Elon Musk has made a striking reversal on Anthropic, leasing SpaceX's massive Colossus 1 supercomputer to the AI rival after previously calling the company "misanthropic and evil." The deal grants Anthropic access to over 222,000 Nvidia GPUs, including powerful H100 and H200 chips alongside next-generation GB200 accelerator systems, plus 300 megawatts of compute power. The move signals a significant thaw in relations between Musk and one of the most prominent AI safety-focused companies in the industry.

What Changed Musk's Mind About Anthropic?

Just months ago, Musk was publicly critical of Anthropic and its flagship Claude AI model. But after spending time with senior members of the company to "understand what they do to ensure Claude is good for humanity," Musk said he was impressed by their approach. In an X post announcing the deal, Musk stated that he gave the green light to lease Colossus 1 after determining that "no one set off my evil detector". This represents a complete reversal from his earlier harsh characterizations of the company.

Musk

The Colossus 1 supercomputer was originally built to power xAI's own Grok models, but Musk is now directing the entire first-generation cluster toward powering one of his direct AI competitors. SpaceX is simultaneously building Colossus 2, which will serve xAI's future needs. The decision underscores how compute capacity has become the primary bottleneck in AI development, making it more economically rational to lease excess capacity than to leave it idle.

How Will Anthropic Use This Massive Computing Power?

• Expanded Claude Access: Anthropic has doubled Claude Code's five-hour rate limits for all paid tiers, including Pro, Max, Team, and Enterprise subscribers, effective immediately upon the deal announcement.
• Removed Peak Hour Restrictions: The company eliminated the peak hours limit reduction that previously applied to Pro and Max users, allowing consistent performance throughout the day.
• Increased API Rate Limits: Anthropic has considerably raised the API rate limits for Claude Opus models, enabling developers to make significantly more requests to the system.

Beyond immediate capacity improvements, Anthropic is also exploring more ambitious infrastructure partnerships with SpaceX. The company has expressed interest in developing "multiple gigawatts of orbital AI compute capacity," essentially data centers in space. According to SpaceX, "the compute required to train and operate the next generation of these systems is outpacing what terrestrial power, land, and cooling can deliver on the timelines that matter".

What Does This Deal Mean for the Broader AI Landscape?

The Colossus 1 lease reflects a broader trend in AI infrastructure strategy. Anthropic has simultaneously announced similar compute expansion deals with Amazon, Google, and Microsoft, all aimed at building gigawatts of additional capacity. This multi-pronged approach suggests that no single provider can meet the explosive demand for AI compute, forcing companies to diversify their infrastructure partnerships.

Anthropic is also signaling that its compute expansion strategy extends far beyond raw processing power. The company plans to deploy future infrastructure in international regions, such as Europe and Asia, to meet growing enterprise demand for local data residency and regulatory compliance, particularly in sectors like healthcare, finance, and government. This geographic diversification reflects the geopolitical reality of AI development, where data sovereignty and regulatory requirements increasingly shape infrastructure decisions.

The deal also introduces a new Claude feature called "dreaming," designed to help AI agents improve themselves between sessions. The feature allows Claude-powered agents to review previous work, identify recurring patterns and mistakes, and reorganize memory files containing user preferences and contextual information, enabling the system to refine its behavior over time rather than starting each session from scratch.

Musk's willingness to lease Colossus 1 to Anthropic, combined with his public praise for the company's safety-focused approach, suggests a potential shift in how competing AI companies view collaboration on infrastructure. Rather than viewing compute capacity as a zero-sum competitive advantage, the industry may be moving toward a model where companies lease excess capacity to rivals, prioritizing revenue and utilization over exclusivity. This pragmatic approach could accelerate AI development across the industry while also distributing the enormous capital costs of building supercomputers across multiple organizations.