xAI’s $230 billion pivot from AI models to data center rental business

xAI made a shocking announcement this week that could reshape how we think about the AI industry’s future. The Elon Musk-backed company revealed a partnership with Anthropic that involves selling all compute capacity at its Colossus 1 data center – roughly 300MW worth of processing power – to the Claude AI maker. The deal, likely worth billions of dollars, immediately allowed Anthropic to raise its usage limits.

This move signals something much bigger than a simple business transaction. It suggests that xAI, valued at $230 billion in January, may be shifting its focus from developing AI software to becoming a data center rental business. That’s a major strategic pivot that puts the company at odds with how other tech giants like Google and Meta approach AI infrastructure.

While Musk explained on X that xAI had already moved training operations to a newer Colossus 2 data center and simply didn’t need both facilities, the timing raises questions about the company’s true priorities. xAI’s main product, Grok, has seen usage drop significantly since image generation problems earlier this year. If the company built more data center capacity than Grok actually needs, renting to competitors like Anthropic creates immediate revenue.

This strategy makes particular sense as xAI prepares for a potential IPO alongside SpaceX. Having Anthropic as a major customer also adds credibility to SpaceX’s ambitious orbital data center plans, showing there’s real demand for compute-as-a-service offerings.

But the Anthropic deal reveals an unusual approach that sets xAI apart from other major tech companies. When Google and Meta face similar capacity constraints, they consistently choose to keep compute resources for their own AI development rather than rent them out to customers.

Just last month, Google’s Sundar Pichai admitted that Google Cloud revenue was lower than it could have been because the company was “capacity constrained.” When forced to choose between renting GPUs or using them internally for AI products, Google picked internal development. Meta took this approach even further, creating an entirely new cloud infrastructure system specifically to ensure enough GPU power for Mark Zuckerberg’s AI ambitions.

As Zuckerberg said when announcing Meta Compute in January: “How we engineer, invest, and partner to build this infrastructure will become a strategic advantage.” Both Google and Meta view computing power as essential for building tomorrow’s most valuable AI products, not just meeting today’s demand.

By choosing the opposite path, xAI positions itself more like what’s called a “neocloud” business – companies that buy GPUs from Nvidia and rent them to model developers. This business model faces significant challenges:

• Squeezed by chip suppliers like Nvidia on one side
• Subject to shifting demand cycles from AI companies
• Lower valuations compared to AI software companies
• Less control over long-term product development

The valuation gap tells the story clearly. While xAI reached a $230 billion valuation, CoreWeave – which manages comparable computing power – is worth less than a third of that amount.

Musk’s version of the neocloud model has ambitious twists, as expected. Some data centers might operate in space by 2035, and xAI plans to manufacture its own chips at the Terafab facility to reduce dependence on Nvidia’s pricing. But these innovations don’t change the fundamental economics of renting compute capacity to other companies.

The strategic shift becomes more significant when you consider xAI’s software ambitions. As recently as February, the company had major plans beyond data centers, including coding partnerships like the recent Cursor collaboration and projects involving digital twins. These long-term software projects typically require dedicated computing resources to succeed.

As long as xAI sells large amounts of compute capacity to competitors like Anthropic, it becomes harder to pursue those software ambitions effectively. The company seems to be choosing immediate revenue from data center rentals over the potential long-term value of developing breakthrough AI applications.

This decision puts xAI in a fundamentally different category than other major AI companies. While Google, Meta, and even OpenAI prioritize keeping compute resources for internal development, xAI appears willing to become the infrastructure provider that powers its competitors’ innovations.