Nebius Group is projecting an annualized revenue run rate of $900 million to $1.1 billion by the end of 2025, escalating to between $7 billion and $9 billion by the close of 2026—a potential increase of up to 1,600% from current levels. These forecasts rest on two cornerstone agreements: a multi-year contract with Microsoft valued at more than $19 billion, and a five-year partnership with Meta Platforms totaling $3 billion. Together, these commitments underpin Nebius’s ability to meet the rapidly expanding compute requirements of large cloud providers, while offering a stable revenue base against which spot-market growth can accrue.

To support its forecasted demand, Nebius is accelerating its infrastructure build-out, expanding from 220 megawatts of installed power today to 800 megawatts by the end of 2026, with plans to reach a full gigawatt of connected capacity in the same timeframe. Contracted power already stands at 2.5 GW, giving the company significant headroom through partner pre-bookings. This expansion program is paired with a reported 20% efficiency advantage over rivals, enabling Nebius to optimize operating margins on high-density GPU clusters essential for training large AI models.

During his Davos address, Nvidia’s CEO highlighted soaring spot rental rates for both current-generation and two-generation-old GPUs, noting that rental prices have rebounded despite last year’s 60%–75% declines. This chip supply tightness—evidenced by orders for 2 million H200 units versus an available stock of 700,000—creates a favorable pricing environment for Nebius’s hourly GPU leasing business. Analysts estimate that sustained high spot rates could lift Nebius’s near-term revenue by 15% to 25%, potentially driving its annualized run rate above $4 billion within 12 months and amplifying margins through dynamic pricing during periods of constrained chip availability.