SpaceXAI announced a compute partnership with Anthropic, giving the AI giant access to Colossus 1 as demand for large-scale model training and inference continues to strain global AI infrastructure.
According to xAI, the agreement will give Anthropic access to one of the world’s largest AI supercomputing clusters, expanding its compute resources with more than 220,000 NVIDIA GPUs, including H100, H200 and next-generation GB200 accelerators, while helping improve service capacity for Claude Pro and Claude Max subscribers.
Colossus expands Claude’s compute capacity
Colossus 1 was built to support complex AI training workloads, including large language models, multimodal systems, scientific simulations and generative AI applications. The cluster’s scale gives Anthropic a new pool of high-performance computing power at a time when leading AI labs are racing to secure chips, power and data center capacity.
Anthropic said the new capacity from Colossus 1 will help ease constraints across its paid Claude products, as the deal is expected to add more than 300 megawatts of compute capacity within the month, helping the company keep pace with rising demand for Claude.
When earth is not enough for AI compute
As part of the agreement, Anthropic expressed interest in working with SpaceXAI on multiple gigawatts of orbital AI compute capacity, a concept aimed at placing large-scale computing systems in space.
Supporters of space-based compute argue that orbital infrastructure could eventually tap abundant solar power while reducing the pressure that massive data centers place on land, water and terrestrial power grids.
SpaceXAI said the need to train and operate next-generation AI systems is beginning to exceed what Earth-based power, cooling and land resources can deliver on the required timelines.
The company framed SpaceX’s launch cadence, cost structure and constellation experience as key advantages that could move orbital compute from a research idea toward an engineering program.
While the concept has significant potential, it remains technically difficult, with unresolved challenges around launch economics, thermal management, maintenance, latency and hardware reliability in orbit.
Claude targets enterprise workflows with Mythos and finance agents
The Colossus deal comes as Anthropic accelerates a wider AI push around Claude, led by Mythos, an unreleased frontier model whose cybersecurity and multi-step reasoning capabilities were powerful enough for Anthropic to keep it inside the restricted Project Glasswing program, while also drawing U.S. government interest in limited access for major federal agencies.
In addition to Mythos, Anthropic recently launched 10 ready-to-use AI agent templates for financial services, targeting banks, asset managers and insurers.
The agents are designed to support research, client coverage, KYC screening, valuation reviews, financial modeling, ledger reconciliation, month-end close and financial statement checks.
Anthropic said the templates can run through Claude Cowork and Claude Code, or as Claude Managed Agents, allowing firms to adapt them to internal workflows, risk controls and approval processes while keeping employees responsible for reviewing outputs before use.