Anthropic Secures Massive SpaceX Colossus Compute Deal, Doubles Claude Code Limits

TLDR

Anthropic has signed a deal to take over the entire compute capacity of SpaceX’s Colossus 1 data center in Memphis, delivering more than 300 MW (over 220,000 NVIDIA GPUs) within the month. This new capacity, combined with existing partnerships, has allowed the company to double five-hour rate limits for Claude Code on paid plans, remove peak-hour reductions for Pro/Max users, and significantly raise API rate limits for Claude Opus models. The move underscores the escalating compute arms race among frontier AI labs.

Compute Partnership Details

The agreement gives Anthropic exclusive use of Colossus 1’s capacity:

• Over 300 megawatts of power.
• More than 220,000 NVIDIA GPUs (including H100, H200, and newer generations).
• Available within the month of the announcement.

This directly boosts capacity for Claude Pro and Claude Max subscribers and supports broader scaling.

It joins Anthropic’s other major infrastructure commitments:

• Up to 5 GW with Amazon (nearly 1 GW of Trainium/Graviton capacity by end of 2026).
• 5 GW agreement with Google and Broadcom (starting 2027).
• $30 billion Azure capacity partnership with Microsoft and NVIDIA.
• $50 billion investment in American AI infrastructure with Fluidstack.

Anthropic trains and runs Claude across AWS Trainium, Google TPUs, and NVIDIA GPUs, and is exploring additional opportunities, including potential orbital AI compute with SpaceX.

Higher Usage Limits

Effective immediately, the following changes improve experience for dedicated users:

• Claude Code five-hour rate limits doubled for Pro, Max, Team, and seat-based Enterprise plans.
• Peak hours limit reduction removed for Claude Code on Pro and Max accounts.
• API rate limits raised considerably for Claude Opus models (specific tiers updated in Anthropic’s documentation).

These adjustments address surging demand, particularly for agentic coding and professional workflows.

Broader Context and Commitments

The expansion supports international needs for regulated industries (finance, healthcare, government) with in-region infrastructure and data residency compliance. Capacity additions prioritize democratic countries with secure supply chains.

Anthropic has committed to covering any consumer electricity price increases caused by its US data centers and is exploring similar approaches internationally.

Customer example: The Government of Alberta has used Claude Code (Opus and Sonnet) since 2025 to review systems, find vulnerabilities, and apply fixes across government infrastructure.

Why this story matters

Frontier AI progress is increasingly gated by access to massive, reliable compute at scale. Anthropic’s deal with SpaceX highlights how hyperscalers, chipmakers, and even non-traditional players like SpaceX are becoming critical infrastructure providers. The immediate payoff—higher limits for users—shows how infrastructure wins translate directly to product experience, while the multi-partner strategy (AWS, Google, Azure, now SpaceX) reduces single-point risks in an era of extreme demand.

Sources

• Anthropic: “Higher usage limits for Claude and a compute deal with SpaceX” (published May 6, 2026). https://www.anthropic.com/news/higher-limits-spacex
• Cross-referenced coverage from Axios, Engadget, The Next Web, and LinkedIn announcements confirming capacity figures, limit changes, and context from prior deals.
• Background on Colossus 1 and related partnerships from contemporaneous reporting.

Featured Image Alt Text

Abstract visualization of massive data center racks with NVIDIA GPUs powering Claude models, overlaid with SpaceX rocket elements and usage limit graphs showing doubling of capacity

Tags

Compute, Infrastructure, Anthropic, SpaceX, Colossus, Claude Limits, GPUs, AI Scaling, Data Centers, Partnerships