Meta built enough AI infrastructure to run its own apps for years, and now Mark Zuckerberg is wondering whether other companies would pay to borrow some of that cloud capacity instead of buying it from Amazon or Microsoft.
Zuckerberg confirmed during a July 9 interview with Bloomberg that Meta is actively exploring a move into the AI cloud business, framing it as a natural next step given how much computing capacity the company has already built for itself. That capacity is not small. Meta plans to spend up to $145 billion on AI infrastructure in 2026 alone, money originally meant to power its own apps rather than to sell as cloud services to outside businesses.
This isn’t a brand-new idea for Zuckerberg. He first raised the possibility publicly at a May 27 shareholder meeting, saying entering the cloud computing market was “definitely on the table,” and noted that outside companies reach out almost every week asking to buy compute or API access directly from Meta’s servers.
Then Bloomberg reported on July 1 that Meta had already started building a dedicated cloud business unit to sell its surplus AI computing power and models to outside customers. The unit includes a system called Muse Spark, and aggressive pricing is reportedly part of the plan to undercut established cloud providers.
Investors responded quickly to the cloud ambitions. Meta’s stock jumped nearly 9% after that July 1 report, a sign the market sees real upside in Meta turning idle infrastructure into a competitive cloud revenue stream.
Stepping into the cloud market means going up against three companies that have spent over a decade building enterprise trust: Amazon Web Services, Microsoft Azure, and Google Cloud. Big Tech’s combined AI infrastructure spending is projected to top $700 billion in 2026, and each of Meta’s potential cloud rivals brings something Meta doesn’t have yet. Microsoft has its OpenAI partnership woven into Azure, Google pairs Gemini with its own custom chips, and AWS still holds the largest cloud market share and the deepest reach into enterprise customers.
Turning spare servers into a real cloud business takes more than available compute sitting idle. It requires sales teams, service-level agreements, security certifications, and a support operation built specifically for enterprise cloud customers, infrastructure Meta has never needed to run before, since none of its consumer products required convincing businesses to trust it with their cloud workloads.
