Anthropic Doubles Claude Limits via SpaceX's 220K-GPU Colossus Deal
Anthropic has doubled token limits across all Claude paid plans and removed peak-hour rate caps for Pro and Max subscribers — live immediately as of May 19. The upgrade is enabled by a new compute deal with SpaceX giving Anthropic access to 220,000+ Nvidia GPUs via the Colossus cluster, the same infrastructure now backing Cursor's latest model training run.
What the Source Actually Says
AlphaSignal's May 19 newsletter, confirmed independently by @ClaudeAI on X, details the concrete changes: context windows go twice as deep without hitting a mid-session wall; peak-hour throttling on Pro and Max accounts is eliminated; and API Tier-1 subscribers see a 1,500% increase in input tokens per minute. AlphaSignal frames it plainly: "You know that feeling when Claude cuts you off mid-project? That just got a lot less common."
The compute backstory carries as much weight as the product change. The SpaceX Colossus cluster — 220,000+ Nvidia GPUs — is the same infrastructure backing Cursor's separate 1M H100-equivalent model training run announced the same day. Two major AI tooling upgrades unlocked by a single physical substrate in one news cycle makes the compute-as-moat thesis concrete rather than theoretical. AlphaSignal's editorial read: "SpaceX is quietly becoming the GPU landlord of the AI industry."
Strategic Take
Any practitioner running Claude in long-context agentic loops — multi-file code analysis, large-doc synthesis, extended Claude Code sessions — benefits immediately with zero config changes. The 1,500% Tier-1 API TPM uplift is the most actionable number for API builders: the prior cap was a real throttle on batch workloads. The xAI/SpaceX GPU-landlord dynamic is worth monitoring as a potential single-point-of-dependency risk across an increasingly Colossus-backed ecosystem.
