A $19.4 Billion Bet on AI’s Next Frontier: Diversifying Power and Purpose

Nebius Group has finalized a landmark $17.4 billion agreement with Microsoft over five years to supply GPU infrastructure, with the potential to expand to $19.4 billion. The deal, announced in late September 2025, will initially see Microsoft gain access to dedicated GPU capacity at a new data center in Vineland, New Jersey, which Nebius plans to launch later in the year. The facility has an initial capacity of 300MW and the potential to expand further by 400MW. Additional GPU clusters will be deployed at facilities in Kansas City, Iceland, the United Kingdom, and Israel.

This strategic partnership is part of a broader shift in the AI infrastructure landscape, as major tech firms seek to diversify their GPU supply chains and reduce dependency on single providers. According to the agreement, Nebius will provide Microsoft with dedicated GPU resources, which are expected to accelerate the deployment of large AI models and reduce processing queue times. The contract’s scale is notable in the context of the growing demand for AI compute, as companies like Microsoft, Google, and OpenAI increasingly rely on custom or alternative infrastructure to support their AI ambitions.

Nebius, formed in 2023 from the non-Russian assets of Yandex after the Ukraine invasion, has been rapidly expanding its global data center footprint. The company previously announced plans to triple its Finland facility’s capacity to 75MW. With this new deal, Nebius is now positioned to significantly enhance its AI cloud business, which serves a range of clients from startups to enterprises. Arkady Volozh, Nebius’s founder and CEO, stated that the contract is a key step in the company’s growth strategy and will help accelerate the development of its AI offerings beyond 2026.

To support the agreement, Nebius plans to raise debt secured against the contract, signaling its confidence in the long-term value of the partnership. Microsoft, meanwhile, already sources GPU capacity from rival provider CoreWeave and is currently using 15,000 H100 GPUs to train its own AI models, including MAI-1-preview. The move to secure additional GPU resources through Nebius reflects a broader trend of tech giants building more resilient and diversified infrastructure to support their AI ambitions.

The deal has had an immediate impact on Nebius’s market valuation, with its shares rising by approximately 60 percent following the announcement. This reflects strong investor confidence in the company’s strategic direction and its ability to secure large-scale contracts with industry leaders. Analysts have noted that the contract underscores the growing importance of dedicated GPU capacity in the AI ecosystem, as demand continues to outstrip supply.

The collaboration between Microsoft and Nebius highlights the increasing competition among cloud and infrastructure providers to meet the surging global demand for AI processing power. As AI models grow larger and more complex, the need for robust, scalable infrastructure becomes a critical factor in maintaining performance and cost efficiency. This deal is expected to influence future infrastructure procurement strategies, particularly for companies seeking to avoid bottlenecks in their AI deployment timelines.