Meta is preparing to enter one of enterprise technology’s most competitive markets by transforming its vast AI infrastructure into a commercial cloud business. This would allow the company to sell excess AI computing capacity and potentially offer developers access to AI models hosted on Meta’s own infrastructure, creating a new revenue stream while positioning the company against established hyperscalers.
According to multiple reports, Meta is evaluating two complementary offerings. One would provide developers with API-based access to AI models running on Meta-managed infrastructure, similar to managed AI model services already available from leading cloud providers. Customers would pay for AI usage while Meta would operate the underlying data centers and accelerator hardware.
The second option would involve selling raw AI computing capacity directly to customers, allowing enterprises and AI developers to rent GPU resources without necessarily using Meta’s own models. This is the business model adopted by specialized AI cloud providers that focus primarily on high-performance compute, like neo-cloud vendors CoreWeave and Nebius.
The initiative is being developed under an internal division known as Meta Compute, which oversees the company’s AI infrastructure strategy. The service could also provide access to Meta’s Muse Spark models, expanding the company’s AI portfolio beyond internal applications and consumer services.
“Meta has been a significant beneficiary from the advances in AI by selling more ads at higher prices which has driven significant revenue acceleration,” Gil Luria, Head of Technology Research at D.A. Davidson, told Techstrong.ai
“It has not gotten credit because it has increased capex even more. If Meta slows down capex and starts monetizing it, we see significant upside to revenue and cash flow. It could stay in the race by going back to its open source roots, which would present more upside.”
Diversified Revenue Stream
The strategy, long rumored, is a major evolution for Meta, whose revenue has depended on digital advertising. Building a cloud business would diversify the company’s income while creating an avenue to generate returns on the enormous capital expenditures required to compete in AI.
Meta has forecast capital expenditures of between $125 billion and $145 billion in 2026 as it invests in constructing AI data centers and buying next-gen GPUs. Like other frontier AI developers, the company has been rapidly expanding its computing capacity amid shortages of AI infrastructure.
CEO Mark Zuckerberg has previously acknowledged that selling compute capacity was under consideration. During Meta’s shareholder meeting earlier this year, he noted that companies regularly approach Meta seeking access to both AI APIs and spare computing resources. While Meta currently expects to utilize most of its infrastructure internally, Zuckerberg indicated that commercializing excess capacity remains an option if the company builds beyond its own requirements.
Investor reaction was enthusiastic. Meta shares climbed sharply following reports of the proposed cloud business, while shares of CoreWeave and Nebius declined as investors weighed the possibility of a powerful new competitor entering the market.
The impact may extend beyond specialized AI infrastructure providers. A commercial AI cloud platform would place Meta alongside AWS, Microsoft and Google, all of which have expanded managed AI services alongside traditional cloud computing.
But a key challenge faces Meta: building a successful cloud business requires more than data centers. Enterprise cloud platforms also depend on mature software ecosystems, customer support, security capabilities and large enterprise sales organizations developed over many years.
