Microsoft’s $17.4B Nebius Deal: Scaling AI Infrastructure
Microsoft and Nebius Group have announced a major AI infrastructure agreement worth up to $17.4 billion, with Microsoft also receiving an option to purchase an additional $2 billion in services through 2031.
The agreement gives Microsoft access to dedicated GPU computing capacity that Nebius plans to deploy at a new data center in Vineland, New Jersey.
The deal comes as major technology companies compete for increasingly scarce AI computing resources. For Microsoft, the agreement adds another source of GPU capacity to its expanding infrastructure portfolio. For Nebius, the contract provides a large, long-term customer for its AI cloud business and supports an accelerated expansion strategy.
💰 A $17.4 Billion AI Infrastructure Commitment #
Under the agreement, Nebius will build and deploy dedicated computing capacity for Microsoft.
The GPU infrastructure is scheduled to come online in multiple phases throughout 2025 and 2026, with the earliest services expected to become available in late 2025.
Each deployment phase is subject to Microsoft’s requirements for capacity, availability, and operational readiness.
Microsoft has not disclosed exactly which internal products or services will consume the Nebius capacity. However, the infrastructure is expected to provide additional GPU resources for Microsoft’s broader AI and cloud operations.
Following the announcement, Nebius shares rose sharply in after-hours trading, at one point gaining more than 60% before settling at a substantially higher level. The company’s market capitalization subsequently moved above $15 billion.
The market reaction reflects the scale and strategic importance of a multiyear infrastructure commitment from one of the world’s largest cloud companies.
🖥️ Microsoft’s AI Computing Crunch #
The deal comes against a backdrop of strong demand for AI computing capacity.
Microsoft CFO Amy Hood warned investors in July that the company expected capacity constraints to continue through the end of 2025.
The shortage has affected the pace at which Microsoft can expand its own AI products and provide additional AI services through Azure.
The underlying problem is not simply a lack of physical data centers. Modern AI infrastructure requires enormous quantities of specialized GPUs, high-bandwidth networking, advanced cooling, power generation, and supporting storage infrastructure.
Even after Microsoft committed billions of dollars to expanding its own infrastructure, demand for AI compute continued to exceed immediately available capacity.
A Multi-Vendor Strategy #
Microsoft has responded by expanding the number of infrastructure providers supplying its AI workloads.
Nebius joins a broader group of external partners that includes OpenAI, CoreWeave, and other infrastructure providers.
This approach gives Microsoft access to additional GPU capacity without relying exclusively on its internally owned data centers.
It also provides greater flexibility as Microsoft attempts to match infrastructure supply with rapidly changing AI demand.
🏢 Nebius’s AI Cloud Business #
Nebius is a relatively new name in the global cloud market, but its origins are tied to the much larger Yandex technology organization.
The company emerged from the restructuring of Yandex N.V. and is headquartered in Amsterdam.
Nebius focuses primarily on AI cloud infrastructure, with NVIDIA GPU-based computing forming a central part of its platform.
The company provides integrated infrastructure and software services designed for organizations developing and operating large-scale AI workloads.
Expansion in the United States #
Nebius has been expanding its presence in the United States, including offices in San Francisco, Dallas, and New York.
The company has said that increasing its U.S. presence is intended to bring its infrastructure and teams closer to American AI customers.
The expansion is particularly relevant as U.S.-based AI companies continue to require large amounts of GPU capacity for model training, inference, and other workloads.
🔗 Microsoft Is Diversifying Its AI Compute Supply #
The Nebius agreement illustrates an increasingly important feature of the AI infrastructure market: large technology companies are building relationships with specialized GPU cloud providers alongside their traditional hyperscale infrastructure.
For Microsoft, this creates another channel for securing computing capacity.
The company already has a major relationship with CoreWeave, while its partnership with OpenAI also involves significant infrastructure requirements.
Adding Nebius creates another source of capacity and potentially reduces dependence on any single external provider.
Why GPU Supply Matters #
AI infrastructure has several layers of supply constraints.
At the compute layer, demand for advanced GPUs remains high.
At the infrastructure layer, those GPUs require:
- High-bandwidth networking.
- Advanced power delivery.
- Large-scale cooling systems.
- High-capacity data centers.
- Storage infrastructure.
- Specialized software and orchestration.
Securing GPUs therefore requires more than simply purchasing accelerator cards. Providers must have the physical infrastructure and power capacity needed to operate them at scale.
The Nebius agreement addresses that broader infrastructure requirement by committing the provider to dedicated capacity for Microsoft.
📈 What the Deal Means for Nebius #
For Nebius, the Microsoft agreement provides a substantial long-term customer and a source of predictable infrastructure demand.
Founder and CEO Arkady Volozh said the economic benefits of the agreement were attractive while emphasizing its role in accelerating the company’s AI cloud business beyond 2026.
The company plans to use a combination of:
- Cash generated from the agreement.
- Contract-backed debt.
- Potential additional financing.
This financing structure could allow Nebius to expand its infrastructure more rapidly than originally planned.
The Microsoft contract therefore functions as both a revenue opportunity and a foundation for additional capital investment.
🌐 AI Infrastructure Is Becoming a Supply-Chain Business #
The agreement highlights a broader transformation in cloud computing.
Traditional cloud infrastructure was built around large centralized providers owning most of the required computing capacity. AI workloads are creating a more distributed infrastructure ecosystem in which hyperscalers, specialized GPU clouds, accelerator manufacturers, data-center operators, and AI model companies increasingly depend on one another.
Microsoft’s relationship with Nebius fits directly into this model.
Rather than waiting for every new data center and GPU cluster to be built internally, Microsoft can secure additional capacity through external providers while continuing to expand its own infrastructure.
For Nebius, meanwhile, a large customer such as Microsoft can provide the demand visibility needed to justify major investments in GPU clusters and data-center capacity.
🔍 A Major Deal Built Around Compute Capacity #
Microsoft’s agreement with Nebius is ultimately about securing access to AI computing resources at a time when demand is growing rapidly.
The $17.4 billion potential value, combined with the additional $2 billion service option through 2031, makes the agreement substantially more than a conventional cloud procurement contract.
For Microsoft, it adds another source of GPU infrastructure and supports its broader strategy of expanding AI capacity across Azure and its AI product ecosystem.
For Nebius, the contract provides a major customer and a financing foundation for accelerating its AI cloud expansion.
The deal also demonstrates how the AI infrastructure market is evolving: access to compute capacity has become a strategic resource, and long-term partnerships between hyperscalers and specialized GPU cloud providers are becoming an increasingly important part of the supply chain.
