Elon Musk Says He Won’t Cut Off Anthropic as SpaceX Hosts Billions in AI Compute

Elon Musk has publicly said he would not use SpaceX’s infrastructure business to undermine Anthropic, addressing concerns that one of the AI industry’s biggest model developers could be vulnerable if it depends on compute controlled by a direct competitor.

The issue surfaced after discussion on X about whether Musk could one day remove Anthropic from SpaceX-hosted infrastructure in order to weaken a rival AI lab. In response, Musk said that was “not my style” and offered unusually strong praise for Anthropic, writing that he had been “clearly wrong” in an earlier assessment of the company.

A major AI infrastructure relationship

The comments matter because Anthropic is reportedly now one of SpaceX’s largest customers. According to TechCrunch, Anthropic signed a deal in May to buy 300 megawatts of compute capacity, tied to infrastructure associated with xAI’s Colossus 1 data center near Memphis. After xAI’s merger with SpaceX earlier this year, that agreement was said to be worth about $1.25 billion per month through May 2029, implying roughly $40 billion in revenue.

That scale turns Musk’s remarks into more than social-media commentary. They touch on a core question in the AI market: whether model companies can safely rely on infrastructure owned by firms whose strategic interests may not be aligned with their own.

Musk praises a rival

Musk described Anthropic as “currently the leader in AI” and said no company had released a model as good as “Mythos/Fable,” while suggesting “Mythos 2” could follow soon. He also said he would not cut Anthropic off “in a way that hurt them badly, even as a competitor.”

That language is notable given Musk’s broader rivalry with major AI labs and his history of sharply criticizing competitors. His comments suggest an attempt to reassure customers that SpaceX and xAI can operate a large-scale infrastructure business without weaponizing control over compute access.

The bigger industry concern

The Anthropic-SpaceX arrangement highlights a growing tension in AI: the same companies building foundation models are increasingly entwined with the firms that supply data centers, chips, energy, and cloud capacity. As training and inference costs rise, labs need access to enormous compute commitments, but those commitments can create concentration risk if a provider is also a market rival.

Musk pointed to Tesla’s past patent stance and the opening of its Supercharger network as examples of competing while still enabling broader ecosystem growth. Whether that analogy convinces enterprise buyers is another question. In AI, switching infrastructure at scale is far harder than changing software vendors, and outages or contract disputes could carry major commercial consequences.

For Anthropic, the immediate takeaway is that Musk is trying to publicly de-escalate fears around the partnership. For the wider market, the episode is a reminder that AI competition is no longer just about model performance. It is also about who controls the physical and financial backbone of the industry.