What happens when one of the world’s most advanced AI companies runs out of compute? It calls SpaceX. On May 6, 2026, Anthropic officially announced a landmark compute partnership with SpaceX, gaining access to Colossus 1 — a data center housing over 220,000 NVIDIA GPUs and more than 300 megawatts of capacity. The immediate result: Claude Code usage limits have been doubled for all paid subscribers, and frustrating peak-hour throttling is gone.
If you’ve ever hit a usage wall mid-session while working with Claude Code, or watched your API throughput crawl during busy hours, this announcement is directly aimed at solving your pain points. Let’s break down everything that changed, why it matters, and what you should expect next.
The Anthropic–SpaceX Colossus Deal: What Happened?
Anthropic signed a compute agreement with SpaceX to utilize the full capacity of Colossus 1, SpaceX’s purpose-built AI infrastructure data center. This isn’t a minor infrastructure tweak — it’s a strategic move that gives Anthropic access to one of the largest GPU clusters on the planet within weeks of signing.
Colossus 1 was originally built to support large-scale model training, but Anthropic is leveraging its inference capacity to directly benefit end users. The 300+ megawatts of power capacity translates into a dramatic improvement in Anthropic’s ability to serve high-demand workloads — especially Claude Code, which requires sustained, intensive compute sessions.
This deal is part of a broader Anthropic compute expansion strategy that also includes a 5GW agreement with Amazon, a 5GW partnership with Google and Broadcom (coming online in 2027), a strategic tie-up with Microsoft and NVIDIA worth $30 billion in Azure capacity, and a $50 billion investment in American AI infrastructure with Fluidstack.
What Exactly Changed: Claude Code Limit Increases Explained
Here’s a precise breakdown of what’s new for Claude Code subscribers as of May 6, 2026:
Five-Hour Rate Limit — Doubled
Claude Code’s five-hour rolling usage limit has been doubled across all paid tiers: Pro, Max, Team, and seat-based Enterprise plans. This means developers and power users can now sustain longer uninterrupted coding sessions — working through complex multi-file refactors, full-stack builds, or extended agentic workflows without hitting a wall.
Peak-Hour Throttling — Removed
Previously, Claude Pro and Max users experienced significant rate reductions during peak usage windows. These restrictions have been fully removed. You now get consistent performance regardless of what time of day you’re working — whether it’s 9 AM Monday or midnight Friday.
Claude Opus API — Massively Expanded Throughput
For API users on Claude Opus models, the numbers are striking. Maximum input tokens per minute for Tier 1 users jumped from 30,000 to 500,000 — a 16x increase. Maximum output tokens per minute rose from 8,000 to 80,000 — a 10x increase. Enterprise and production-scale applications that previously required careful batching and rate management can now operate much more freely.
Why Claude Code Has Been the Bottleneck
Claude Code is Anthropic’s agentic coding environment — it doesn’t just answer coding questions, it executes multi-step development tasks autonomously. That level of capability requires significantly more sustained compute than a standard chat session. A single Claude Code session can involve dozens of tool calls, file reads, test executions, and iterative refinements within minutes.
This is computationally expensive. Anthropic’s previous infrastructure, while robust, struggled to scale with the explosive adoption of Claude Code following its widespread release. Users on paid plans were increasingly vocal about hitting limits during complex projects — a friction point that directly affected developer productivity and enterprise adoption.
The SpaceX Colossus partnership resolves this at the infrastructure level rather than through artificial throttle adjustments. It’s a supply-side fix, not a patch.
How This Compares to OpenAI and Google’s Compute Strategies
Anthropic’s aggressive compute expansion mirrors moves by its major competitors. OpenAI updated ChatGPT’s default model to GPT-5.5 Instant in early May 2026, prioritizing speed and availability for all users. Google is preparing to unveil Gemini 4 at Google I/O 2026 (May 19), a flagship model expected to feature deeper multimodal reasoning and cross-service integration.
The compute arms race is real — and Anthropic is making it clear that scaling infrastructure is now as important as scaling model capability. The company that can serve the best model reliably at scale wins the enterprise. Doubling Claude Code limits is a direct competitive response to OpenAI’s GPT-5.5 rollout and Google’s impending I/O announcements.
What This Means for Users
Individual developers on Claude Pro or Max: You can now run longer uninterrupted coding sessions. Extended refactors, test-driven development loops, and multi-file architectural changes are all significantly more feasible within a single session window.
Teams on the Team plan: Concurrent usage across team members during peak hours — say, during a sprint or crunch period — will no longer cause degraded experience for the whole group. The doubled five-hour limit effectively means your shared capacity has grown substantially.
Enterprise API users: The 16x increase in input tokens per minute for Opus opens the door to use cases that were previously blocked by rate limits — large-scale document processing, multi-agent orchestration pipelines, and real-time production coding assistants.
Startups and indie developers: Even on standard Pro plans, the removal of peak-hour degradation levels the playing field. You’re no longer penalized for working during business hours alongside larger enterprise customers.
Key Takeaways
- Anthropic secured a compute deal with SpaceX to use Colossus 1, a 220,000+ GPU, 300+ MW data center.
- Claude Code five-hour rate limits are now doubled for Pro, Max, Team, and Enterprise users.
- Peak-hour throttling for Pro and Max Claude Code users has been completely removed.
- Claude Opus API throughput increased by up to 16x for Tier 1 users (input tokens per minute: 30K → 500K).
- This is part of a wider multi-petawatt compute expansion strategy including deals with Amazon, Google, Microsoft, and NVIDIA.
- The move directly addresses the #1 pain point for Claude Code users: hitting limits mid-session.
- Positions Anthropic competitively ahead of Google I/O 2026 and OpenAI’s GPT-5.5 rollout.
Frequently Asked Questions (FAQ)
What is the new Claude Code usage limit after the SpaceX deal?
As of May 6, 2026, Anthropic has doubled the five-hour rolling rate limit for Claude Code across all paid plans — Pro, Max, Team, and Enterprise. This means you can now sustain twice as many interactions within any five-hour window compared to the previous limits.
Does the SpaceX compute deal affect free users of Claude?
The immediate announced changes apply specifically to paid Claude plans (Pro, Max, Team, and Enterprise). However, increased infrastructure capacity generally benefits all users over time, as it reduces system-wide strain. Free tier improvements were not part of the May 6 announcement.
What is Colossus 1 and why does it matter for Claude?
Colossus 1 is SpaceX’s AI data center housing over 220,000 NVIDIA GPUs and more than 300 megawatts of compute capacity. It was originally built for large-scale model training. Anthropic is using its inference capacity to power Claude’s services — giving the company a massive boost in its ability to handle high-demand workloads like Claude Code without throttling users.
How does Anthropic’s compute expansion compare to OpenAI and Google?
All three companies are aggressively expanding compute. OpenAI upgraded ChatGPT to GPT-5.5 Instant in May 2026. Google is expected to announce Gemini 4 at Google I/O on May 19, 2026. Anthropic’s strategy is unique in its multi-partner approach — spanning SpaceX, Amazon, Google, Microsoft, and NVIDIA — which diversifies infrastructure risk while rapidly scaling total capacity.
Will Claude Code limits continue to increase in the future?
Very likely. Anthropic’s infrastructure deals — including 5GW agreements with Amazon and Google/Broadcom — are staggered, with more capacity coming online through 2026 and 2027. As new compute becomes available, users can expect continued improvements in usage limits and response speeds across all Claude plans.
What does this mean for businesses using the Claude API?
Enterprise and API users on Tier 1 are seeing Claude Opus input tokens per minute jump from 30,000 to 500,000 and output tokens per minute rise from 8,000 to 80,000. This dramatically expands what’s possible with production-scale deployments — including real-time document processing, multi-agent systems, and AI-powered development pipelines.
Conclusion
Anthropic’s SpaceX Colossus deal is more than a headline-grabbing partnership — it’s a tangible infrastructure upgrade that removes one of the most frustrating bottlenecks for Claude Code power users. Doubling five-hour limits, eliminating peak-hour throttling, and supercharging API throughput sends a clear message: Anthropic is ready to compete at scale.
With Google I/O 2026 just days away and OpenAI’s GPT-5.5 already deployed, the AI compute race is accelerating. Watch this space — the battle for developer mindshare is being fought not just in model benchmarks, but in reliability, limits, and infrastructure resilience. Anthropic just made a strong move on all three fronts.
