Summary
Anthropic announced a major compute expansion at its inaugural Code with Claude developer conference in San Francisco, anchored by a new infrastructure partnership with SpaceX. The deal gives Anthropic access to 300 megawatts of capacity and over 220,000 Nvidia GPUs — directly addressing the persistent outages and rate-limit complaints that frustrated Claude users throughout the prior quarter.
Nate Herk breaks down three concrete changes effective immediately: Claude Code’s 5-hour usage limits are doubled across Pro, Max, and Team plans; the peak-hours throttling that reduced limits during weekday mornings has been removed entirely; and API rate limits for Claude Opus models have been substantially increased across all tiers, with output tokens per minute jumping from 8,000 to 80,000 — a 10x increase that makes parallel agent architectures far more viable. Herk also contextualizes the broader infrastructure buildout, noting active compute agreements with Amazon, Google, Broadcom, Microsoft, Nvidia, Fluid Stack, Goldman Sachs JV, and Blackstone.
The video explains why compute procurement is a complex balancing act — unused capacity wastes money, while insufficient capacity degrades user experience and pushes developers toward competitors. For teams building production agents or high-throughput applications on the Claude API, the previous 8,000 output-token-per-minute ceiling was a hard constraint on scale; the new limits substantially change the calculus for those workloads.
📺 Source: Nate Herk | AI Automation · Published May 07, 2026
🏷️ Format: News Analysis
