Meta has poured billions of dollars into artificial intelligence and the data centers needed to run it. Now, according to Bloomberg, the company may be looking for a more direct way to earn money from that infrastructure: selling access to AI compute.
The reported plan would move Meta beyond using its AI capacity mainly for its own products and research. It would also place the company in a market already defined by major cloud providers such as Amazon Web Services, Google Cloud, and Microsoft Azure.
Meta Compute would turn infrastructure into a product
Bloomberg reported on Wednesday that Meta is developing plans for a cloud infrastructure business. The offering would sell access to both AI compute power and models, according to the report.
The initiative is reportedly called Meta Compute. It is led by head of infrastructure Santosh Janardhan, Meta Superintelligence Labs leader Daniel Gross, and president Dina Powell McCormick.
The basic idea is straightforward. Meta has built and committed to more AI infrastructure than its current public revenue lines appear to show. If some of that capacity is not immediately needed for Meta’s own work, the company could sell access to others that need compute.
Bloomberg reported that Meta may copy CoreWeave’s business model by selling access to “raw” compute capacity. It is also considering a model closer to AWS, where customers could access various AI models hosted on Meta’s AI infrastructure. That could include Muse Spark, Meta’s recently launched closed-weight model.
The move follows a similar SpaceX pattern
Meta is not the only company looking at excess AI compute as a business opportunity. The source article notes that SpaceX, through xAI, announced similar plans weeks earlier.
In early May, SpaceX signed a deal with Anthropic to buy out all of the compute capacity at SpaceX’s Colossus 1 data center. SpaceX has also signed similar leases with Google and Reflection AI.
That pattern matters because it shifts attention from AI models alone to the physical and technical systems that support them. If demand for compute remains strong, companies that control data centers may be able to monetize that position even when they are not selling the leading AI service directly to end users.
In that reading of the AI market, the advantage may not belong only to the companies with the most popular models or applications. It may also belong to those with enough infrastructure to sell scarce capacity to others.
The bet depends on demand staying strong
The opportunity comes with a major condition: demand for AI compute has to continue. The value of the strategy also depends on whether data centers keep their value over time.
The source article points to skepticism around the broader AI infrastructure buildout. Some skeptics warn that the current race may be creating a bubble, especially because it relies heavily on rapidly depreciating chips. Others question whether AI companies can earn enough revenue from end users to support trillion-dollar bets.
Those concerns have not stopped Meta from spending heavily. As of the end of the first quarter, Meta had committed to spending $182.9 billion on AI infrastructure in the coming years.
That spending includes massive ongoing projects in Louisiana and Ohio. The Ohio project, which Zuckerberg said would be the size of Manhattan, is expected to come online this year.
For Meta, selling compute could help answer a practical business question. If the company has already committed to enormous AI infrastructure spending, can some of that investment generate revenue before Meta’s own AI products become large standalone earners?
Meta’s AI revenue picture remains unclear
Unlike Google and OpenAI, Meta has not seen significant demand for its own AI models and services, according to the source article. Meta also does not break out revenue from Meta AI or from Llama, its open-weight AI model family, in its earnings.
That makes the financial picture harder to read from the outside. Public statements from executives have mostly emphasized the internal corporate uses of AI, rather than presenting Meta AI or Llama as clearly material standalone revenue lines.
A cloud infrastructure business would give Meta another path. Instead of relying only on consumer-facing or developer-facing AI products, Meta could sell the underlying capacity that other AI companies, model builders, or cloud customers may need.
This would not remove the risks around AI spending. It would, however, make the data center itself part of the commercial strategy. The company would not only be building infrastructure to support its own push toward AI “superintelligence”; it would also be exploring whether that infrastructure can become a market-facing product.
Zuckerberg had already left the door open
The Bloomberg report lines up with Zuckerberg’s May comments that a Meta cloud computing business is “definitely on the table.” The stated reason was to get a return on some of the massive investment behind Meta’s AI strategy.
TechCrunch said it reached out to Meta for comment. Based on the reported plans, the larger question is not only whether Meta can build powerful AI systems, but whether it can turn the expensive infrastructure behind them into a reliable business.
If Meta Compute moves ahead, it would add another large technology company to the contest for AI infrastructure customers. It would also make the AI race less about models alone and more about who owns the capacity needed to run them.