Anthropic rented SpaceX's Colossus 1 supercomputer cluster — 220,000 NVIDIA GPUs and 300 megawatts of power — to address an 80x annualized demand surge driven largely by Claude Code adoption. The immediate developer impact: five-hour rate limits for Claude Code doubled across Pro, Max, Team, and Enterprise plans, and peak-hour throttling was removed. The deal is notable given Elon Musk had publicly called Anthropic 'evil' just three months prior, but both parties had clear financial incentives — Colossus 1 was underutilized after xAI migrated to the newer Colossus 2 cluster, and Anthropic needed capacity in weeks, not years. The post also covers the broader AI compute race, Anthropic's multi-partner infrastructure stack (Amazon, Google, SpaceX), and a contractual clause allowing SpaceX to reclaim compute if Anthropic's AI 'harms humanity.'
Table of contents
What Colossus 1 Actually IsWhy Anthropic Needed ThisThe Musk ReversalWhat Actually Changed for Claude UsersThe Compute Race Is Now a First-Class Business ProblemThe Weird Politics at the Edge of This DealWhat This Means for Developers Using ClaudeThe Short VersionSort: