Elon Musk publicly praised Anthropic's Mythos and Fable models on X, calling Anthropic the current leader in AI and stating that no company has released a model as capable -- a striking reversal from his September 2025 post declaring that "winning was never in the set of possible outcomes for Anthropic." Musk directly acknowledged the about-face: "I was clearly wrong about Anthropic," he wrote.
The more consequential part of Musk's comments wasn't the praise itself but a specific promise: he said he would never cut Anthropic off from SpaceX's compute infrastructure in a way that seriously damaged the company, even though Anthropic is a direct competitor to his own xAI/SpaceXAI models. The comments came after users on X floated the idea that Musk could use SpaceX's server access as leverage to hobble a rival lab -- a scenario Musk explicitly rejected. "That's not my style," he said, invoking Tesla's 2014 decision to open-source its patents and later open the Supercharger network to competing automakers as evidence of a consistent approach to competing on product rather than infrastructure denial.
The commercial backdrop makes the promise more than rhetorical. Anthropic is one of SpaceX's largest infrastructure customers, having signed a deal in May 2026 to buy 300 megawatts of compute from xAI's Colossus 1 data center near Memphis, Tennessee. That means Musk's own business has a direct financial incentive in Anthropic's continued growth, running in parallel with his personal and competitive rivalry with the company through xAI's own model lineup.
“The comments came after users on X floated the idea that Musk could use SpaceX's server access as leverage to hobble a rival lab -- a scenario Musk explicitly rejected.”
The episode is a useful data point in the broader trend of frontier AI labs becoming financially interdependent even as they compete head-on for the same enterprise and developer customers -- Microsoft's stake in OpenAI, Google and Amazon's tranches in Anthropic's $50 billion round, and now SpaceX selling meaningful compute capacity to a direct model-layer competitor. Public rivalry increasingly coexists with commercial entanglement rather than replacing it.
For founders building on any single lab's infrastructure, Musk's comments are a reminder that even the most vocally competitive infrastructure providers currently have strong commercial reasons not to weaponize access against customers who happen to also be competitors -- but that reasoning depends on the commercial relationship remaining profitable for both sides, not on any structural guarantee. For AI-lab watchers, the reversal on Anthropic itself is notable: Musk's September 2025 dismissal of Anthropic's chances was widely quoted at the time, and this walk-back is a real acknowledgment that Anthropic's model quality has outpaced his own prior public assessment.
The bear case: verbal commitments from Musk about not cutting off a competitor carry limited enforceability, and the relationship's durability ultimately depends on Colossus capacity remaining commercially attractive for Anthropic relative to Google Cloud, AWS and Azure alternatives it also uses. What to watch next: whether Anthropic expands or reduces its Colossus compute commitments in future disclosures, and whether Musk's public praise translates into any deeper commercial or model-benchmarking collaboration between the two labs.