Anthropic has announced a compute partnership with SpaceXAI to access the Colossus 1 AI supercomputer. The agreement increases compute capacity for Claude services, updates usage limits, and supports future exploration of large-scale AI infrastructure, including orbital compute development.
Colossus 1 supercomputer and compute capacity
SpaceXAI’s Colossus 1 is one of the world’s largest and fastest-deployed AI supercomputers, built to support large-scale AI workloads.
Key specifications include:
- Over 220,000 NVIDIA GPUs
- H100, H200, and next-generation GB200 accelerators
- Around 300 megawatts of compute capacity
- Designed for AI training, fine-tuning, inference, and high-performance computing
The system supports workloads such as large language models, multimodal systems, scientific simulations, and generative AI at scale.
Claude usage limits and API updates
The expanded compute capacity enables changes across Anthropic’s Claude platform.
Claude Code
- Five-hour rate limits are doubled for Pro, Max, Team, and Enterprise plans
- Peak-hour usage limit reductions for Pro and Max accounts are removed
API changes
- API rate limits for Claude Opus models are increased
- Detailed API rate limit table:
Compute partnership and infrastructure access
Anthropic has signed an agreement with SpaceXAI to use the full capacity of the Colossus 1 data center.
This provides:
- Over 300 megawatts of compute capacity
- More than 220,000 NVIDIA GPUs
- Infrastructure for large-scale AI training and inference
The capacity will directly support scaling for Claude Pro and Claude Max users.
Orbital AI compute exploration
Anthropic and SpaceXAI have expressed interest in developing multi-gigawatt orbital AI compute systems.
This is driven by:
- Increasing global AI compute demand
- Constraints in Earth-based power, cooling, and land availability
The concept explores space-based compute as a potential long-term infrastructure model, with reduced environmental impact on Earth.
SpaceX is identified as a key enabler due to its launch operations, orbital deployment experience, and constellation infrastructure capabilities.
Broader compute expansion strategy
This agreement is part of Anthropic’s wider infrastructure expansion across multiple providers:
- Up to 5GW agreement with Amazon (including near-term expansion)
- 5GW partnership with Google and Broadcom
- $30 billion Azure capacity collaboration with Microsoft and NVIDIA
- $50 billion US AI infrastructure investment with Fluidstack
Anthropic operates across multiple compute systems including AWS Trainium, Google TPUs, and NVIDIA GPUs.
International infrastructure expansion
Anthropic is expanding infrastructure globally to support enterprise and regulated-sector requirements.
Focus areas include:
- In-region infrastructure for compliance and data residency
- Expansion of inference capacity in Asia and Europe
- Secure supply chains for hardware and networking systems
- Deployment in jurisdictions with stable regulatory frameworks
Energy and infrastructure commitments
Anthropic has stated it will cover any consumer electricity price increases in the United States caused by its data centers. The company is also exploring extending this approach to other regions and reinvesting in local communities hosting infrastructure.
Availability
Anthropic will gain access to Colossus 1 compute capacity within one month of the agreement.
This includes:
- Deployment of over 300MW compute capacity
- Integration of 220,000+ GPUs into Anthropic systems
- Immediate scaling of Claude Pro, Max, and API workloads after activation
- Usage improvements directly tied to incoming compute availability