Meta’s Cloud Push: Renting Out AI Compute to Rival Giants

Meta plans to rent out its unused AI computing power. The company has spent two years building massive AI infrastructure and now wants to monetize the excess capacity.

This effort is called Meta Compute, a new unit aiming to sell both access to AI models hosted on Meta’s hardware and raw computing power itself. The move would put Meta in direct competition with AWS, Google Cloud, and Microsoft Azure.

Meta’s AI infrastructure supports its core apps—Facebook, Instagram, WhatsApp—and ad products for its roughly three billion daily users. But the massive build-out leaves idle compute resources that Meta wants to turn into revenue rather than sunk cost.

Meta has not confirmed the plan or shared pricing and launch details. Sources say the strategy is still in flux. Meta’s stock rose over 10% after the report leaked, signaling Wall Street’s appetite for this pivot.

Meta’s capital spending is huge and growing. Actual expenditures hit $72.2 billion in 2025. Guidance for 2026 points to roughly $135 billion, an 87% jump. Much of this will go to chips, land, and power for AI infrastructure.

Meta owns a hyperscale campus in Louisiana and is building another data center in the Midwest. It also has external deals like a 1.6-gigawatt capacity agreement with Crusoe. This infrastructure build dwarfs many competitors.

While Meta’s AI ambitions grow, it faces limits. Google recently rationed Meta’s access to its Gemini models because of compute restrictions. That bottleneck likely feeds Meta’s desire to control more infrastructure directly.

The cloud market Meta aims for is crowded. Yet experts say Meta’s impact may hit smaller “neocloud” providers harder than the hyperscalers. Companies like CoreWeave rely on Meta’s capacity; Meta may cut them out.

SpaceX offers a cautionary tale. It rents spare capacity from Elon Musk’s xAI Memphis data center to Anthropic, potentially earning over $50 billion by 2028. Meta hopes to unlock similar revenue streams.

Meta’s recent earnings report shook investors. The stock dropped about 10%, wiping out $150 billion in market cap. The raised capital spending and heavy AI investments sparked doubts about returns.

Still, Meta’s CTO Andrew Bosworth claims progress at Meta Superintelligence Labs is fast. Their models are “very good” after just six months. This AI push demands more infrastructure and keeps spending on a steep climb.

The big question: can Meta turn its costly AI build into a profitable cloud business? It’s a gamble, but Meta has the scale and cash to try. The next year will show if the cloud can ease Meta’s AI spending woes or deepen the cost hole.