Meta builds cloud business to sell excess AI capacity

Meta has spent years telling investors that its massive AI infrastructure buildout would eventually pay off. Now it’s putting a business plan behind that promise.

Bloomberg reported on July 1 that Meta Platforms is developing a cloud services unit called Meta Compute, designed to sell off excess AI computing capacity and provide customers with hosted access to its AI models. The initiative is being led by infrastructure head Santosh Janardhan, Superintelligence Labs leader Daniel Gross, and president Dina Powell McCormick.

Mark Zuckerberg had telegraphed this move back in May, noting that monetizing excess capacity externally was a strong possibility. Meta’s market responded accordingly: the stock jumped over 6% in pre-market trading following the Bloomberg report.

What Meta Compute actually is

Think of it in two parts. The first is raw compute capacity, the kind that specialized providers like CoreWeave sell to companies that need GPU access without building their own data centers. The second is hosted AI model access, similar to how AWS Bedrock lets developers tap into foundation models through an API rather than running them locally.

The timing makes sense given the scale of what Meta is spending. The company projects capital expenditures of $125 billion to $145 billion on data centers and AI hardware in 2026 alone. Spreading that cost across external customers who pay for access is not just smart business, it is almost financially necessary to justify the number to shareholders.

Meta has also reportedly signed a $21 billion infrastructure deal with CoreWeave running through 2032, alongside a contract potentially worth up to $27 billion with Nebius for AI infrastructure. The company is both buying compute from third parties and building its own capacity at the same time, which reflects how tight the GPU market remains.

Why this matters beyond Meta’s stock price

The cloud infrastructure market is currently dominated by three players: AWS, Google Cloud, and Microsoft Azure. Meta’s competitive advantage, if it has one, is that it already runs one of the largest AI workloads on the planet across its apps.

The more direct competitive threat is probably to specialized AI compute providers like CoreWeave, which have built their business on exactly the kind of GPU-on-demand model that Meta Compute is now entering. This is notable given that Meta is simultaneously one of CoreWeave’s largest customers under that $21 billion deal.

For investors, Meta Compute represents a potential new revenue line that would reduce the company’s dependence on advertising, which currently funds essentially everything. AI infrastructure spending has reportedly more than doubled year-over-year at Meta, and that spending has weighed on margins even as the core business remains highly profitable. A cloud unit that generates meaningful revenue from that infrastructure changes the math on those capital expenditures from pure cost to partial investment with return potential.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.