Meta is getting into the cloud business. As CNBC and TechCrunch reported on July 1, the company is building a new division called “Meta Compute.” The idea: sell excess AI computing capacity to external customers — similar to what SpaceX does with its Colossus data center.
What Meta is planning
Meta Compute is led by infrastructure chief Santosh Janardhan, Meta Superintelligence Labs leader Daniel Gross, and president Dina Powell McCormick. It’s still unclear whether Meta will sell raw compute only or also offer hosted AI models. Either way, it’s a direct challenge to AWS, Google Cloud, and Microsoft Azure.
The numbers
By the end of Q1 2026, Meta had committed $182.9 billion to AI infrastructure — including massive data centers in Louisiana and Ohio. Not all of that capacity is needed internally. Rather than letting it sit idle, Meta wants it to generate revenue.
Market reaction
Meta’s stock jumped nearly 9% on the day of the news. The counter-move from neocloud providers was brutal: CoreWeave and Nebius each dropped about 12%. When a company with Meta’s resources enters your market, things get tight.
My take
Meta has a luxury problem: too much GPU capacity, too little return from its own AI agents (Zuckerberg admitted that same day that agent development is behind schedule). Cloud rental is a pragmatic way out. But whether Meta can hold its own as a cloud provider — against AWS, Google, and Microsoft — remains an open question. Those three have decades of enterprise customer experience. Meta’s experience is in advertising. These are different worlds.
Sources: