Summary
Meta has signed a new long-term AI infrastructure agreement with Amsterdam-based Nebius Group that could be worth up to $27 billion over five years, marking one of the biggest external compute commitments yet in the artificial-intelligence buildout race. The base portion of the contract calls for $12 billion in dedicated capacity across multiple locations, with delivery set to begin in early 2027.
The structure of the deal matters. Beyond the initial $12 billion, Meta also committed to buy up to $15 billion in additional available compute capacity tied to certain upcoming Nebius clusters over the same five-year period, but only if that capacity is not sold to other customers first. That means the headline figure is not a guaranteed $27 billion from day one; it is a combination of firm committed capacity plus a large conditional backstop on future supply.
This is also not a routine cloud rental agreement. Nebius said the deployment will be based on one of the first large-scale rollouts of NVIDIA’s Vera Rubin platform, linking the Meta contract directly to the next generation of AI infrastructure now coming into view. That detail is significant because the industry is no longer competing only on model quality or product features. It is increasingly competing on access to power, data-center space and advanced GPU systems at scale.
The agreement deepens a relationship that was already moving quickly. Nebius had previously disclosed an earlier approximately $3 billion, five-year AI infrastructure agreement with Meta in November 2025. It also announced a major Microsoft AI infrastructure deal in September 2025, underscoring how aggressively large tech buyers are locking in outside compute rather than relying only on their own internal buildouts.
For Nebius, the Meta expansion lands just days after another major vote of confidence: NVIDIA said on March 11 it would invest $2 billion in Nebius, and the company disclosed that the stake would amount to 8.3%. NVIDIA said the partnership is meant to help Nebius scale a full-stack AI cloud platform and support deployment of more than 5 gigawatts of NVIDIA systems by the end of 2030, though that broader capacity buildout remains a forward-looking target rather than near-term delivered infrastructure.
The broader takeaway is straightforward: the AI infrastructure market is tightening around whoever can secure chips, power and physical capacity first. That is why companies like Meta are increasingly turning to specialized AI-cloud providers instead of waiting for every new megawatt to come from their own campuses. Nebius sits inside that growing layer of AI-first infrastructure firms that sell high-performance computing capacity while trying to grow into major cloud platforms in their own right.
Investors reacted quickly. Nebius shares were trading at $112.95 on Monday, according to market data, after news of the agreement. The company said its 2026 guidance remains unchanged, which suggests management is framing the Meta deal less as an immediate revenue shock and more as a long-duration foundation for future scale. In practical terms, that may be the clearest signal in the whole announcement: this is about locking in the next phase of the AI compute economy before the next supply crunch hits.
Correction note: This report may be updated if either company discloses additional terms, deployment timelines or regulatory filings related to the agreement.
