Anthropic boosts Claude limits, signs SpaceX deal

This is Anthropic turning compute into a user-facing product feature: more GPUs, fewer caps, better throughput for heavy Claude users. It also shows frontier AI is now as much an infrastructure race as a model race.

• –The biggest immediate win is for power users who were hitting Claude Code and Opus limits in real workflows
• –Anthropic is signaling it will buy capacity wherever it can get it, not just from its existing cloud partners
• –The orbital compute angle is mostly long-term theater; the real story is terrestrial capacity and demand management
• –For developers, higher limits should make long-running sessions, batch jobs, and agentic workflows less brittle
• –This is another data point that model quality alone is not enough; distribution increasingly depends on reliable compute access