Nebius is positioning itself to monetize AI compute scarcity more rapidly than hyperscalers can internalize incremental capacity. The company has committed over $5 billion in capital expenditures since inception and is focused on converting that heavy CapEx into high-utilization megawatts at scale. By leasing whitespace in existing data centers and deploying modular, high-efficiency pods, Nebius claims it can bring new capacity online in as little as 12 weeks, compared with 24–36 months for traditional cloud providers. This time-to-capacity advantage has allowed the firm to secure contracts for over 200 MW of AI compute demand under multi-year agreements.

Nebius has built a differentiated software stack to manage heterogeneous AI workloads across its global footprint. Its open-source Soperator platform automates GPU orchestration, while the Papyrax framework optimizes data movement and model parallelism. These tools reduce customer lock-in and enable seamless integration with major AI frameworks such as PyTorch and TensorFlow. Strategic partnerships with Nvidia grant Nebius early access to next-generation H100 and forthcoming Blackwell GPUs, further reinforcing its competitive edge against proprietary solutions offered by rivals like CoreWeave.

Management forecasts a build-out to 2 GW of installed AI compute capacity by 2028, supporting an estimated $29 billion in annual revenue and $2–4 billion in net income at full utilization. These figures imply a revenue density of approximately $14.5 million per MW and an EBITDA margin north of 30 percent once operating leverage is achieved. With a current market capitalization of $22 billion, Nebius’s projected net income range suggests significant upside should execution track the plan and utilization rates exceed 80 percent.

The re-rating thesis for Nebius hinges on three core execution moats: accelerated time-to-capacity, pre-contracted demand and speed of deployment. To date, the company has pre-sold over 60 percent of its 2025 build-out, providing revenue visibility and de-risking large-scale rollouts. Its modular data center design enables parallel construction across six geographies, with the next phase targeting North America, Western Europe and Southeast Asia. By the end of this year, Nebius expects to add 400 MW of new capacity, locking in average contract tenors of five years and average pricing above $3 per GPU hour.