Nebius is an AI-focused cloud provider and the central business of Nebius Group N.V., an Amsterdam-headquartered company listed on Nasdaq. It sells GPU computing capacity together with storage, networking, managed services and software for building and running AI workloads. Customers pay for usage or reserve capacity through contracts; they are buying cloud infrastructure, not a chip or an AI model.
What Nebius is—and what the group includes
Nebius AI Cloud is the group’s core business: an integrated cloud platform aimed at AI developers and organizations, from startups to enterprises. Nebius describes its infrastructure and software as built in-house and optimized for AI workloads. Its offering spans model development and deployment through application management and inference.
Nebius Group is broader than the cloud business. Its other businesses include Avride and TripleTen, and it holds equity stakes in ClickHouse and Toloka. Those activities should not be conflated with Nebius AI Cloud’s revenue or operating results. The company’s 2025 annual report describes the group structure and its cloud operations.
How Nebius AI Cloud works
Compute, storage and networking
AI workloads need more than access to GPUs. Nebius offers GPU clusters alongside storage and networking, plus managed services and development tools. Together, these components let customers provision infrastructure for tasks such as training models and serving them to users. Nebius is therefore best understood as an infrastructure and software provider, not as a GPU manufacturer or an AI model vendor.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →#1 Best Overall
Usage-based and reserved capacity
Customers can obtain capacity on a pay-as-you-go basis or reserve it under fixed contracts. Usage-based access ties charges to cloud services consumed; reserved-capacity arrangements commit a customer to capacity over an agreed period and give Nebius a contracted basis for planning infrastructure. The terms and economics depend on each customer agreement; the public figures below do not establish a standard customer price.
What the reported growth figures show
Nebius AI Cloud revenue rose from $68.3 million in 2024 to $480.3 million in 2025, an increase of $412.0 million, or 603%, according to the company’s 2025 annual report. That is a historical segment figure, not a forecast.
For the quarter ended June 30, 2026, Nebius Group reported $582 million in total group revenue, up 454% year over year. This is a group-wide figure and should not be read as AI cloud revenue alone. In the same quarter, the AI cloud segment reported a 50% adjusted EBITDA margin. Adjusted EBITDA is a non-GAAP measure; it is not net income or cash flow. Nebius also reported that Token Factory production inference workloads increased more than threefold in the quarter. The company’s Q2 2026 results and shareholder materials provide the period and segment context.
How large contracts support the growth model
Customer contracts and infrastructure investment
Large reserved-capacity deals can provide contracted demand that supports data-center deployment and financing. In September 2025, Nebius announced a multi-year agreement to deliver dedicated capacity to Microsoft from its Vineland, New Jersey data center, with delivery expected to begin in late 2025. The announcement said cash flow from the deal and debt secured against the contract would help fund associated capital expenditure. It does not, by itself, establish current utilization, the contract’s value or its realized financial contribution. See the Microsoft agreement announcement.
Q2 2026 deal metrics
Nebius said four AI cloud deals signed in Q2 2026 averaged more than $1 billion in total contract value each and more than $20 million per megawatt in yield. The company also reported that 70% of those deals included prepayments, covering 50–60% of associated capital expenditure. These are company-reported metrics for a small set of deals, including future capacity and company estimates; they are not guaranteed unit economics or returns for the business as a whole.
NVIDIA partnership and planned capacity
In March 2026, Nebius and NVIDIA announced a strategic partnership covering AI factory design, inference software and models, infrastructure deployment and fleet management. NVIDIA announced a $2 billion investment. The companies also described a deployment ambition of more than 5 gigawatts by 2030; that is forward-looking, not capacity already delivered. The partnership announcement outlines the collaboration.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What can constrain the business
AI cloud is capital-intensive. Nebius identifies data-center operations, electricity and utilities, maintenance, personnel, and depreciation of servers and networking equipment among its costs. Growth therefore depends not only on customer demand, but also on whether the company can build facilities, secure power, finance equipment and deliver contracted capacity on schedule.
- Buildout and power: Data-center construction and access to electricity can limit how quickly planned capacity becomes usable.
- Financing and execution: Large contracts may support investment plans, but they also create delivery obligations and capital needs.
- Customer concentration: Large enterprise or hyperscaler deals can contribute substantially to demand, making results more exposed to individual customers and agreements.
- Competition and technology: Cloud pricing, available GPU generations, software capabilities and customer requirements can change, putting pressure on utilization or margins.
These risks matter when interpreting rapid growth: announced contracts, planned capacity and past segment results are different from delivered capacity and future earnings.
Efficiency figures need context
Nebius reported an average portfolio power usage effectiveness (PUE) of 1.25 for 2025 and cited a global industry average of 1.54. The figures appeared in the company’s July 2026 sustainability announcement and are company-reported; the comparison is not independently verified here. PUE compares total data-center energy use with energy used by IT equipment, so it is an indicator of facility energy overhead rather than a complete measure of environmental impact. See the 2025 sustainability report announcement.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
