Anthropic has been struggling to serve its users in recent times, and xAI has stumbled with its new model releases, and this situation seems to have made the two rivals come together.
SpaceXAI — the compute arm of Elon Musk’s xAI — has signed an agreement to provide Anthropic with access to Colossus 1, one of the world’s largest AI supercomputers. The deal gives Anthropic a significant infusion of computing power at a moment when it has been visibly straining under demand, and gives xAI a high-profile customer for infrastructure it has already built.
Anthropic’s Compute Crisis
The partnership comes after a prolonged period of capacity pain for Anthropic. Claude’s daily active users on iOS and Android have grown roughly 5x since October 2025, climbing from around 2 million to over 11 million — and the infrastructure has not kept up. In response, Anthropic has been forced into a series of increasingly unpopular moves: introducing weekly rate limits for Pro and Max subscribers, tightening 5-hour session windows during peak hours, and cutting off third-party tools like OpenClaw from subscription access entirely.
The company has also been dealing with self-inflicted wounds. Anthropic recently acknowledged three separate bugs that degraded Claude’s response quality for users, and a similar post-mortem was required when Claude Code’s performance deteriorated due to a series of stacked infrastructure changes. The turmoil gave OpenAI’s Sam Altman enough ammunition to publicly mock Anthropic, with his “ok boomer” jab going viral after Anthropic pulled Claude Code from its Pro plan for a subset of new users.
What Colossus 1 Brings
The compute injection is substantial. Colossus 1 features over 220,000 NVIDIA GPUs — including dense deployments of H100, H200, and GB200 accelerators — and was built from the ground up in record time. It is designed for large language model training, fine-tuning, inference, and high-performance computing at frontier scale. Anthropic plans to use the additional capacity to directly improve availability for Claude Pro and Max subscribers, the segment of its user base that has borne the brunt of the recent throttling.
A Strange Bedfellow
The partnership is an unusual one. Anthropic and xAI have not been friendly. Earlier this year, Anthropic cut off xAI’s access to its coding models on Cursor — forcing xAI engineers to scramble for alternatives — prompting X’s product head to call for banning Anthropic from the platform entirely. xAI co-founder Tony Wu told staff the move would “hurt productivity” but push the team to build their own coding products faster.
For its part, xAI has had a turbulent stretch. Half of its original co-founders have departed, Elon Musk himself admitted the company wasn’t built right first time around and is being rebuilt from scratch, and the company has faced a high-profile lawsuit after a former engineer allegedly stole its entire codebase before defecting to OpenAI. Meanwhile, Grok 4.3 — its latest model release — trails GPT-5.5 and Claude Opus 4.7 on the major composite intelligence benchmarks, a far cry from the brief high-water mark Grok 4 achieved when it was released in July 2025.
In other words, both companies needed something from the other.
The Bigger Play: Orbital Compute
The most striking element of the deal is buried at the end. Anthropic has expressed interest in partnering with xAI to develop multiple gigawatts of orbital AI compute capacity — essentially moving AI infrastructure to space. The rationale is that terrestrial power, land, and cooling constraints are already becoming a bottleneck for next-generation AI systems, and SpaceX is the only organization with the launch economics and constellation operations experience to make space-based compute a near-term engineering project rather than a thought experiment.
Whether the engineering challenges can actually be overcome remains to be seen. But the ambition signals where both companies think the compute ceiling is heading — and how far outside conventional infrastructure they’re willing to look.
For now, the more immediate win is simply more GPUs for Claude users. After months of throttling, rate limits, and public stumbles, Anthropic’s subscribers will be hoping the Colossus capacity translates into a noticeably more reliable product.