Elon Musk Says He Was ‘Clearly Wrong,’ Calls Anthropic ‘The Leader in AI’

Elon Musk said he was “clearly wrong” about Anthropic, calling the AI company “currently the leader in AI” in a sharp reversal of comments he made last year dismissing its chances of winning the race to build frontier models.

In a series of posts on X, Musk praised Anthropic’s Mythos and Fable models as the industry’s strongest while saying he would not use compute infrastructure controlled by his companies to hurt the AI startup, even as the two compete at the frontier of artificial intelligence.

The remarks came just days after SpaceXAI unveiled Grok 4.5, its latest frontier AI model, and offered a rare public acknowledgement from one of the industry’s most influential executives that a direct rival currently holds the technological lead.

Musk Reversed His Earlier View of Anthropic

Musk’s latest comments marked a notable shift from his position less than a year ago.

In September 2025, responding to technology commentator Yacine Mostafa (@yacineMTB), Musk wrote that “winning was never in the set of possible outcomes for Anthropic,” dismissing the startup’s long-term prospects in the increasingly competitive AI industry.

On Thursday, however, Musk publicly withdrew that assessment. “I was clearly wrong about Anthropic,” he wrote. “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.”

Musk did not explain what prompted the change in his assessment but credited Anthropic’s recent model releases as setting the current benchmark for frontier AI.

The Comments Followed SpaceXAI’s Launch of Grok 4.5

The reversal came only days after SpaceXAI launched Grok 4.5, its latest flagship model.

Ahead of the public release, Musk described Grok 4.5 as roughly comparable to Anthropic’s Opus 4.7 while arguing that its advantages lay elsewhere.

“Our internal assessment is that Grok 4.5 is roughly comparable to Opus 4.7, but much faster,” Musk wrote. “The combination of capability, faster speed and lower cost is what makes it competitive.”

He also said the company was focusing on practical performance rather than benchmark rankings, adding that engineers at Tesla and SpaceX were already using Grok 4.5 in day-to-day work.

The comments suggest Musk sees frontier AI competition extending beyond model capability alone to include inference speed, operating costs and real-world utility.

The Exchange Began With a Debate Over AI Infrastructure

Musk’s remarks came in response to a post from technology commentator Yacine Mostafa, who argued that Anthropic remained dependent on compute capacity leased from SpaceXAI and suggested Musk could theoretically use that relationship against a competitor.

Musk rejected the idea. “And I would never cut them off in a way that hurt them badly, even as a competitor,” he wrote. He argued that doing so would run counter to the way his companies have historically operated.

Musk cited Tesla’s decision to open-source its patents, the company’s move to make its Supercharger network available to rival automakers and SpaceX’s practice of launching competing satellite operators without imposing different commercial terms.

“Even my worst enemies can attack me on this platform,” he added, referring to X.

Anthropic Is One of SpaceXAI’s Largest Compute Customers

The exchange also highlighted the increasingly intertwined nature of the frontier AI industry.

According to SpaceX’s IPO filings, Anthropic agreed to pay approximately $1.25 billion per month for compute capacity at SpaceXAI’s Colossus data centers under a multi-year agreement running through May 2029. The arrangement gives Anthropic access to one of the world’s largest GPU clusters for training and operating its AI models.

Google has also signed a separate compute agreement with SpaceXAI, underscoring the company’s growing role not only as an AI developer through Grok but also as a supplier of large-scale AI infrastructure.

The relationship means some of the industry’s fiercest competitors increasingly rely on shared compute resources as demand for advanced AI infrastructure continues to accelerate.

The Grey Terminal Note

For much of the past two years, the AI race has been framed around benchmark scores and model launches. That narrative is becoming incomplete. 

Frontier AI is increasingly shaped by who controls the infrastructure beneath the models, from GPUs and power to data center capacity. Musk’s reversal is notable not simply because he praised a rival, but because it comes as competitors become customers, sharing the same compute backbone while competing for technological leadership. The next phase of the AI race may be defined as much by infrastructure ownership as by model performance.