had already moved training to Colossus 2," he posted. He also claimed "no one at Anthropic set off my evil detector" after meeting senior team members, CNBC reported.

Analyst Antoine Chkaiban of New Street Research provided the financial assessment to Fortune.

Colossus 1 was originally built in 2024 to train xAI's Grok model. But Grok's annualized revenue reportedly sits below $1 billion — a fraction of what the infrastructure could theoretically support. Utilization at Colossus 1 had reportedly dropped to roughly 11% after xAI shifted training to Colossus 2, according to Forbes.

SpaceX is expected to begin its public roadshow next month, targeting a valuation between $1.75 trillion and $2 trillion. Having Anthropic as a marquee compute customer helps position SpaceX as a credible cloud‑infrastructure contender alongside AWS, Azure, and Google Cloud. "He who controls the data center, really does control the application of artificial intelligence right now," Andrew Moore, former head of Google Cloud AI, told Fortune. Moore added: "I think both sides of this wedding of convenience will be a little stressed out by it."

The deal includes an unusual escape clause: SpaceX "reserves the right to reclaim the compute" if Anthropic's AI "engages in actions that harm humanity."

The compute isn't just future capacity — Anthropic began using it almost immediately.

Within hours of the announcement, the company doubled five‑hour rate limits on Claude Code for Pro, Max, Team, and Enterprise seat‑based plans, Forbes reported. Peak‑hours reductions for Pro and Max users were also removed, and API rate limits for Claude Opus models were "considerably" increased.

The Wall Street Journal confirmed the deal will "substantially increase" Anthropic's overall capacity, with Anthropic specifically citing higher usage limits for Claude Code and the Claude API as direct outcomes.

Anthropic's existing compute commitments are enormous but mostly future‑dated. It has 5‑gigawatt agreements with both Amazon and Google, plus a $30 billion Azure partnership with Microsoft and Nvidia, according to Forbes. But most of that capacity won't arrive until late 2026 or 2027. As Forbes columnist Jon Markman put it: "Anthropic needs compute now, not in eighteen months."

This deal also signals a deeper structural shift. "In 2026 AI, compute scarcity is the constraint that overrides every other strategic preference, including who you have to do business with to solve it," Markman wrote. Anthropic and SpaceX also said they've "expressed interest in partnering to develop multiple gigawatts of orbital AI compute capacity," CNBC reported — a nod toward space‑based data centers, long a Musk priority.

The Colossus 1 facility hasn't been without controversy. While building and operating the data center, xAI reportedly ran dozens of natural gas‑burning turbines without federal permits, worsening air pollution in the Memphis area. The NAACP sued xAI in April 2026 for unlawful emissions, and community groups have staged persistent protests.

Anthropic, a company that has built its brand on AI safety and responsible development, is now indirectly tied to a facility with active civil rights litigation over environmental harm — adding another layer of tension to an already complicated partnership.

For builders using Claude, the immediate effect is concrete: fewer rate limits, higher API caps, and more reliable access during peak hours. The deal also signals that Anthropic is willing to make uncomfortable partnerships to solve compute constraints — something developers should expect to see more of as the AI infrastructure bottleneck tightens.

Longer‑term, the orbital compute discussion hints at an even more ambitious vision. If even a fraction of those planned gigawatts in space become operational by the late 2020s, it could fundamentally reshape where and how AI computation happens. But for now, the most important story is simpler: in the 2026 AI industry, even calling someone "evil" isn't a dealbreaker when GPUs are on the table.