Elon Musk’s $40B Anthropic Truce Hides AI Hostage Risk

Elon Musk was expected by some X users to hold a knife over Anthropic’s compute supply. Instead, he praised the rival AI lab and promised he would not “cut them off” from SpaceX infrastructure. That is reassuring theater, but the Elon Musk Anthropic relationship still should be governed like a strategic hostage risk, not a friendship.

TechCrunch reports that Anthropic signed a May deal to buy 300 megawatts of compute, the full output of xAI’s Colossus 1 data center near Memphis, Tennessee. The company agreed to pay $1.25 billion per month through May 2029, a deal worth about $40 billion in revenue for SpaceX’s xAI unit after xAI merged with SpaceX in February.

That number changes the story. This is not a normal vendor contract. It is a rival AI company renting core production capacity from a Musk-controlled company.

The Elon Musk Anthropic truce doesn't erase the conflict

Musk’s tone has flipped hard. In September 2025, he posted on X that “Winning was never in the set of possible outcomes for Anthropic.” Now, he says he misread the company.

“I was clearly wrong about Anthropic,” Musk wrote on Thursday.

He went further, calling Anthropic the current leader.

“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,” he wrote.

That is a striking endorsement from the founder of xAI, a direct competitor. It also proves the problem. Musk is not a neutral infrastructure executive. He is both landlord and rival.

XOOMAR analysis: public praise can lower panic, but it does not remove incentives. Anthropic needs uptime, predictability, and neutrality. Musk’s statement offers admiration. A board cannot build risk controls out of admiration.

Compute hosting gives Musk leverage Anthropic can't ignore

Model hosting is not back-office plumbing. For a frontier AI lab, infrastructure shapes product reliability, customer confidence, deployment speed, and operating freedom. If a provider controls the pipes, that provider has leverage even without pulling the plug.

The harsh version is a shutdown. The softer versions matter too:

• Access: delays or technical friction can slow deployments.
• Priority: resource allocation can shape performance under load.
• Support: poor escalation can turn incidents into customer-facing failures.
• Pricing: future contract terms can become pressure points.
• Visibility: hosting workloads can expose operational patterns that competitors rarely see.

Anthropic’s enterprise position depends on being boring in the best sense. Customers want the model to work, the terms to hold, and the vendor chain to stay quiet. A week of uncertainty can do reputational damage before any legal claim is filed.

The Elon Musk Anthropic deal also sits beside Google’s own agreement to rent SpaceX infrastructure through June 2029 for $920 million per month, according to the same TechCrunch report. That strengthens the case that SpaceX wants to be taken seriously as AI infrastructure. It also makes the neutrality question bigger than Anthropic.

For adjacent XOOMAR coverage on how AI model risk is becoming a board-level problem, read Model Risk Lands on AI Firms as Trump Rejects FDA for AI. For the customer-side version of the same dependency question, see Model Lock-In Cracks as Vercel AI Agents Pick Labs.

Musk's examples help his case, but they don't settle it

Musk offered evidence that he does not squeeze competitors. He pointed to Tesla’s 2014 patent pledge, now housed under its patent pledge, saying Tesla would not initiate patent lawsuits against anyone using its technology in good faith. He also cited Tesla opening its Supercharger network and charging port design to competitors.

He added that SpaceX launches competing satellite systems “with no increase in price or use of unfair terms,” and said even his “worst enemies” can attack him on X.

Those examples matter. They show Musk understands the reputational value of appearing fair when he controls strategic infrastructure. That is the strongest argument for trusting him.

But it is not the whole record supplied here. TechCrunch also notes Musk sued OpenAI. And during his trial against OpenAI, Musk acknowledged that AI “distilling” is real, a process where one model maker uses many fake accounts to prompt a rival model and learn how it works.

“Generally AI companies distill other AI companies,” Musk said, according to the New York Times account cited by TechCrunch.

Anthropic accused three Chinese model makers in February of doing this to Claude. There is no claim in the supplied material that SpaceX or xAI has done that to Anthropic. Still, infrastructure proximity creates a different kind of exposure. Hosting a rival’s compute can reveal operational rhythms, support needs, scaling constraints, and deployment behavior.

XOOMAR analysis: the issue is not that Musk will necessarily abuse the relationship. The issue is that Anthropic would be reckless to assume he never could.

Before and after Musk's public reassurance

The counterargument is real. Compute is hard to secure at this scale. Anthropic may have decided that 300 megawatts from Colossus 1 was worth the discomfort of relying on a competitor. Musk also has a clear financial incentive to keep the deal alive. $1.25 billion per month is not loose change, even in frontier AI.

That is why the right answer is not “never work with Musk.” It is “never depend on Musk without hard exits.”

Anthropic needs contracts that survive the next fight

If Anthropic is going to rely on Musk-controlled infrastructure, it needs protections that do not care what anyone posts on X.

The minimum package should include:

• Multi-site redundancy: Anthropic should be able to move critical workloads away from Colossus 1 without a crisis.
• Strict uptime guarantees: service commitments need teeth, not aspirational language.
• Non-discrimination clauses: Anthropic should receive no worse treatment because it competes with xAI.
• Portability requirements: deployments, data flows, and tooling should be designed for rapid migration.
• Financial penalties: unjustified service disruption should be expensive enough to deter gamesmanship.
• Audit and escalation rights: Anthropic needs clear visibility when infrastructure decisions affect model delivery.

This is where the board should be ruthless. If about $40 billion in SpaceX/xAI revenue is tied to Anthropic’s contract, both sides have incentives to keep the machine running. But incentives are not safeguards. They are weather. Contracts are shelter.

Anthropic also needs to assume that public relationships can sour. Musk’s praise for Mythos/Fable may be sincere. It may also be temporary. The company cannot let a rival’s current mood become a single point of failure.

AI companies should trust redundancy, not billionaire reassurance

The Elon Musk Anthropic story is not mainly about whether Musk means what he says today. He may. The sharper question is whether any AI company should place critical model capacity inside a competitor’s house and then rely on personal style as a safety mechanism.

Musk’s line, “That’s not my style,” is useful as a signal. It is useless as a control.

Anthropic should demand enforceable guarantees, keep credible alternatives ready, and disclose concentration risks where appropriate. SpaceX may prove to be a reliable AI infrastructure provider. Musk may decide that neutrality is more profitable than pressure. Good. Put it in writing, test the exits, and rehearse the failover.

In AI, the company that controls compute can end up controlling everyone else’s options. Anthropic should not hand that power away cheaply.

The Bottom Line

• Anthropic is relying on a Musk-controlled company for core AI compute despite competing with xAI.
• The deal’s scale makes infrastructure neutrality and uptime a board-level risk, not just a vendor issue.
• Musk’s public promise may calm concerns, but contractual safeguards matter more than personal assurances.