SpaceX reportedly rented out Colossus 1 after latency problems

SpaceX reportedly rented out capacity at its Colossus 1 AI data center after xAI encountered problems connecting it with two other Memphis-area campuses. Bloomberg, summarized by The Verge, attributes the trouble to hardware variation, aging network infrastructure, and latency between sites more than 10 miles apart. The report says SpaceX then struck large capacity deals with Anthropic and Google. The story complicates the idea that simply assembling more accelerators guarantees useful training scale: networking, hardware consistency, and geography can determine whether a cluster works as one system. It also shows SpaceX turning an internal infrastructure problem into a major external compute business.

Key details: June 13, 2026 report, Colossus campuses are more than 10 miles apart, Hardware variation and network lag reportedly caused problems, Capacity reportedly rented to Anthropic and Google.

Continue swiping for more AI Brief stories.