Meta is planning to enter the cloud computing market to sell AI computing capacity to external clients, competing directly with Amazon Web Services, Microsoft Azure, Google Cloud, and IBM. The news was positively received by markets, with Meta’s stock gaining over 8% at the open, while shares of Nebius fell 14% and CoreWeave fell 11%. According to Bloomberg, citing people familiar with the matter, the internal project is named Meta Compute and aims to monetize unused computing power from Meta’s massive infrastructure investments in chips and data centers. The initiative is managed by Santosh Janardhan, Meta’s head of infrastructure, along with Daniel Gross, head of Meta Superintelligence Labs, and Meta president Dina Powell McCormick. Meta is considering two options: one plan involves selling access to various AI models hosted on Meta’s existing AI infrastructure, similar to AWS’s Bedrock model, where Meta would manage data centers and chips powering AI models including its own Muse Spark models and charge for access. A second possibility is granting access to raw computing capacity, modeled on CoreWeave. A Meta spokesperson declined to comment. Meta has made developing superintelligence a top priority and raised its capital expenditure for 2026 to 145 billion dollars, focusing on AI infrastructure. Entering the cloud sector is seen as a way to recoup part of that investment amid enormous demand for data center computing capacity.