Elon Musk said on Thursday he would not cut off Anthropic from the 300 megawatts of SpaceX compute it depends on, calling the AI lab the current leader in the industry and reversing a hostile posture he held as recently as September 2025. The pledge, made in a post on X, addresses one of the more awkward dependencies in the AI market: Anthropic pays $1.25 billion per month through May 2029 to rent the entire output of xAI's Colossus 1 data center near Memphis, Tennessee, a contract worth roughly $40 billion in revenue for SpaceX's xAI unit.
"I was clearly wrong about Anthropic," Musk wrote, referring to his September 2025 claim that "winning was never in the set of possible outcomes for Anthropic." At the time of that original post, Anthropic was already reported to hold the largest AI market share among enterprise buyers.
“I was clearly wrong about Anthropic”— Elon Musk, CEO of xAI and SpaceX
The compute contract was signed in May, three months after xAI merged into SpaceX in February. It gives Anthropic exclusive use of Colossus 1's full 300 MW output and locks in pricing through May 2029. Google signed a parallel arrangement to rent SpaceX infrastructure through June 2029 at $920 million per month, meaning two of Anthropic's largest model-market rivals now sit on the same landlord's balance sheet.
Key facts
- 01Anthropic pays SpaceX's xAI unit $1.25B per month through May 2029 for 300 MW of compute at the Colossus 1 data center near Memphis.
- 02The deal is worth roughly $40B in revenue for SpaceX's xAI unit over its term.
- 03Google signed a parallel SpaceX hosting deal at $920M per month through June 2029.
- 04Musk publicly reversed his September 2025 claim that 'winning was never in the set of possible outcomes for Anthropic.'
- 05Anthropic in February accused three Chinese model makers of distilling Claude.
That arrangement is the reason Musk's Thursday comments matter. X users had begun openly speculating that Musk could throttle or terminate Anthropic's hosting to hobble a competitor. His response was to praise the company by name and cite prior examples of what he framed as competitive restraint: Tesla's 2014 patent pledge, the opening of the Supercharger network and its connector design to rivals, and SpaceX's satellite-launch pricing for competing constellations.
Musk singled out Anthropic's models specifically, using the internal codenames Mythos and Mythos 2.
“They are obviously currently the leader in AI. No company has released a model as good as Mythos/Fable and they will undoubtedly have Mythos 2 ready soon. And I would never cut them off in a way that hurt them badly, even as a competitor. That's not my style”— Elon Musk, CEO of xAI and SpaceX
The commercial logic reinforces the pledge independently of style. Cutting off a customer paying $1.25 billion a month would trigger contractual consequences, and xAI's revenue base would collapse without the Anthropic and Google contracts anchoring Colossus 1. SpaceX's engineering teams also gain operational knowledge from running production workloads for a frontier lab, the same dynamic that has made Amazon a significant beneficiary of hosting Anthropic on AWS.
Proximity cuts both ways. Hosting a competitor's inference and training workloads gives the landlord unusual visibility into how those workloads behave, what hardware they stress, and how they scale. That is not the same as access to model weights, but it is more insight than any arms-length competitor would normally have.
Musk himself has publicly acknowledged that model-to-model extraction is a live practice in the industry.
“Generally AI companies distill other AI companies”— Elon Musk, CEO of xAI and SpaceX
He made the comment during his lawsuit against OpenAI, where a lawyer asked whether xAI had distilled OpenAI technology. Anthropic itself accused three Chinese model makers in February of distilling Claude through fake-account prompting. Both Anthropic and Google presumably have contractual and technical safeguards against their landlord doing anything similar, but the structural conflict is unusual by any standard of cloud hosting.
The skeptical read is that Musk's assurances are worth what any three-year contract is worth when relationships sour. He has litigated against former partners before, most notably OpenAI, and the current warmth toward Anthropic is a full reversal from a stance he held ten months ago. Contract terms, not temperament, are what protect Anthropic here.
For the AI market, the more interesting signal is the acknowledgment itself. Musk publicly ranking Anthropic ahead of xAI on model quality, while collecting $1.25 billion a month to power its training runs, is a clean illustration of how thoroughly compute economics now outweigh competitive posture at the top of the market. The frontier labs are each other's largest customers, landlords, and rivals simultaneously, and the shape of the next three years of AI competition will be dictated as much by data-center leases as by model releases.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
