Meta Plans Cloud Business to Sell Its 'Excess' AI Computing Power

Meta is developing plans for a cloud infrastructure business that would sell access to its AI computing power and models, according to a Bloomberg report — a move that would open a new front of competition with Amazon Web Services, Microsoft Azure and Google Cloud, and that sent ripples through AI supply-chain stocks across Asia.

At the center of the effort is an internal organization called Meta Compute, created to oversee the buildout and operation of the company's vast AI infrastructure. The unit is led by Santosh Janardhan, Meta's head of infrastructure, alongside Daniel Gross, a senior leader inside Meta Superintelligence Labs, and Meta President Dina Powell McCormick.

Two paths to market

Meta is reportedly weighing two service models. The first would sell developers access to AI models hosted on Meta's own infrastructure — including its closed-weight Muse Spark model — in an arrangement similar to Amazon's Bedrock service. The second would rent out raw computing capacity, the way neocloud providers such as CoreWeave do.

The rationale is straightforward: Meta is spending an estimated $125–145 billion on AI infrastructure in 2026, up from roughly $72 billion last year — likely more capacity than its own products currently need. Turning that surplus into revenue would help offset one of the largest capital expenditure programs in corporate history.

Markets react — in both directions

Meta's stock jumped more than 6% in premarket trading on July 1 after the report surfaced. But the news landed very differently in Asia, where investors read it as a signal that hyperscale AI spending could become more disciplined. South Korea's KOSPI slid as Samsung Electronics and SK Hynix fell sharply, before rebounding later in the week.

The plan echoes a broader industry shift: SpaceX has floated selling orbital compute, and Google, Microsoft and Amazon have long monetized their infrastructure overbuild. If Meta follows through, the last major holdout among AI hyperscalers will have joined the cloud wars — with hundreds of thousands of GPUs as its opening bid.