Anthropic Rents All of SpaceX’s Colossus 1 Cluster, Eyes Orbital Compute
Anthropic and SpaceX announced a compute partnership on May 6, 2026 that gives Anthropic access to all 300+ megawatts of capacity at xAI’s Colossus 1 data center in Memphis, Tennessee — a Musk-owned facility built around more than 220,000 NVIDIA GPUs. The deal lands within a month of signing, lifts usage limits for Claude Pro and Max subscribers immediately, and contains an unusual side-clause: Anthropic and SpaceX will explore developing multiple gigawatts of orbital, space-based AI compute.
General Audience
What’s in the Deal
SpaceX (which absorbed xAI earlier this year) is renting Anthropic the entire output of Colossus 1, the Memphis supercluster xAI assembled at record speed. The cluster mixes dense H100, H200, and next-generation GB200 deployments and is described by xAI as “one of the world’s largest and fastest-deployed AI supercomputers.” Anthropic will route the new capacity directly to its consumer products, easing the rate-limit complaints that have followed the popularity of Claude Code.
Concretely, Anthropic announced three immediate user-facing changes alongside the deal:
- Doubled five-hour Claude Code limits across Pro, Max, Team, and seat-based Enterprise plans
- Removal of peak-hours throttling for Pro and Max accounts
- Substantially higher API rate limits for Claude Opus models
An Unlikely Pairing
The partnership is striking because Elon Musk has been a vocal Anthropic critic and is currently litigating against another competitor, OpenAI. Musk merged xAI into SpaceX earlier in 2026, and on the day of the announcement he posted on X that he had recently spent time with senior Anthropic leaders and “no one set off my evil detector.” Anthropic, for its part, gets a marquee compute supplier; SpaceX gets a flagship customer to point at as it courts public-market investors.
The Orbital Wildcard
The most forward-looking piece of the agreement is a stated interest in building “multiple gigawatts of orbital AI compute capacity.” xAI’s announcement frames the rationale plainly: terrestrial power, land, and cooling cannot keep pace with frontier-model demand on the timelines that matter, and SpaceX’s launch cadence and constellation experience make space-based compute “a near-term engineering program rather than a research concept.” The companies stop short of committing to a launch schedule, but the language is significantly more concrete than the speculative space-data-center pitches that have circulated in industry decks for the past year.
Where It Fits in Anthropic’s Buildout
The Colossus deal joins a rapidly growing list of Anthropic compute commitments: up to 5 GW with Amazon (with nearly 1 GW of new capacity by the end of 2026), 5 GW with Google and Broadcom starting in 2027, a $30 billion Azure partnership with Microsoft and NVIDIA, and a $50 billion American AI infrastructure investment with Fluidstack. Anthropic also reiterated a commitment to cover consumer electricity price increases caused by its US data centers and said it is exploring similar policies internationally.
Read against the broader market, the move signals that capacity, not model quality, is the binding constraint of this stretch of the AI race — enough to push competitors who publicly dislike each other into a same-roof arrangement.
Related Coverage
- Anthropic Plugs Claude Into Adobe, Blender, and Ableton With Nine New Connectors — Anthropic’s recent push to embed Claude into creative software
- Anthropic Releases Claude Opus 4.7 With Sharper Coding and 3x Vision Resolution — the model family that will benefit from the new compute
- Anthropic’s 2026 Agentic Coding Trends Report: From Assistants to Agent Teams — context on the compute demands behind Claude Code
沪公网安备31011502017015号