Mark Zuckerberg told Meta Platforms’ annual shareholder meeting on Wednesday that leasing compute capacity to outside customers is “definitely on the table,” the clearest signal yet that the company is willing to convert its AI infrastructure into a fifth hyperscaler business if its spending outruns internal demand.

Of the four major U.S. hyperscalers, Meta is the only one that doesn’t sell cloud infrastructure externally. Zuckerberg said outside companies approach Meta “almost every week” asking about API services or raw compute. The reason Meta hasn’t said yes, he explained, is that its own AI workloads keep absorbing whatever capacity it builds. The framing matters: overbuilding becomes a feature, not a risk.

“Obviously if we get to a point where we feel that we have overbuilt, then that is an option that we have, and that is partially what gives us confidence in investing in building this out,” Zuckerberg said.

That confidence carries a price tag. On April 29, Meta raised its 2026 capex guidance to $125 billion to $145 billion, up from $115 billion to $135 billion, citing higher component prices and additional data centre costs for future capacity. The company spent $72.2 billion on capex in 2025. Shares fell after the April release on AI spending concerns.

The cloud option functions as a hedge against that anxiety. If Meta’s own products can’t monetise the buildout fast enough, AWS and Azure customers will. Meta also said Wednesday it’ll begin testing paid Meta AI subscriptions at $7.99 and $19.99 a month in Singapore, Guatemala, and Bolivia, the first time it has charged consumers for AI features, while a longer-term monetisation model for business AI services on WhatsApp is in development.

Every revenue surface is being tested at once. That’s what a capex line approaching twice last year’s looks like in practice.

Sources