Meta pivots to selling AI compute power as it seeks returns on infrastructure spending

Meta has invested billions into building out artificial intelligence capabilities and the data center infrastructure that underpins them. Now the company is considering how to recoup some of those costs by offering access to that computing power to other businesses.

According to reporting from Bloomberg, Meta is developing what it calls "Meta Compute," a new business unit designed to rent AI processing capacity and proprietary AI models to external customers. The initiative would position Meta in direct competition with established cloud providers including Amazon Web Services, Google Cloud, and Microsoft Azure.

A trend among AI infrastructure leaders#

The strategic pivot mirrors a similar move by SpaceX, which announced comparable plans earlier this year. SpaceX's AI subsidiary xAI has already signed deals to lease computing capacity at its Colossus 1 data center to companies including Anthropic, Google, and Reflection AI. The parallel approaches from both organizations suggest an emerging consensus: controlling computing hardware may ultimately prove more valuable than building the AI applications themselves.

Investment and scope#

Meta's commitment to computing infrastructure is substantial. The company has pledged $182.9 billion toward AI infrastructure development over coming years, with major projects underway in Louisiana and Ohio. The Ohio facility, which Mark Zuckerberg said would be roughly the size of Manhattan, is expected to launch later this year.

According to Bloomberg, Meta is contemplating two revenue models for the new service. One would follow CoreWeave's approach by selling raw computational capacity to developers. The other would mimic Amazon's cloud strategy of packaging and selling access to AI models—potentially including Meta's recently released closed-weight model, Muse Spark, as well as its open-source Llama family.

The underlying bet#

Meta's move underscores a strategic wager: that ownership of computing infrastructure will generate returns even if Meta's own AI products don't capture dominant market share. Unlike Google and OpenAI, Meta has not disclosed significant commercial demand for its proprietary AI offerings. The company does not separately report revenue from Meta AI services or Llama, and executives have primarily highlighted internal applications of AI technology.