What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Nebius Group N.V. is an Amsterdam-headquartered, Nasdaq-listed AI cloud company. Its main operating business, Nebius AI Cloud, sells access to large-scale computing infrastructure and a software platform used to train, fine-tune, deploy and run artificial-intelligence models. Customers range from AI-native startups to large enterprises. Nebius earns money primarily by selling compute capacity and related platform services, while large multiyear capacity contracts and a newer partner-owned data-center model provide additional routes to revenue.
The most important qualification is that several numbers associated with Nebius describe different things. Reported revenue is money recognized for services delivered; annualized run-rate revenue (ARR) extrapolates one recent month; backlog and contract values represent commitments over time; and contracted, connected and active power describe different stages of infrastructure readiness.
What Nebius Group does
Nebius Group is the listed corporate entity. Nebius AI Cloud is the central operating business inside that group. The cloud is designed as a full-stack environment for the AI lifecycle rather than merely a warehouse of servers: customers can obtain capacity for data processing, model training, fine-tuning, inference and production deployment, together with software and services intended to help them move between those stages.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsIn practical terms, a customer does not buy a graphics card from Nebius. It buys access to computing resources and a managed technology layer. Nebius may provide that capacity from a facility it owns, from colocation space, or—under the partner model announced in July 2026—from infrastructure owned and operated by another company. The customer-facing service remains cloud compute and platform functionality.
#1 Best Overall
Nebius says its target users include developers, startups and enterprises building AI products, agents and other services. Publicly announced large counterparties include Microsoft and Meta, while the company describes its broader customer base as spanning AI-native businesses through large enterprises.
How Nebius AI Cloud makes money
Compute capacity
The primary commercial product is access to AI compute. Customers pay to use installed infrastructure for workloads such as training a model, fine-tuning an existing model, generating outputs through inference, or running an AI application in production. The economic value is the usable capacity and the ability to obtain it when needed, not the electricity, rack space or physical hardware in isolation.
Platform software and services
Nebius presents its offer as more than raw GPU rental. Its software and service stack supports the workflow around the hardware, including the tools and operational layer needed to put models into production. That full-stack positioning can let Nebius sell a broader service relationship than a wholesale provider that supplies bare-metal machines alone. The available company disclosures do not provide a quantified split between compute charges, software fees and other service revenue.
Free tools Windows power users keep installed
One-click scans. No signup required.
Committed and dedicated capacity
Large customers can reserve capacity through multiyear agreements. A dedicated-capacity arrangement gives the buyer an allocation over an agreed period and gives Nebius a contractual demand signal that can support data-center and equipment investment. Revenue is recognized as the contracted services are delivered under the agreement; the headline value of a contract is not revenue already earned.
Partner-enabled economics
In July 2026, Nebius announced a model in which infrastructure partners finance, own and operate facilities and hardware. Nebius supplies the systems architecture, software and services stack, maintains that stack, and sells capacity through its global sales organization. The company said possible economics include revenue sharing, licensing fees, commissions and arrangements involving committed capacity. It disclosed that initial arrangements had been entered into, but did not quantify their revenue contribution or profitability.
Chief executive Arkady Volozh described the approach as an “asset-light model” that could give infrastructure partners access to a wider customer base and better margins than conventional wholesale bare-metal contracts. That is management’s characterization, not an independently measured margin comparison.
What the reported numbers mean
Nebius reported $399.0 million of group revenue for the first quarter of 2026. Nebius AI Cloud contributed $389.7 million, or approximately 98% of group revenue in that quarter. The percentage applies to the reported Q1 2026 period; it should not automatically be used as a permanent description of every future quarter or as proof that every group activity is part of the AI cloud segment.
Recommended Free Tools
| Measure | Reported figure or definition | How to interpret it |
|---|---|---|
| Q1 2026 group revenue | $399.0 million | Revenue recognized by the group during that quarter, according to Nebius. |
| Q1 2026 Nebius AI Cloud revenue | $389.7 million | AI-cloud revenue recognized during the quarter. |
| AI Cloud share of Q1 group revenue | Approximately 98% | A period-specific share calculated from the company’s Q1 figures. |
| 2025 year-over-year revenue growth | 479% | Company-reported historical growth for 2025, not a forecast. |
| Cash reported in the March 2026 update | $3.7 billion | Cash disclosed by the company at that update; it is not revenue or profit. |
| End-2026 ARR target | $7 billion to $9 billion | Forward-looking management guidance, not realized 2026 revenue. |
ARR is a run-rate, not annual revenue
Nebius defines annualized run-rate revenue by taking AI cloud revenue from the final month of a quarter and multiplying it by 12. If the final month produces an unusually high or low result, the annualized figure will reflect that month’s pace. ARR therefore indicates the scale of a recent recurring run rate; it does not mean Nebius has already earned that amount during the year.
Backlog is not recognized revenue
A March 2026 company update described more than $20 billion of contract backlog. Backlog represents contracted or committed future business under the company’s stated definition. It becomes recognized revenue as services are provided and accounting requirements are met, so it should not be added to current-year revenue.
How the AI cloud operating model works
- Secure power and facilities. Nebius assembles data-center capacity through owned sites and colocations. Its newer partner model adds facilities financed and operated by infrastructure partners.
- Equip the sites. The company deploys AI infrastructure, including large-scale GPU capacity, and connects that equipment to the systems needed to deliver cloud services.
- Expose usable capacity. Nebius makes the infrastructure available through its cloud platform so customers can run training, fine-tuning, inference and production workloads.
- Sell service and platform value. Customers pay for access to capacity and the software and operational services around it. The exact pricing mix by workload or customer is not disclosed in the available materials.
- Allocate capacity to major buyers. Dedicated or additional-capacity agreements can reserve resources for large customers over multiple years.
- Expand through partners. Where a partner owns the facility, Nebius can provide the architecture and software stack and take capacity to market without owning every underlying asset.
Contracted, connected and active power: three different capacity numbers
Nebius reports capacity in stages that answer different questions. Combining them into one headline number can make the business look further along—or smaller—than it actually is.
| Term | Meaning | Reader’s question answered |
|---|---|---|
| Contracted power | Power secured through agreements for future or planned capacity. | How much power has Nebius arranged access to? |
| Connected power | Power connected into built and equipped data centers. | How much capacity is physically connected to facilities? |
| Active power | Power consumed by installed, operational IT equipment available to generate revenue. | How much capacity is currently usable by customers? |
Nebius reported more than 3.5 gigawatts of contracted power, with more than 75% owned, and targeted 800 megawatts to 1 gigawatt of connected power by the end of 2026. Those figures are not interchangeable: contracted power can precede construction, connected power can precede full equipment deployment, and active power is the closest of the three to immediately revenue-generating capacity.
Major customer commitments and what they imply
Microsoft dedicated capacity agreement
Nebius announced a dedicated-capacity agreement with Microsoft in September 2025. The existence of the agreement demonstrates demand from a major technology customer and a route to planning capacity over a longer horizon. The announcement does not make the agreement equivalent to revenue recognized in 2025 or 2026.
Meta five-year agreement
In March 2026, Nebius announced a five-year agreement with Meta for $12 billion of dedicated capacity and up to $15 billion of additional compute purchases. Nebius stated a potential total contract value of approximately $27 billion. The $15 billion portion is described as additional possible purchases, so the full amount is a potential value over five years rather than guaranteed current-period sales. Neither the $12 billion nor the approximately $27 billion should be presented as money already earned.
These commitments can help Nebius justify infrastructure expansion and give investors visibility into potential demand. They also create execution requirements: facilities, power, equipment and service quality must be delivered on schedule, and customer usage must develop under the contract terms.
How the partner-owned infrastructure model changes the economics
Traditional ownership requires a cloud provider to fund or lease facilities, buy equipment and carry more of the construction and operating burden. Nebius’s announced partner model shifts facility and hardware ownership to partners. Nebius contributes the architecture, software, maintenance and customer distribution.
This can reduce the capital Nebius must commit per unit of marketed capacity and can make expansion faster where qualified partners already control suitable sites and power. It also means Nebius may not receive the same economics as it would from selling capacity from wholly owned infrastructure. Revenue shares, licenses, commissions and committed-capacity structures can produce different revenue timing, margins and risk profiles. The company has not published a quantified breakdown, so the profitability of the model cannot be inferred from the announcement alone.
Who pays Nebius and for what?
- AI startups and developers: access to scalable capacity for experimentation, training, fine-tuning and launching applications.
- AI-native businesses: infrastructure and platform services for products whose core workloads depend on model execution.
- Large enterprises: managed capacity and production support for internal or customer-facing AI services.
- Large technology companies: dedicated or additional compute capacity arranged through multiyear commitments.
The customer mix matters because on-demand usage, reserved capacity and partner-mediated sales can have different utilization and margin characteristics. Public disclosures identify customer categories and major counterparties, but do not provide a complete revenue table by customer, workload, geography or pricing mechanism.
How to evaluate Nebius’s growth claims
The company reported 479% year-over-year revenue growth for 2025, more than $20 billion of backlog in its March 2026 update, and a $7 billion-to-$9 billion ARR target for the end of 2026. These figures point to rapid expansion, but they are not the same type of evidence.
- Revenue growth is historical performance based on recognized revenue.
- Backlog is a measure of contracted future business and depends on delivery over time.
- ARR guidance is management’s forward-looking target based on its run-rate definition.
A careful reader should check subsequent earnings releases for actual revenue, the composition and remaining term of backlog, progress from contracted to connected and active power, cash spending, and any change in guidance. Management descriptions of attractive partner margins and future demand are useful context, but they are not independent audits.
What Nebius does not disclose publicly in the available material
The available company announcements do not establish a peer-normalized unit cost for GPU capacity, a complete price list, a quantified revenue split among compute, software, licensing, commissions and revenue sharing, or the profitability of the initial partner arrangements. They also do not independently verify how much of every announced contract has been utilized or recognized as revenue.
That limits conclusions about margins and customer concentration. A large contract can improve visibility while still requiring substantial spending before capacity becomes active. Likewise, a high ARR can coexist with different cash-flow and accounting outcomes depending on equipment purchases, colocation costs, power costs and partner terms.
How Nebius differs from a hardware seller
Nebius’s customer relationship is a cloud-service relationship. The buyer receives compute and platform access; Nebius decides whether the underlying capacity comes from owned equipment, colocation or partner-owned infrastructure. This distinction explains why comparing Nebius with a server or GPU retailer on units shipped would miss the company’s business model.
For comparisons with another AI cloud provider, use five consistent axes:
- Service breadth across raw GPU access, managed software, training, fine-tuning, inference and deployment.
- Capacity status—contracted, connected and active—and the timetable for bringing additional capacity online.
- Infrastructure ownership, including owned sites, colocations and partner facilities.
- Customer and contract mix, separating on-demand usage from concentrated multiyear commitments.
- Reported revenue and cost economics, keeping recognized revenue separate from ARR, backlog and management targets.
Bottom line for readers and investors
Nebius Group is building and selling AI cloud infrastructure and the software layer around it. In Q1 2026, that business generated nearly all group revenue, and the company is pursuing rapid capacity expansion backed by large customer commitments. Its revenue engine is straightforward—customers pay for AI compute and platform services—but the path from demand to reported revenue depends on construction, equipment deployment, utilization, contract terms and the accounting timing of service delivery.
The partner model could let Nebius market more capacity with less direct ownership of data centers and hardware, but its revenue mix and profitability remain unquantified. The clearest way to understand future results is to track recognized AI-cloud revenue alongside ARR methodology, backlog conversion, active power and the share of capacity supplied through each infrastructure model.
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.
