Anthropic Leases xAI's Colossus 1 Data Center

On May 6, 2026, Anthropic announced it had leased the full capacity of xAI's Colossus 1 data center — over 300 MW and 220,000+ NVIDIA GPUs — to address compute constraints driven by explosive demand growth, with Claude Code rate limits doubled simultaneously as a customer benefit.[1] The deal pairs Anthropic with infrastructure tied to Elon Musk, who reportedly holds unilateral power to reclaim the compute if he judges Anthropic's AI to 'harm humanity.'[3] A concrete regulatory development has now emerged: a state permit hearing has been scheduled for xAI's gas turbines in Southaven, and the EPA has updated its rules on temporary generator exemptions — the first administrative movement on the environmental compliance questions the deal raised.[4] The orbital compute dimension of the announcement continues to attract parallel industry activity, with multiple startups and proposals now pursuing space-based AI inference.[7][5]

The Musk reclaim clause remains the sharpest unresolved governance risk: Anthropic's compute access is contingent on one private actor's subjective harm judgment, in direct tension with its public 'secure supply chains' framing.[1][3] The emerging regulatory process around xAI's turbines — now moving to formal permitting hearings — means the environmental liability embedded in this deal is becoming harder to ignore, both for regulators and for Anthropic's reputational positioning.

On May 6, 2026, Anthropic announced it had signed an agreement with SpaceX to access the full capacity of the Colossus 1 data center — over 300 megawatts and more than 220,000 NVIDIA GPUs — with availability expected within the month.[1] The announcement framed the deal as a direct customer benefit: Claude Code's five-hour rate limits were simultaneously doubled for Pro, Max, Team, and Enterprise plans, and peak-hour restrictions were removed for Pro and Max subscribers.[1] Anthropic also disclosed that it has expressed interest in partnering with SpaceX to develop multiple gigawatts of orbital AI compute capacity, a moonshot ambition signaling how aggressively the company is expanding its infrastructure vision.[1] The official announcement emphasized a broader multi-provider compute portfolio totaling over 10 GW across Amazon, Google/Broadcom, Microsoft/NVIDIA, and Fluidstack, and stressed a commitment to partnering only with democratic countries whose supply chains are secure.[1]

The financial context behind the urgency is significant. Anthropic's ARR grew from $9B to $44B in 2026, gross margins on inference infrastructure rose from 38% to over 70%, and the company is reportedly weighing a funding round at a valuation above $900B.[2] Zvi Mowshowitz characterized the deal as a rational response to 80x demand growth and framed xAI's role as a structural shift: having retained the larger Colossus 2 for its own model training, xAI appears to be pivoting away from frontier model competition and toward functioning as a compute provider.[2]

The sharpest public scrutiny came from Simon Willison, who raised two pointed concerns. First, Colossus 1 carries a documented environmental record: its gas turbines were operated without Clean Air Act permits or pollution controls by classifying them as 'temporary,' a history that sits uneasily with Anthropic's public safety identity and with the already politically charged debate over AI energy consumption.[3] Second, and more structurally alarming, the agreement reportedly includes a clause giving Elon Musk unilateral power to reclaim the compute if Anthropic's AI 'harms humanity' — with Musk himself as sole arbiter of that determination.[3] Willison frames this as 'a new form of supply chain risk,' reinforced by xAI's history of giving developers only two weeks' notice before deprecating models.[3]

A concrete regulatory development has since emerged: a state permit hearing has been scheduled for xAI's gas turbines in Southaven (the Mississippi portion of the greater Memphis area), and the EPA has separately updated its rules governing temporary generator exemptions.[4] This is the first formal administrative movement on the environmental compliance questions the deal raised, and its outcome may affect operating costs or capacity constraints at Colossus 1. Separately, the orbital compute ambition Anthropic gestured at has drawn parallel industry activity: multiple startups and proposals — including space-based inference ventures and lunar factory concepts for AI satellites — are now pursuing the same territory, suggesting the idea has moved from science fiction to active competition, even as technical feasibility and timelines remain undefined.[5][6][7]