Elon Musk publicly expressed admiration for the company Anthropic on the social network X on Thursday (the 9th) and assured that he would never restrict this company's access to SpaceX infrastructure. This statement was made in response to concerns from platform users who speculated that Musk might remove Anthropic from the servers to harm a competitor.
Musk admitted his past mistake, writing: 'I was clearly wrong about Anthropic.' He referred to his September 2025 post, in which he claimed that 'victory was never a possible outcome for Anthropic.' This change of opinion comes at a time when Anthropic is already considered an AI market leader, surpassing even OpenAI. In July 2026, Anthropic is one of SpaceX's largest clients.
In May 2026, Anthropic signed a major agreement to purchase 300 megawatts of computing power. This capacity matches the total production capability of xAI's Colossus 1 data center, located near Memphis, Tennessee, USA. The cost of this contract is $1.2 billion (6.4 billion rubles) monthly until May 2029, bringing approximately $40 billion (205.4 billion rubles) in revenue to the xAI division within SpaceX. It is important to note that xAI merged with SpaceX in February 2026. Furthermore, Google signed a separate agreement to lease SpaceX infrastructure until June 2029 for $920 million (4.7 billion rubles) monthly.
In his post on Thursday, Musk directly praised the competing lab, stating: 'Clearly, they are the current leader in AI. No company has released such a good model as Mythos/Fable, and they will undoubtedly release Mythos 2 soon. And I will never restrict them in a way that seriously harms them, even as a competitor. That is not my style.' As an example of his behavior, Musk cited Tesla's decision in 2014 not to file patent lawsuits against those who used its technology in good faith. This decision is recorded in the company's remote blog and is now documented in Tesla's patent policy. He also mentioned that Tesla opened its Supercharger network and designed charging ports for competitors, while SpaceX releases competing satellite systems without raising prices or using unfair terms, and that even its 'worst enemies' can attack it on X.
Historically, Musk has had contradictory episodes with competitors: he sued OpenAI, which lost in May 2026. In addition to Musk's statements, Anthropic has contractual guarantees in case of service interruption. SpaceX is also interested in maintaining the existing agreement: besides monthly payments, its engineers can gain experience working on creating and supporting the infrastructure used by Anthropic, similar to how Amazon engineers do. Nevertheless, there is an unresolved issue: hosting Anthropic's infrastructure could give SpaceX more information about the company's operating methods than most competitors. This issue became relevant during Musk's lawsuit against OpenAI, when the lawyer asked if xAI was 'extracting' technology from the competitor—a process where a company creates fake accounts to send requests to a competitor and study its operations. Musk replied: 'Generally, AI companies extract data from other AI companies.' In February 2026, Anthropic accused three Chinese model manufacturers of conducting similar attacks against Claude. Presumably, Anthropic and Google expect protection from SpaceX against this practice, but this issue remains open as the three-year contract progresses.
