In a deal that subordinates rivalry to resource scarcity, Elon Musk’s SpaceXAI is now renting its most powerful supercomputer to direct competitor Anthropic.
In a deal that subordinates rivalry to resource scarcity, Elon Musk’s SpaceXAI is now renting its most powerful supercomputer to direct competitor Anthropic.

A landmark deal giving Anthropic access to over 220,000 Nvidia GPUs in a SpaceXAI data center shows that compute scarcity is now the single most powerful force in the artificial intelligence industry. The agreement, announced May 6, converts Elon Musk’s underutilized AI assets into a new cloud provider, forcing a strategic pivot that complicates the valuation narrative for SpaceX’s anticipated IPO later this year.
"No one at Anthropic set off my evil detector," Musk said in a public statement, a sharp reversal from his previous criticisms of the company as "woke" and "evil." The necessity of the deal for both parties seemingly overrode prior public animosity.
Under the terms, Anthropic gains immediate and exclusive access to Colossus 1, the Memphis, Tennessee data center xAI built in 2024. The site provides over 300 megawatts of compute capacity, including H100, H200, and the newest GB200 accelerators from Nvidia. The deal provides a critical bridge for Anthropic, whose own massive compute agreements with Amazon and Google are not expected to deliver significant capacity until late 2026 and 2027.
For investors, the move reframes SpaceXAI from a pure AI model developer into a direct competitor with cloud providers like CoreWeave and the hyperscalers themselves. It allows SpaceX to monetize a data center asset that was reportedly operating at just 11 percent utilization, while providing a marquee customer to build a new AI infrastructure business line ahead of its public offering.
The decision to rent out its powerful infrastructure stems from the underwhelming market adoption of SpaceXAI’s own Grok model. While leading AI labs like Anthropic and OpenAI are compute-constrained, SpaceXAI found itself with a surplus of expensive, idle capacity. Renting to a rival is a clear strategy to generate revenue from an underperforming asset.
The situation mirrors the business model of specialized cloud providers like CoreWeave, which focus purely on supplying high-demand AI hardware. However, for a company intended to be at the forefront of AI model development, becoming a landlord is a difficult narrative. The rebranding from the standalone xAI to SpaceXAI, a product line within SpaceX, may be an attempt to soften this strategic shift. For a space and logistics company to offer a "cloud product" is more palatable than an AI research lab admitting its core product isn't using its own resources.
The deal's urgency was high enough to overlook significant external issues. Anthropic is currently in litigation with the U.S. government after the Pentagon designated it a "supply chain risk," while the Defense Department is actively integrating SpaceXAI's Grok. Furthermore, the Colossus 1 data center itself is controversial for its use of natural gas turbines in a historic Memphis neighborhood, drawing criticism from civil rights groups over environmental concerns.
For the pending SpaceX IPO, this pivot is a double-edged sword. On one hand, it creates a new, tangible revenue stream and demonstrates an ability to adapt. On the other, it invites direct valuation comparisons to more established cloud companies like Amazon Web Services, potentially discounting the futuristic appeal of its AI division.
To counter this, Musk is promoting a narrative of "orbital AI data centers." The agreement with Anthropic included a clause where both parties "expressed interest in partnering to develop multiple gigawatts of orbital AI compute capacity." While this remains purely speculative, it aligns with SpaceX's long-term vision and attempts to frame its data center business as a step toward a much larger, space-based infrastructure play that competitors cannot easily replicate.
The deal's impact on Anthropic was immediate. The company announced it was doubling rate limits for its Claude Code models and increasing API capacity for its most advanced Opus model, showing the new compute power was being deployed within hours. For SpaceX, landing the AI industry's highest-growth customer provides a powerful proof point for its nascent cloud business.
The move solidifies a new reality in the AI sector: access to compute is more important than corporate rivalry or public perception. The companies that benefit most are the chip suppliers like Nvidia and the infrastructure players who can deliver capacity now, not in two years. While SpaceX may not have intended to be a cloud provider, the market has forced its hand, creating a formidable new competitor in the process.
This article is for informational purposes only and does not constitute investment advice.