Anthropic announced a major infrastructure partnership with SpaceX that will give Claude access to the full capacity of the Colossus 1 AI supercomputer in Memphis, one of the largest AI compute clusters in the world.
The deployment includes:
- 300+ megawatts of compute
- 220,000+ NVIDIA GPUs
- H100, H200, and next-generation GB200 chips
The new capacity is expected to come online within weeks and will immediately expand Claude’s capabilities and usage limits.
The announcement is strategically significant because SpaceX acquired xAI earlier this year, making this effectively a partnership between two of the biggest non-OpenAI AI players.
Key points
Massive increase in AI compute
Anthropic said it will use:
“the entire compute capacity of SpaceX’s Colossus 1 data center in Memphis.”
The scale is extraordinary:
“300+ megawatts, more than 220,000 NVIDIA GPUs.”
This places Anthropic among the most compute-intensive AI companies globally.
Claude usage limits are increasing immediately
The added infrastructure is already being reflected in product limits.
Anthropic announced:
“Claude Code’s five-hour rate limits are doubling for Pro, Max, Team, and Enterprise plans.”
It also said:
“Peak-hours throttling is being removed for Claude Code on Pro and Max.”
API limits for Claude Opus models are also increasing substantially.
Anthropic is building a hyperscale compute network
The SpaceX agreement is part of a broader infrastructure expansion that includes:
- 5GW partnerships with Amazon,
- 5GW with Google and Broadcom,
- and $30B of Azure capacity with Microsoft and NVIDIA.
This demonstrates how frontier AI companies are increasingly operating at utility and industrial scale.
Orbital AI infrastructure is now being discussed seriously
One of the most striking details was Anthropic’s interest in:
“multiple gigawatts of orbital AI compute with SpaceX.”
The idea points toward future space-based AI data centers powered by solar energy, potentially addressing long-term energy and cooling constraints on Earth.
Key questions raised
- Is AI now fundamentally constrained by compute and energy rather than algorithms?
- Will only a handful of companies be able to afford frontier-scale AI infrastructure?
- Can electrical grids sustain multi-gigawatt AI clusters?
- Does this deepen the industry’s dependence on NVIDIA hardware?
- Is orbital AI infrastructure realistic or strategic signaling?
Why important?
This announcement shows that the AI race is increasingly becoming an infrastructure and energy race, not just a software race. Access to GPUs, power, and data-center scale may determine which companies can build frontier AI systems.
The deal also signals a new phase of industry consolidation where rivals may cooperate on infrastructure because the cost and scale requirements are becoming so extreme.
Most importantly, the reference to “orbital AI compute” suggests leading AI firms are already thinking beyond terrestrial limits, treating energy and compute capacity as the defining bottlenecks of the next generation of AI. This will give competitors a lot to compute.
