July 2, 2026, (Inside AI) — Meta Platforms is quietly building a cloud service to rent out its idle AI computing power, according to a Bloomberg report. The internal program, called Meta Compute, aims to turn the company's massive infrastructure spending into a new revenue stream. Meta declined to confirm the plan, but the move could reshape how the social media giant makes money.
The initiative would let outside businesses and developers tap into Meta's AI systems on a pay-as-you-go basis. It signals a strategic pivot beyond advertising, directly challenging cloud incumbents like AWS, Microsoft Azure, and Google Cloud. The project remains in development and may evolve before any launch.
Inside Meta Compute: Two-Pronged Cloud Play
Meta Compute reportedly features two distinct offerings. The first gives developers access to Meta's proprietary Muse Spark AI models, hosted and served on its own infrastructure. This mirrors Amazon's Bedrock service, which provides managed access to foundation models. The second product sells raw GPU computing capacity directly, targeting companies that need high-performance hardware for training or inference.
Three senior leaders are steering the effort: infrastructure chief Santosh Janardhan, Daniel Gross of Meta Superintelligence Labs, and President Dina Powell McCormick. Their involvement underscores the project's seriousness. If launched, it would mark the first time Meta opens its vast AI infrastructure to external customers, potentially monetizing billions in sunk costs.
Meta has invested heavily in AI data centers, with capital expenditures soaring in recent years. The company's 2025 annual report revealed over $30 billion in AI-related infrastructure spending. Much of that capacity sits underutilized outside peak training periods, creating a financial drag. Renting it out could improve margins and offset future investment risks.
Market Shockwaves and Rival Reactions
Investors applauded the news, sending Meta shares up more than 10% on Monday. The rally trimmed a nearly 15% year-to-date decline, reflecting hopes for a new profit engine. But the announcement hammered neocloud competitors heavily dependent on Meta's business. CoreWeave and Nebius both plunged about 11% to 12%, as the market priced in direct competition.
Those firms have deep supplier relationships with Meta. Nebius reportedly holds a $27 billion deal with the tech giant, while CoreWeave's agreement is valued at $21 billion. Analysts warn that Meta may now need them less, turning a key customer into a rival. One industry observer likened the situation to SpaceX renting out its data center capacity—a move that would disrupt existing cloud dynamics.
Meta's entry could pressure pricing across the AI cloud market. Its scale and existing infrastructure give it a cost advantage that smaller providers cannot match. However, the company faces hurdles in enterprise sales, support, and reliability—areas where established clouds have decades of experience. Success will depend on execution and whether businesses trust Meta with sensitive workloads.
Beyond financials, the move reflects a broader industry trend of AI infrastructure commoditization. As more companies build massive GPU clusters, the line between consumer and provider blurs. Meta's pivot from AI consumer to infrastructure supplier could accelerate this shift, forcing rivals to differentiate on software and services rather than raw compute.
