Meta Platforms has committed up to $27 billion over five years to Dutch AI cloud provider Nebius Group for dedicated and flexible compute capacity, the largest external infrastructure procurement deal a hyperscaler has disclosed this cycle.
The agreement is structured in two parts. Nebius will provide $12 billion in dedicated capacity across multiple locations, including what is expected to be one of the first large-scale rollouts of NVIDIA's next-generation Vera Rubin chips. Meta also committed to purchase up to $15 billion in additional available capacity on a flexible basis.
The deal reflects a structural shift in how hyperscalers are approaching AI infrastructure. Despite Meta's own AI capital expenditure guidance of $115 billion to $135 billion for 2026, the company is simultaneously securing substantial long-term capacity from a third-party provider. That combination is a departure from how hyperscalers have traditionally viewed infrastructure: not build versus buy, but both at scale. It signals that even companies with aggressive internal build programs cannot provision AI compute quickly enough through owned infrastructure alone.
Nebius, founded in 2022 after a restructuring of Yandex's operations outside Russia, listed on Nasdaq in 2024 and has since emerged as one of the more significant European pure-play AI cloud providers. Its stock has risen more than 400 percent since its IPO and jumped another 14 percent on news of the Meta agreement.
The timing is notable. NVIDIA invested $2 billion in Nebius the week before, giving the chip maker a direct stake in one of its largest hardware deployers. The Meta commitment follows a pattern in which Nebius has been accumulating institutional backing from companies that also supply or consume its infrastructure. Nebius signed a comparable deal with Microsoft in September 2025, covering up to $19.4 billion in compute over five years.
The inclusion of Vera Rubin capacity carries its own signal. NVIDIA's Vera Rubin architecture, the successor to the Blackwell line, has not yet seen broad commercial deployment. A hyperscaler-scale commitment through Nebius would represent one of the first anchor deployments of that generation, with implications for how quickly the next GPU cycle reaches production scale.
Meta is part of a group of hyperscalers including Amazon, Alphabet, and Microsoft that analysts project will collectively spend around $700 billion on AI infrastructure this year. Nebius CEO Arkady Volozh described the deal as part of a strategy to "secure large, long-term capacity contracts" as the company accelerates its AI cloud build-out.
–
By the Control Plane Editorial Team