Elon Musk says he was wrong about Anthropic
Anthropic runs some of the most powerful AI models in the world on computing hardware it doesn't own. The company signed a deal in May to lease the full capacity of SpaceX's Colossus 1 supercomputing facility in Memphis, paying roughly $1.25 billion a month through 2029. That arrangement gives ...
Overview
Anthropic runs some of the most powerful AI models in the world on computing hardware it doesn't own. The company signed a deal in May to lease the full capacity of SpaceX's Colossus 1 supercomputing facility in Memphis, paying roughly $1.25 billion a month through 2029. That arrangement gives Musk's company an unusual kind of leverage over one of his biggest rivals.
Someone on X pointed that out to Musk directly. The user suggested that if Musk wanted to, he could effectively kill Anthropic by cutting off its compute access. Musk's response was not what most people expected.
What Musk posted and why it's significant
Musk replied on July 10 and did something he rarely does: admitted a rival had beaten him. The full post walked back everything he'd said about Anthropic over the past year.
"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. Tesla open sourced its patents and we made the Supercharger network available to all competitors, even though we could have made it a walled garden. SpaceX launches competing satellite systems with no increase in price or use of unfair terms. Even my worst enemies can attack me on this platform," Musk wrote on X.
More Elon Musk:
- Elon Musk has a radical fix for an AI-dominated economy
- Elon Musk reveals his grand ambitions for SpaceX ahead of IPO
- Elon Musk sets SpaceX IPO price in blunt message to Wall Street
He wasn't randomly praising a competitor. He was responding to a user who pointed out that SpaceX controls the computing infrastructure Anthropic relies on to run its models: over 220,000 Nvidia GPUs at the Colossus 1 data center in Memphis, which CNBC reported Anthropic is paying roughly $1.25 billion a month to access, according to TechCrunch.
The user suggested Musk could "kill" Anthropic if he wanted. Musk's response was to say publicly he wouldn't.
How sharply this reverses what Musk had been saying
Musk's comments on Anthropic over the past year tell a very different story. In September 2025, he wrote that "winning was never in the set of possible outcomes for Anthropic." In January 2026, he joked that the company's "fate is to be misanthropic."
It got sharper in February, when Anthropic raised $30 billion at a $380 billion valuation. Musk accused the company of stealing training data "at massive scale," called its Claude models "misanthropic and evil," and wrote that "Anthropic hates Western Civilization." In March, he asked publicly: "Is there a more hypocritical company than Anthropic?"
The shift started in May, though. On May 6, the same day the Colossus 1 partnership was announced, Musk wrote on X that he'd spent time with senior Anthropic team members and was impressed.
"Everyone I met was highly competent and cared a great deal about doing the right thing. No one set off my evil detector," he said.
Details
Even then, he held the door open for competition, noting SpaceXAI is only three years old and suggesting the picture could look different in three more years. July 10 dropped that caveat entirely.
Henry/Getty Images
What Anthropic's models actually achieved to force this
Musk specifically named Claude Mythos 5 and Claude Fable 5. Anthropic launched both models in June 2026 before the Trump administration restricted access under an executive order requiring national security reviews of frontier AI systems.
The U.S. government was concerned about suspected Chinese hackers accessing Mythos. Fable 5 was restored for users on July 1 after export controls were lifted. Mythos remains under tighter access restrictions.
Independent benchmarks back up what Musk said. Artificial Analysis, a service that rates AI models based on independent performance data, places three Anthropic models at the top of its rankings.
Grok 4.5, which SpaceXAI launched on July 8, two days before Musk's post, is described as the "top non-Anthropic model." Musk acknowledged Anthropic's lead the same week he released a model specifically designed to compete with them.
What this means for the AI race and investors watching it
Musk's post lands differently than a typical compliment from a competitor because of what it comes alongside.
He leads SpaceXAI, which only officially rebranded this week after merging with xAI earlier this year. His company is competing directly with Anthropic on models while simultaneously acting as Anthropic's infrastructure provider. He controls the compute Anthropic depends on and just said publicly that Anthropic's models are the best available.
For investors, it's a useful data point on where the AI race actually stands. Musk made a comment on May 26 that framed the longer picture: SpaceXAI is three years old, Anthropic is six, OpenAI is twelve. He wasn't conceding the race then. He's not conceding it now. What he's saying is that Anthropic has the lead today, and it's earned.
Anthropic is private, so there's no direct ticker to watch. But what happens to Anthropic shows up everywhere: in the AI spending decisions of public companies using its models, in the valuations of cloud providers like Google supplying its compute, and in the pressure it puts on every public AI company competing against Claude. When the CEO of SpaceXAI calls Anthropic the current leader, that shapes how the people making those spending decisions think about who's ahead.
Related: Anthropic restores access to Mythos 5 for select organizations
Source
Originally published at www.thestreet.com.