Meta reportedly plans a cloud service to sell excess AI computing capacity

Meta Platforms is exploring a cloud business aimed at selling spare artificial intelligence computing capacity to outside customers, according to a report carried by Yahoo Finance and attributed to Bloomberg.

The concept centers on what counts as “computing capacity,” the processing power used to train and run artificial intelligence models. For companies building large AI systems, the costs of data centers and specialized hardware can be heavy, and turning that infrastructure into a revenue stream is one way to offset fixed costs.

Meta has largely used its advanced AI infrastructure internally, the report said, which would make selling access to developers a notable shift in strategy. If the plan moves forward, the company would be taking a step from operating as an AI customer and user to also becoming a supplier in the infrastructure layer of the AI economy.

The market reaction described in the TradingView write-up was swift. Meta shares jumped about 8.8% to $613.34 in the session in which the Bloomberg report circulated, while competitors focused on AI cloud infrastructure saw declines. CoreWeave shares fell about 14%, and Nebius dropped about 17%, reflecting investor concern about potential competition.

TradingView also reported that Meta already works with both of those companies. It said Meta expanded its CoreWeave partnership through 2032 in April and signed a long-term infrastructure agreement with Nebius starting in 2027. That context matters because it suggests Meta’s reported initiative would not be starting from zero, even if the new business would still introduce a different commercial dynamic.

In the broader sector context, the AI compute market has become a battleground, with traditional cloud providers and specialized “neocloud” infrastructure firms competing to offer faster deployment, capacity guarantees, and access to models and tools. A move by Meta to commercialize excess capacity could increase supply and intensify pricing pressure for providers that rely on demand from other AI builders.

Still, much of the plan remains unconfirmed in public statements from Meta. The reporting described exploration of a new cloud offering, but it did not lay out timelines, pricing, the exact product scope (for example, whether customers would buy raw capacity, hosted services, or both), or how existing partnerships would be structured.

For now, investors and developers will likely watch for indicates from Meta’s leadership, product teams, or investor communications, including any mention of a formal launch, customer pilots, or changes in the company’s data-center build-out strategy. The next update may also come from competitors’ guidance, as they assess whether demand for third-party AI compute could be diluted by a new in-house supplier entering the market.