Meta's Cloud Ambitions Turn Its Biggest Liability Into a Revenue Engine — But Can Advertising's Giant Really Compete With AWS?

Shares surged nearly +9.7% on July 1 after Bloomberg reported that Meta is building a cloud business to sell excess AI computing power to outside customers — a move that reframes the company's massive, controversial infrastructure spending as a potential profit center rather than a pure cost drain.

A $130 Billion Spending Spree Finally Gets a Business Case. Meta raised its 2026 capital budget to between $125 billion and $145 billion, close to double what it spent the year before. Until today, investors had no clear path to recoup that outlay beyond better ad targeting. Now Meta is forming a business to sell access to AI computing power and models to outside customers , including both hosted AI model access and raw computing capacity. The move could reduce Meta's reliance on advertising revenue and help it take on major cloud companies. That narrative shift alone added roughly $50 per share in a single session.

Google's Refusal Proved the Demand Is Real. Google imposed strict usage limits after Meta requested more computing capacity than Google could deliver in March 2026.

CEO Sundar Pichai acknowledged: "We are compute-constrained in the near term."

Wedbush Securities called Google's rationing "the latest sign that compute demand continues" to outpace supply. If the world's biggest cloud providers can't fill orders, Meta selling spare capacity looks less speculative and more like plugging a genuine market gap.

Neocloud Competitors Are Already Bleeding. CoreWeave and Nebius shares tumbled over 6% Wednesday on the news. Meta's infrastructure — spanning up to 6GW of AMD chips , millions of NVIDIA GPUs , and a multi-year Broadcom partnership supporting its custom AI chips through 2029 — dwarfs what smaller cloud upstarts can assemble. If Meta prices aggressively, it could squeeze margins across the sector while generating incremental revenue on infrastructure it has already committed to building.

The Risk: Execution in an Unfamiliar Business. The company's plans are still in development, and it's possible the strategy could change. Meta has never run a commercial cloud operation, has no enterprise sales force for it, and the market has not been rewarding cloud businesses lately — hyperscalers recently traded at their lowest forward valuation since the launch of ChatGPT. Selling "excess" compute also implicitly signals Meta may have overbuilt, a concern Zuckerberg himself acknowledged: "If we get to a point where we feel that we have overbuilt, then that is an option that we have." Shareholders are betting the upside of a new revenue stream outweighs that admission.