Elon Musk Admits He Was "Clearly Wrong" About Anthropic, Pledging Safe Compute Access Under $40 Billion Deal
In a dramatic public reversal, Elon Musk has declared that his long-time rival Anthropic is "currently the leader in AI," praising its flagship Claude Mythos and Fable models as the best in the industry. The public U-turn, posted on X on July 9, 2026, marks a complete departure from Musk's previous dismissals of the company and highlights the complex, trillion-dollar interdependencies of the frontier AI compute economy.
This public realignment comes as Anthropic operates as one of SpaceX's largest commercial customers. In May 2026, Anthropic signed a monumental contract to lease 300 megawatts of compute—representing the entire output of xAI’s Colossus 1 data center in Memphis, Tennessee (which was absorbed into SpaceX in February)—for $1.25 billion per month through May 2029. This single contract is worth approximately $40 billion in revenue for SpaceX.
Pledging Safeguards Against Infrastructure Sabotage
The public praise was triggered after users on X speculated that Musk could arbitrarily boot Anthropic from SpaceX's data centers to kneecap a direct competitor to his own xAI Grok models. Musk responded directly to these concerns, writing:
"I was clearly wrong about Anthropic. 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."
Musk cited Tesla's open-sourcing of patents and the opening of the Supercharger network, as well as SpaceX launching competing satellite systems, as historical proof that he does not engage in anti-competitive infrastructure squeezing.
The Strategic Proximity of Competitor Compute
While the $40 billion contract provides immense financial liquidity for SpaceX, hosting Anthropic’s compute carries significant strategic advantages for Musk's conglomerate. In recent legal proceedings against OpenAI, Musk acknowledged the prevalence of "AI distillation"—the practice of using a competitor's model outputs to train one's own system—stating that "Generally AI companies distill other AI companies."
Hosting Anthropic's massive model workloads (alongside a separate $920 million per month compute lease signed by Google in June 2026) gives SpaceX engineers unprecedented operational visibility into how these frontier models run at scale. While Anthropic and Google maintain strict digital safeguards, the physical proximity of their workloads to SpaceX infrastructure remains a source of immense competitive intelligence.1
-
An instance of Platform bans cannot prevent rivals from distilling superior model outputs to train their own. — Hosting a competitor's workloads on adjacent hardware provides operational visibility that helps the landlord study and run the competitor's high-value model patterns. ↩︎