Nebius is an AI-focused cloud provider within Nebius Group N.V., an Amsterdam-headquartered company listed on Nasdaq. Its cloud combines GPU computing with storage, networking, managed services and software for developing, deploying and running AI workloads. Customers pay for usage or reserve capacity through contracts. The parent group also includes other businesses and investments, so its total results are not the same as Nebius AI Cloud’s results.
What is Nebius?
Nebius AI Cloud is the group’s central business: it sells infrastructure and software to organizations that build or operate AI systems. The company describes its hardware and software as built in-house and designed around AI workloads, rather than as a general-purpose cloud adapted for AI. It is a cloud infrastructure provider—not, on the evidence in its filings, a chip maker or an AI-model vendor.
Nebius Group N.V. is the parent company. Its other businesses include Avride and TripleTen, and it holds equity stakes in ClickHouse and Toloka. Those activities matter when reading group-level announcements and financial results: a group total should not automatically be attributed to the AI cloud segment. The company’s 2025 annual report describes its business structure and services.
What does Nebius AI Cloud provide?
The service is intended to support the AI workload lifecycle, from model development and training through deployment, application management and inference. The components work together: compute runs workloads, storage holds data and model assets, networking connects systems, and software and managed services help customers use and operate the infrastructure.
Free tools Windows power users keep installed
One-click scans. No signup required.
#1 Best Overall
- GPU compute: AI-optimized clusters provide processing capacity for demanding workloads.
- Storage and networking: Supporting infrastructure connects compute resources and provides access to data.
- Software and managed services: Tools and operational support help customers develop, deploy and manage AI workloads.
- Inference services: Nebius’s Token Factory is one part of its offering for production inference workloads.
The available company disclosures describe the product categories, but do not provide a like-for-like comparison with other cloud providers on GPU availability, regional coverage, networking performance, pricing or reliability.
How does Nebius make money?
Nebius earns revenue by providing cloud services under customer contracts. Customers can pay for capacity as they use it or make fixed arrangements to reserve capacity. The first approach ties spending more closely to usage; reserved-capacity deals give Nebius a contracted commitment around future infrastructure, while obliging it to deliver the agreed capacity.
Long-term contracts can support infrastructure deployment and financing. For example, Nebius announced a multi-year agreement to provide dedicated capacity to Microsoft from its Vineland, New Jersey data center, with delivery expected to begin in late 2025. The company said cash flow from the deal and debt secured against the contract would help fund associated capital expenditure. The announcement does not establish the facility’s current utilization, the contract’s value, or the realized financial contribution. See the Microsoft agreement announcement for the terms the company disclosed.
How large is the business, and how quickly is it growing?
Nebius AI cloud revenue increased from $68.3 million in 2024 to $480.3 million in 2025, a rise 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. That figure covers the group, not just AI cloud. The company also reported a 50% adjusted EBITDA margin for the AI cloud segment in the quarter. Adjusted EBITDA is a non-GAAP measure; it is not net income or cash flow. The Q2 2026 results and shareholder materials give the company’s quarter-specific figures.
The company also reported that production inference workloads in Token Factory increased more than threefold in Q2 2026. In the same quarter, four AI cloud deals each averaged more than $1 billion in total contract value and more than $20 million per megawatt in yield. Nebius said 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 estimates—not guaranteed returns or evidence that every contract will have the same economics.
What partnerships and expansion plans matter?
In March 2026, Nebius and NVIDIA announced a strategic partnership spanning AI factory design, inference software and models, infrastructure deployment and fleet management. NVIDIA also announced a $2 billion investment. The partners described a plan to deploy more than 5 gigawatts of capacity by 2030; that is a forward-looking ambition, not capacity already delivered. Details are in the partnership announcement.
Nebius has also reported an average portfolio power usage effectiveness (PUE) of 1.25 for 2025, compared with a global industry average of 1.54 cited by the company. PUE is a data-center energy-efficiency measure; a lower value generally indicates less overhead energy relative to the energy used by IT equipment. Both figures are from Nebius’s sustainability announcement and are company-reported, not independently verified by this comparison. See the 2025 Sustainability Report announcement.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What are the main risks and constraints?
AI cloud growth depends on building and operating expensive physical infrastructure. Nebius identifies data-center operations, electricity and utilities, maintenance, personnel, and depreciation of servers and networking equipment among its costs. That makes capital availability, construction timelines and access to power important constraints, even when customer demand is strong.
- Delivery and execution: The company must build facilities and bring capacity online on time to meet reserved-capacity and hyperscaler commitments.
- Power and operations: Data centers require dependable electricity, utilities, maintenance and skilled personnel.
- Financing and capital intensity: Expanding GPU clusters and data centers requires substantial capital. Contract cash flows or financing arrangements can help, but do not eliminate execution or funding risk.
- Customer concentration: Large contracts can anchor investment plans while increasing exposure to the performance and timing of a smaller number of customer commitments.
- Competition and technology shifts: Changing AI hardware, software and customer preferences can affect the value of infrastructure and put pressure on prices.
These risks are distinct from reported growth rates: historical revenue does not by itself establish that planned capacity will be delivered, that contract economics will recur, or that margins will persist.
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.
