Meta Platforms is developing plans for a cloud infrastructure business that would sell outside customers access to its AI computing power and models, Bloomberg reported Wednesday — a move that would put the company into direct competition with Amazon Web Services, Microsoft Azure and Google Cloud for the first time. Meta shares jumped more than 10 percent on the report, their best day of the year.
The service under development would let developers access AI models hosted on Meta's infrastructure — including Muse Spark, the model the company unveiled in April but has not yet released to developers — and pay for the computing power needed to run them, an arrangement similar to Amazon's Bedrock. Meta is also considering selling raw computing capacity, the business model of so-called neoclouds such as CoreWeave and Nebius. The plans remain in development and could change; Meta declined to comment.
The economics behind the move are straightforward. Meta has projected between $125 billion and $145 billion in capital expenditure this year, most of it for AI data centers, part of an industry-wide buildout now exceeding $700 billion annually. Unlike Microsoft, Amazon and Google, Meta has never had a cloud business to turn that infrastructure into revenue — its capacity serves its own apps and research. Chief Executive Mark Zuckerberg signaled the shift at May's shareholder meeting, saying companies were approaching Meta "almost every week" to buy access to its models or spare computing power, and that entering cloud computing was "definitely on the table."
The market immediately sorted winners from losers. While Meta rallied, CoreWeave fell 10.8 percent and Nebius dropped 12.4 percent — investors reasoning that Meta both becomes a competitor and potentially reduces its own purchases from third-party capacity providers. Analyst Gil Luria noted that the impact of Meta's capacity hitting the market "is more likely to be on neoclouds than the big hyperscalers," whose customers buy integrated services rather than raw GPUs.
There is a more skeptical reading, too. Muse Spark has no announced launch date for developers, and some analysts see the cloud plan as an acknowledgment that Meta's frontier-model efforts have not kept pace with leading AI labs despite tens of billions in spending — making it more attractive to sell the shovels than to keep digging alone. The company's AI division has been through repeated reorganizations, including the 8,000-person restructuring announced in June.
Execution would not be trivial. Building an enterprise cloud business means sales teams, service guarantees, developer tooling and security certifications Meta has never needed. But the strategic logic — turning the industry's second-largest capex budget from a cost center into a product — is the same one that turned Amazon's spare retail infrastructure into AWS two decades ago. Investors, at least for one day, priced in the possibility that history rhymes.