Meta Launches Meta Compute to Sell Excess AI Capacity

Meta Platforms is establishing a new business unit called Meta Compute to monetize its massive investments in artificial intelligence infrastructure. The initiative aims to sell raw GPU computing power and provide cloud access to proprietary models, such as Muse Spark, to external customers. This strategic shift allows Meta Platforms to diversify revenue beyond advertising and offset capital expenditures projected to reach $145 billion in 2026.

Led by executives Santosh Janardhan, Daniel Gross, and Dina Powell McCormick, the venture positions Meta as a direct competitor to established hyperscalers including Amazon Web Services, Microsoft Azure, and Google Cloud. It also targets the "neocloud" sector, competing with providers like CoreWeave and Nebius Group. Mark Zuckerberg had previously indicated during a May shareholder call that selling excess capacity was an option if the company overbuilt its infrastructure.

The announcement triggered immediate market volatility. Meta shares rose over 10% as investors reacted positively to the new revenue stream. Conversely, shares of CoreWeave and Nebius Group plummeted by as much as 17% due to fears that Meta would transition from a major client to a competitor. CoreWeave and Nebius currently hold multi-billion dollar contracts with Meta, but analysts warn that Meta's entry into the rental market challenges the industry narrative that AI compute remains scarce.

Meta's infrastructure expansion includes the Hyperion data center campus in Louisiana. While the company continues to purchase millions of chips from Nvidia, AMD, and Google, the Meta Compute unit seeks to ensure that idle servers do not waste cash or energy.