Musk says Anthropic won't be cut off from SpaceX servers, praises Mythos/Fable

Musk's reversal on Anthropic follows the company's May deal to purchase compute from xAI's Colossus 1 data center near Memphis, Tennessee. That transaction made Anthropic one of SpaceX's largest customers and occurred after xAI merged with SpaceX in February. The deal is significant because Musk had been critical of Anthropic's prospects, but the substantial financial arrangement—worth about $40 billion（約6.4兆円） to SpaceX through May 2029—alongside Anthropic's status as the reported leader in enterprise AI market share, appears to have shifted his public stance.

Musk's assurance that he will not "cut off" Anthropic addresses a real concern: proximity to a competitor's compute operations could enable intelligence gathering or operational interference. The body notes that AI model makers sometimes use "distilling"—creating fake accounts to probe a competitor's system—and that Musk acknowledged this practice is common in the industry. Anthropic accused three Chinese model makers of doing this to Claude in February. However, the contract itself provides legal consequences if Musk breaches it, and SpaceX's engineers may gain valuable knowledge supporting Anthropic's infrastructure, creating mutual benefit. Musk pointed to Tesla's 2014 patent-pledge decision and SpaceX's practice of launching competing satellite systems without price increases as evidence of his competitive style. Yet the body also notes he has sued OpenAI, suggesting his restraint varies by relationship. Over the contract's three-year term through May 2029, the partnership's stability and any competitive risks will likely become clearer.