Anthropic Leases Musk's Colossus 1 In Compute Deal
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic struck a deal to lease full capacity of SpaceX's Colossus 1 data center in Memphis, with The Decoder reporting the company's 80x growth had outstripped its own infrastructure
- Elon Musk said on X he spent the prior week with senior Anthropic staff to understand "what they do to ensure Claude is good for humanity," adding that "no one set off my evil detector" — a sharp turn from his past public attacks on the company
- The lease terms include a clause reserving SpaceX's right to reclaim the compute "if their AI engages in actions that harm humanity," according to a clause highlighted by X user @deredleritt3r
- SpaceXAI had already migrated its own training to Colossus 2, freeing Colossus 1 capacity for Anthropic, Musk's post indicated
- Anthropic separately plans to spend roughly $200 billion on Google Cloud and chips over five years, representing more than 40% of the revenue backlog Google disclosed last week, per The Information
- Commentators including Derek Thompson framed the deal as a symptom of a broader AI compute crunch, with frontier labs "desperate for compute" even as Musk's own models lag behind Claude
- Instapundit flagged a separate filing showing Musk's Terafab chip factory in Texas could cost up to $119 billion
Why it matters: Anthropic is now committing tens of billions of dollars to multiple external infrastructure partners — SpaceX for capacity and Google for cloud and chips — underscoring how compute scarcity is forcing frontier labs into dependency on a shrinking pool of suppliers. The termination clause tied to "actions that harm humanity" gives Musk a defined lever over a direct competitor's training capacity, a structural advantage xAI could not win on model quality alone.
