Anthropic Rents the Entire Colossus 1 Data Center — 300+ MW and 220,000 NVIDIA GPUs From SpaceX
After reporting 80x year-over-year usage growth, Anthropic signed a deal to lease all of SpaceX's Colossus 1 facility in Memphis — over 220,000 NVIDIA GPUs across 300+ MW — and is immediately lifting rate caps for most paid subscribers.
Anthropic announced on May 6 that it has leased the entirety of Colossus 1 — SpaceX’s Memphis, Tennessee data center — giving it access to over 220,000 NVIDIA GPUs across more than 300 megawatts of capacity. Compute will be online within the month.
The deal is a direct consequence of the company’s growth numbers. Anthropic reported 80x year-over-year revenue and usage growth in Q1 2026. That kind of trajectory doesn’t give you time to build your own data centers; you acquire whatever large-scale capacity exists and is available now. Colossus 1 fits that description — it was purpose-built for massive AI workloads and is one of the largest single-site GPU clusters in existence.
The Musk dimension is notable. SpaceX and xAI merged earlier in 2026, meaning Elon Musk’s companies now own the facility Anthropic is renting. Musk commented publicly on the deal: “Everyone I met was highly competent and cared a great deal about doing the right thing. No one set off my evil detector.” That’s a warmer endorsement than Musk typically gives to AI labs he doesn’t control, and it signals that even across competitive lines, raw compute economics are dictating partnerships.
The user-facing changes start now. As a direct consequence of the capacity expansion, Anthropic is lifting the five-hour rate caps that have been throttling usage for Pro, Max, Team, and Enterprise subscribers. Peak-hour usage reductions are also being removed. If you’ve been hitting Claude’s limits during heavy workdays, that constraint is going away.
The longer-term ambition is larger. The announcement includes Anthropic expressing interest in working with SpaceX to develop “multiple gigawatts” of compute capacity in space. Whether orbital or Starship-adjacent data infrastructure becomes real within a reasonable timeframe is speculative — but the statement indicates Anthropic is thinking about compute supply at a scale that dwarfs current cloud infrastructure, not just filling a near-term gap.
For context: 300 MW is roughly the power draw of a mid-sized city. The H100 cluster at Colossus was already considered one of xAI’s central competitive assets. Anthropic now has access to all of it.
The infrastructure gap between AI frontier labs and everyone else has been widening steadily. This deal accelerates that dynamic: a company that had 80x growth in a single quarter is now operating from what is effectively one of the largest AI compute sites on the planet. The race is no longer just about model quality — it’s about who can secure the physical infrastructure to serve hundreds of millions of users simultaneously without degrading quality.
Competitors are watching closely. Any lab without a comparable secured-compute arrangement is now structurally disadvantaged for the next 12-24 months.