Meta Platforms is reportedly preparing to enter the cloud computing business. It would offer its unused AI computing power to outside customers. Bloomberg News first reported the plan, which Meta declined to confirm.
The move could mark a major shift in the company’s AI strategy. Meta wants new revenue from the billions it has poured into AI infrastructure. The effort falls under an internal program called Meta Compute. It reframes that heavy spending as a possible profit center.
The plan would let businesses and developers access Meta’s AI systems through the cloud. Customers would pay only for the computing resources they use. The project is still in development and could change before launch. It signals Meta’s ambition to compete beyond social media and advertising.
Meta Compute would reportedly offer two products. The first gives developers access to Meta’s own Muse Spark AI models. Meta would host and serve those models from its own infrastructure. That approach mirrors Amazon’s Bedrock service. The second offering would sell raw GPU computing capacity directly.
Three executives are said to lead the effort. They are infrastructure chief Santosh Janardhan, Daniel Gross of Meta Superintelligence Labs, and President Dina Powell McCormick. If launched, the service would rival AWS, Microsoft Azure, and Google Cloud.
Investors reacted strongly to the report. Meta shares rose more than 10%, easing a nearly 15% decline this year. However, neocloud rivals fell sharply on the news. CoreWeave and Nebius both dropped around 11% to 12%.
Those firms have deep ties to Meta as key suppliers. Nebius holds a reported $27 billion Meta deal, and CoreWeave a $21 billion agreement. Analysts said Meta may now need them less while competing directly. One analyst compared it to SpaceX renting out its data center capacity. If successful, Meta could shift from AI consumer to major infrastructure provider.
