Meta is stepping up moves to run a cloud business using data centres it built with large capital outlays.

Bloomberg News reported on Tuesday that Meta is preparing a cloud infrastructure business that sells AI computing power together with access to models. It would compete with major cloud providers such as Amazon Web Services, Google Cloud and Microsoft Azure.

Meta is considering renting out computing capacity, like CoreWeave, to recoup heavy investment in its data centres. It is also weighing an option to provide access to various AI models that run on its infrastructure, including the closed model Muse Spark that it recently launched, in an AWS-style approach.

The project is being pursued under the name "Meta Compute" and is led by infrastructure chief Santosh Janardhan (산토시 자나르단), Meta Superintelligence Labs chief Daniel Gross (다니엘 그로스) and president Dina Powell McCormick (디나 파월 매코믹).

Meta's move is similar to SpaceX. SpaceX signed a contract with Anthropic in early May and sold all computing capacity at its Colossus1 data centre. It later signed similar leasing deals with Google and ReflectionAI.

TechCrunch said Meta adopting a similar tactic is a sign that the AI competition could shift from the ability to build the best models to how many data centres a company has. It added that this assumes computing demand stays in place and that data centre values do not fall.

Some sceptics also warn of a bubble driven by an AI infrastructure investment race that relies excessively on chips that depreciate quickly. Others question whether AI companies can generate enough end-user revenue to justify trillion-dollar investments.

Despite such concerns, Meta has not stopped investing in AI computing infrastructure. As of the end of the first quarter, Meta signalled it would spend $182.9 billion on AI infrastructure going forward. That includes large projects under way in Louisiana and Ohio. The Ohio project, which Zuckerberg said is the size of Manhattan, is expected to begin operating this year.

Unlike Google and OpenAI, Meta has not secured clear demand for its AI models and services. In earnings announcements, Meta does not separately disclose revenue from Meta AI or the open-weight model family Llama. Executives have also mostly highlighted internal AI use cases in public settings. This is seen as meaning Meta's AI business has not yet established itself as an independent source of revenue.