Elon Musk admits he was wrong about Anthropic, praises rival as current AI leader

Elon Musk Praises Anthropic After Years of Criticism

Elon Musk has publicly acknowledged that Anthropic leads the AI industry, marking a dramatic reversal from his previous dismissive stance toward the artificial intelligence startup. Writing on X on July 10, Musk admitted he was "clearly wrong" about Anthropic, a company he had repeatedly criticized over the past year1

1

. The acknowledgment is particularly significant given that Musk's own companies, SpaceX and xAI, directly compete with Anthropic through their Grok models, which challenge Anthropic's Claude models in the market2

2

.

"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 wrote, specifically highlighting the Mythos and Fable models that Anthropic released on June 91

1

. This praise comes just months after Musk claimed in September 2025 that "winning was never in the set of possible outcomes for Anthropic," and after he accused the company in February 2026 of stealing training data and called its models "misanthropic and evil"5

5

.

SpaceXAI Compute Partnership Creates Unusual Leverage

The context behind Musk's comments reveals a complex business relationship that gives him unusual power over his competitor. Anthropic signed a deal in May 2026 to lease the entire 300-megawatt output of xAI's Colossus 1 data center near Memphis, Tennessee, paying $1.25 billion per month through May 2029—a deal worth approximately $40 billion in total revenue for SpaceX's xAI unit1

1

. This arrangement grants Anthropic exclusive access to over 220,000 Nvidia GPUs and more than 300 megawatts of compute capacity at the Colossus 1 supercomputer2

2

.

Musk's comments came in direct response to an X user who suggested he could effectively "kill" Anthropic by cutting off its compute access. His reassurance was explicit: "I would never cut them off in a way that hurt them badly, even as a competitor. That's not my style"1

1

. He cited Tesla's 2014 decision to open-source its patents, the opening of Tesla's Supercharger network to competitors, and SpaceX's practice of launching competing satellite systems without price increases as evidence of his approach.

Competitive AI Development Intensifies Amid Market Shifts

The timing of Musk's admission is notable. SpaceXAI released Grok 4.5 on July 8, just two days before his post, with the model described by Axios as aimed at coding and agentic work4

4

. Independent benchmarks from Artificial Analysis place three Anthropic models at the top of AI performance rankings, with Grok 4.5 described as the "top non-Anthropic model"5

5

. This suggests that despite SpaceXAI's efforts in competitive AI development, Anthropic maintains a technical edge.

Anthropic's business momentum supports Musk's assessment. The company's valuation reached $1.2 trillion on secondary markets, up from a $965 billion post-money valuation during its $65 billion Series H round announced in late May2

2

. Annual recurring revenue surged from $9 billion at the end of 2025 to over $60 billion, according to SemiAnalysis, with the company generating more than $1 billion in quarterly profit2

2

. Anthropic's Claude Fable 5 model is now available across major platforms including Amazon Bedrock, Google Cloud, and Microsoft Foundry, signaling strong enterprise traction2

2

.

Risks and Benefits of Infrastructure Dependence

While contractual protections likely safeguard Anthropic from sudden infrastructure cutoffs, the arrangement presents both opportunities and risks. Google has also signed a deal to rent SpaceX infrastructure through June 2029 for $920 million per month, demonstrating the scale of SpaceX's infrastructure business1

1

. For SpaceX, hosting Anthropic's compute power provides not only substantial revenue but also valuable learning opportunities, as engineers gain experience building for and supporting rapidly growing AI models1

1

.

However, this proximity raises questions about competitive intelligence. During his trial against OpenAI, Musk acknowledged that AI "distilling"—where one model maker uses fake accounts to send prompts to competitors to learn how their systems work—is a real practice. When asked if xAI had distilled technology from OpenAI, Musk replied: "Generally AI companies distill other AI companies"1

1

. Anthropic accused three Chinese model makers of this practice in February1

1

.

Anthropic is moving toward public markets, having confidentially filed for a U.S. initial public offering on June 1, though timing and terms remain undisclosed4

4

. As the AI race accelerates and the three-year compute contracts age, the question remains whether Musk's cooperative stance will hold, or whether competitive pressures will shift the dynamics between these supercomputing infrastructure providers and the AI models they power.