Per a Bloomberg report, the social media company is preparing a cloud business to sell AI compute and its own AI models to outside customers, challenging Amazon and Google's dominant cloud services and giving Wall Street a path to monetize its massive AI spending.
Meta is preparing to rent out the same AI computing power and resell the same AI models it built for itself, turning a capital-expenditure liability that has spooked investors into a potential enterprise revenue stream that would directly challenge Amazon, Microsoft and Google on their own cloud turf. Shares of the social media company jumped more than 6% on Wednesday after Bloomberg reported the plan. The reaction suggests Wall Street is reading Meta's next chapter as infrastructure rather than apps.
The frame matters more than the headline number. Meta has poured hundreds of billions of dollars into data centres and AI chips in pursuit of what CEO Mark Zuckerberg calls "superintelligence," a bet that has looked like pure cost on the income statement and has fuelled investor concern about AI capex returns. Per a Yahoo Finance summary of the Bloomberg report, the proposed service would sell AI compute capacity to outside companies and resell Meta's internally designed AI models to business customers, putting the company in direct competition with Amazon Web Services, Microsoft Azure and Google Cloud.
This is infrastructure-surplus monetization, not a new product. TechCrunch frames the move as analogous to SpaceX turning Starlink and launch capacity built for internal use into external revenue, the same pattern playing out at xAI's Memphis data centre, which has begun renting capacity to Anthropic and striking deals with Google. Meta is making the same play with AI compute: capacity it sized for its own AI and ads workloads is being recast as inventory the rest of the industry can rent. Reuters, via TradingView, confirmed the Bloomberg reporting.
Zuckerberg had already opened the door. On Meta's May earnings call he told shareholders that selling excess compute or launching a paid business-customer AI service was "definitely on the table," remarks the market had read at the time as a hedge against AI-spend fatigue. The cloud plan would convert that hedge into a product. Meta has existing major compute deals with CoreWeave, Google and Oracle that already help it flex capacity; a first-party cloud business would let it capture more of the markup and turn those relationships from cost management into revenue.
The competitive stakes are real, but the bar is high. AWS, Azure and Google Cloud have years of enterprise contracts, region coverage, compliance certifications and tooling depth that Meta would have to build out from a standing start. The market is treating the plan as upside because AI capex returns remain well below break-even. If Meta can monetize even a fraction of the compute capacity it has built, a cloud business could add tens of billions in annual revenue and convert a feared cost line into a recurring stream.
The execution risk sits on the AI models, not the data centres. Critics and investors have questioned whether Meta can compete with OpenAI, Anthropic and Google at the model frontier, and whether its AI infrastructure bet will pay off at all. If enterprise buyers reach the same conclusion, the surplus-capacity story collapses back into a capex problem. Wall Street is betting Zuckerberg can turn a feared pile of compute into a product line. The next test is whether enterprise customers agree.