Anthropic's SpaceX Deal Doubles Claude Code Limits and 16x Opus API Throughput

TL;DR: SpaceX's Colossus 1 data center (220k GPUs) gives Anthropic 300MW of new capacity — Claude Code's 5-hour limits are now doubled and Tier 1 Opus API throughput jumped 16x overnight.

Anthropic signed an agreement to access all compute capacity at SpaceX's Colossus 1 data center — 300+ megawatts and over 220,000 NVIDIA GPUs — with capacity coming online within the month.

The limit changes are live:

Claude Code (all paid plans):

• 5-hour rate limits doubled across Pro, Max, Team, and seat-based Enterprise
• Peak-hours limit reduction removed for Pro and Max accounts

Claude API — Opus models:

• Tier 1: 30,000 → 500,000 tokens/minute (16x increase)
• Higher tiers: proportional boosts

This joins a stack of compute commitments: 5 GW from Amazon (1 GW online end-of-2026), 5 GW from Google + Broadcom (2027), and $30B in Azure capacity via Microsoft + NVIDIA.

Why it matters: For automated pipelines, the Opus API jump is the most actionable change — tasks that previously hit Tier 1 ceilings on complex code review or long-context reasoning can now run at full speed on Opus without upgrading your tier. Doubled Claude Code limits mean background agents can run more work per rolling window.

Sources: Anthropic announcement