Meta considers Meta Compute cloud, selling excess AI compute

Meta is developing plans to sell access to AI compute power and hosted models from its data centers, a move that would turn its infrastructure spending into a commercial cloud business. Bloomberg reported the plans, which TechCrunch says would live under an initiative reportedly dubbed Meta Compute.

What will Meta sell and who will run it?

Meta would sell both raw AI compute capacity and access to hosted models, and the new unit is reportedly led by Santosh Janardhan, Daniel Gross, and Dina Powell McCormick. The report says Meta may copy CoreWeave's model and sell "raw" compute capacity, and it is also considering following AWS's lead by selling access to hosted models, including its closed-weight model Muse Spark.

Meta has already committed substantial capital to the hardware those offerings would run on: as of the end of the first quarter, Meta had committed to spending $182.9 billion on AI infrastructure in the coming years, and the company expects an Ohio project that Mark Zuckerberg said would be the size of Manhattan to come online this year.

How does this compare with recent industry moves?

SpaceX and smaller cloud players have already started turning excess AI compute into a product, and Meta’s shift places it in direct competition with major cloud providers. In early May, SpaceX, via xAI, signed a deal with Anthropic to buy out all of the compute capacity at SpaceX’s Colossus 1 data center, and SpaceX has signed similar leases with Google and Reflection AI. Bloomberg framed Meta’s plans as a parallel to those moves and said the company would be pitting itself against Amazon Web Services, Google Cloud, and Microsoft Azure.

The TechCrunch analysis notes Meta has not yet seen significant end-user demand for its own AI services: the company does not break out revenue for Meta AI or the Llama model family, and executives have emphasized internal corporate uses. That context helps explain why monetizing excess capacity would be an attractive path to recoup infrastructure spend.

Why does this matter?

Meta’s shift would recast a portion of its multibillion-dollar infrastructure bet into a commercial cloud play, turning data-center ownership into a potential revenue stream. If demand for AI compute stays strong, control of physical capacity may become as valuable as model performance, changing the competitive dynamics between model developers and cloud owners. Skeptics remain: some warn the build-out risks creating a bubble around rapidly depreciating chips, and others question whether AI services can generate enough end-user revenue to justify the scale of investment.

What does Meta already have in place?

Meta’s reported plan is anchored to very large commitments and active projects. The company has committed $182.9 billion to AI infrastructure spending and is running major projects in Louisiana and Ohio; the Ohio facility is expected to come online this year. The new business line would be organized under the reported Meta Compute initiative and led by Santosh Janardhan, Daniel Gross, and Dina Powell McCormick.

What to watch

Watch for formal announcements from Meta about a Meta Compute product, any pricing or service tiers it releases, and early customer contracts that mirror the SpaceX-Anthropic deal for Colossus 1. A concrete signal that the strategy is moving from planning to execution would be hosted-model listings or commercial compute leases announced under the Meta Compute name.