Behind the US$27bn AI Infrastructure Deal by Meta & Nebius

Meta and Nebius have signed a multi-billion dollar infrastructure agreement that signals a significant expansion in enterprise AI capabilities as demand for hyperscale computing continues to grow.

The deal, valued at up to US$27bn, focuses on large-scale data centre capacity specifically designed to support advanced AI workloads across multiple sites.

Under the five-year supply arrangement, Nebius is providing US$12bn in dedicated compute infrastructure built on the NVIDIA Vera Rubin platform.

This infrastructure enables rapid scaling of AI cloud services, offering the type of high-performance computing environments increasingly required for training and deploying sophisticated AI models.

Meta has committed US$15bn to purchase upcoming Nebius clusters, with Nebius intending to sell capacity primarily to third-party customers of its AI cloud business while Meta acquires the remaining available capacity.

Arkady Volozh, Founder and CEO of Nebius, says: "We are pleased to expand our significant partnership with Meta as part of securing more large, long-term capacity contracts to accelerate the build-out and growth of our core AI cloud business. We will continue to deliver."

Infrastructure built for AI

The partnership centres on hyperscale infrastructure designed to support large-scale AI operations.

Nebius is delivering dedicated compute capacity across several locations, with deployments based on the NVIDIA Vera Rubin platform.

These systems represent next-generation accelerated computing infrastructure tailored for AI cloud environments.

The Vera Rubin NVL72 model used for these operations is built on the third-generation NVIDIA MGX NVL72 rack design.

According to Nebius, the platforms will benefit from support provided by over 80 MGX ecosystem partners.

Their deployment across Nebius facilities could reflect growing demand for high-density computing environments capable of handling the computational requirements of large language models, machine learning training and other resource-intensive AI applications.

For Meta, the agreement secures long-term access to dedicated infrastructure whilst allowing Nebius to expand its AI cloud platform footprint.

Deliveries are scheduled to begin from 2027 as new capacity comes online across Nebius clusters.

Capacity allocation and commercial structure

Beyond the initial US$12bn commitment, Meta has also agreed to purchase additional available compute capacity from upcoming Nebius clusters.

The arrangement allows Nebius to maintain a broader cloud customer base while securing a guaranteed buyer for unused capacity.

According to Nebius, the company intends to sell cluster capacity primarily to third-party customers using its AI cloud services.

The remaining capacity across those clusters will then be purchased by Meta under the terms of the agreement.

This combined structure brings the potential contract value to approximately US$27 bn across the five-year term and could reflect the scale of investment now shaping global AI infrastructure development.

Long-term AI infrastructure expansion

The agreement forms part of Nebius's wider effort to expand its AI cloud infrastructure through long-term capacity contracts with major technology companies.

By securing large commitments from hyperscale customers, the company could scale new data centre clusters and deploy next-generation computing platforms designed specifically for AI workloads.

Nebius has confirmed that its guidance for 2026 remains unchanged as the infrastructure rollout and capacity delivery timeline begins from 2027, when the first dedicated capacity from the agreement is expected to be deployed.

The deal could further strengthen the partnership between the two companies as Nebius continues expanding its global infrastructure footprint, potentially offering enterprises greater access to the computing resources required for advancing AI capabilities.