Nvidia has deepened its role in the global artificial intelligence buildout with a roughly $2 billion investment in AI cloud company Nebius, a move that highlights how the next phase of AI competition is increasingly about physical infrastructure rather than software alone. The investment, structured through pre-funded warrants, gives Nvidia a minority stake while helping finance a large-scale expansion of AI computing capacity.
At the center of the announcement is an ambitious target: Nebius expects its AI infrastructure platform to exceed 5 gigawatts of Nvidia-accelerated computing capacity by 2030. That figure stands out because it places AI infrastructure in a category closer to major industrial systems than conventional cloud growth. Gigawatt-scale deployments imply not only vast numbers of GPUs, but also enormous demands for electricity, cooling, land, and supply-chain coordination.
A Strategic Bet on the Next AI Cloud Layer
Nebius is one of the companies increasingly described as part of the “neocloud” segment, meaning providers built specifically for AI workloads rather than adapted from legacy enterprise cloud environments. Based in Amsterdam and listed on Nasdaq under the ticker NBIS, the company has positioned itself as an AI-native infrastructure operator spanning the United States, Europe, and Israel.
The company’s corporate history is also notable. Nebius emerged in 2024 after the Dutch parent of Yandex divested its Russian businesses for about $5.4 billion. The remaining international operations were reorganized into Nebius, which has since focused on building infrastructure for training large AI models, supporting inference workloads, and serving the rising demand associated with agentic AI systems.
For Nvidia, the logic is straightforward. By backing Nebius, the chipmaker is not only making a financial investment; it is also supporting a cloud operator likely to consume substantial volumes of Nvidia hardware for years to come. The collaboration spans the full AI stack, including data center design, inference deployment, and GPU fleet management.
From H100 and H200 to Future Nvidia Platforms
Nebius already operates large GPU clusters using Nvidia hardware, including the H100 and H200 accelerators. Under the expanded partnership, the company also plans to adopt future Nvidia systems such as the Rubin architecture, Vera CPUs, and BlueField storage platforms. That broadens the relationship beyond a simple chip supply arrangement and turns it into a deeper ecosystem alignment.
Nvidia founder and CEO Jensen Huang described the partnership as part of a new inflection point in AI, arguing that agentic AI is driving another surge in compute demand and accelerating the buildout of underlying infrastructure. Nebius CEO Arkady Volozh framed the company as an AI cloud provider built for this moment from the ground up, rather than a legacy cloud player repurposing existing systems.
Together, those statements reflect a broader industry reality: AI demand is no longer constrained only by model innovation. It is increasingly constrained by who can secure power, deploy advanced semiconductors, and scale physical infrastructure quickly enough to meet demand.
Missouri Project Signals the Scale of Expansion
The scale Nebius is discussing is not theoretical. Just days before Nvidia’s investment was announced, the company received approval for a 1.2-gigawatt AI data center campus in Independence, Missouri. The site is described as one of the largest planned AI infrastructure developments in the United States.
According to the disclosed figures, the Missouri project is expected to create roughly 1,200 construction jobs and generate an estimated $650 million in economic impact over two decades. While those local development benefits are significant, the project’s broader importance lies in what it reveals about the trajectory of AI infrastructure: hyperscale buildouts are beginning to resemble industrial policy, energy planning, and logistics strategy all at once.
The Missouri campus is only one part of a larger expansion roadmap. Nebius has said it could operate up to 16 global data center sites by the end of 2026, with contracted power capacity approaching 3 gigawatts within the next year. Those numbers suggest the company is moving rapidly to secure physical capacity before demand outpaces available infrastructure.
Capex Is Soaring Faster Than Revenue
Nebius’s spending profile shows just how capital-intensive this strategy has become. The company reported about $2.1 billion in capital expenditures in the fourth quarter of 2025 alone. For 2026, projected spending is expected to range from $16 billion to $20 billion, an extraordinary figure for a company still in the process of scaling its revenue base.
Revenue remains comparatively modest for now. Nebius reported roughly $530 million over the trailing twelve months. However, management has pointed to long-term contracts and backlog as evidence that growth could accelerate as new facilities come online and more AI workloads migrate into production.
Among the company’s disclosed agreements is a multi-year deal with Microsoft tied to a data center project in Vineland, New Jersey, estimated at between $17 billion and $19.4 billion. Nebius has also cited a separate $3 billion arrangement with Meta Platforms. These contracts help explain why investors may be willing to tolerate heavy upfront spending: the company is attempting to build ahead of expected demand, not react after shortages appear.
Market Reaction and the Infrastructure Thesis
Investors responded positively to the Nvidia announcement. Nebius shares rose between 13% and 16% following the news, lifting the company’s market capitalization above $24 billion during early trading. The reaction reinforced the idea that an investment from Nvidia often serves as a powerful market signal, particularly in sectors closely tied to AI infrastructure.
More importantly, the transaction underlines a central theme in the current AI cycle: the bottleneck is shifting. Algorithms still matter, but at this scale, the decisive constraints are increasingly power availability, cooling systems, advanced chips, and the speed of data center construction. In other words, AI may be expressed through software, but it is being enabled by industrial capacity.
Nvidia’s investment in Nebius therefore carries significance beyond the size of the check. It suggests that the company sees long-term value in expanding the broader ecosystem around its hardware, especially with demand from training, inference, and agentic AI continuing to rise. For Nebius, the deal provides both funding and validation at a time when competition in AI infrastructure is intensifying.
If the next decade of AI is defined by who can build the backbone fastest, this partnership indicates that both companies intend to compete not just at the level of chips and models, but at the level of energy, geography, and scale.