Nebius is scaling rapidly as an AI cloud infrastructure provider, with Q2 revenue up 454% year over year and a $3.0B annualized run-rate. The stock still earns a Hold as valuation, heavy capex, and financing needs offset the growth story.
Nebius Group N.V. (NBIS) has become a high-growth AI cloud company with strong operating momentum and an unusually large infrastructure ambition. In Q2 2026, group revenue reached $582.3M, up 454% year over year, while Nebius AI revenue reached $575M, or 98% of group revenue. Annualized run-rate revenue reached $3.0B, up 598% year over year, and adjusted EBITDA reached $236M at a 41% margin.
The investment case rests on three facts: management is selling capacity quickly, customers are signing large contracts, and the business is adding a software and asset-light layer around GPU infrastructure. Management reaffirmed 2026 revenue guidance of $3.0B to $3.4B, annualized run-rate revenue of $7B to $9B, adjusted EBITDA margin of approximately 40%, and capital expenditures of $20B to $25B.
The counterweight is valuation and financing intensity. NBIS carries a $73.1B market capitalization, a 68.5x forward P/E, and a 57.3x enterprise-value-to-revenue multiple. FY2025 debt reached $4.21B against $3.68B of cash, while 2025 free cash flow was negative $3.68B. For a moderate-risk investor with a medium-term horizon, the growth evidence supports continued ownership, but the valuation leaves little room for a slow capacity ramp. The recommendation is Hold.
Nebius Group is a Netherlands-based technology company that builds full-stack infrastructure for the global AI industry. Its core offering combines large GPU clusters, an AI cloud platform, developer tools, and services for training, fine-tuning, inference, and production workloads. The company had 1,543 employees and is led by founder and CEO Arkady Volozh.
The company changed its name from Yandex N.V. to Nebius Group N.V. in August 2024 after completing the sale of its Russian businesses. The April 30, 2026 20-F states that the divested businesses represented more than 95% of consolidated revenue and approximately 95% of assets and employees before the transaction. That makes the current NBIS story a strategic reset rather than a simple continuation of the former operating model.
Nebius also owns or operates adjacent businesses, including Avride, an autonomous-driving and delivery-robotics company, and TripleTen, an education technology platform focused on technology careers. Toloka became an equity investment after the company lost majority voting control, so its results are no longer consolidated. The current financial profile is therefore dominated by the AI cloud rather than a balanced portfolio of equally sized businesses.
Nebius AI is the economic engine. Q2 2026 Nebius AI revenue was $575M, representing 98% of group revenue, and the business generated adjusted EBITDA of $236M at a 50% margin according to management's quarterly commentary. Group profitability was lower because Avride and TripleTen remain early-stage investments.
The AI business uses three contract structures. Core AI cloud agreements generally run for one to three years. Short-duration capacity contracts can run for up to six months and carry premium pricing. Long-term investment-grade customer contracts help finance infrastructure expansion and create contracted cash flows that can support asset-backed borrowing.
Management said four Q2 landmark deals with Reflection, Cohere, a scaled U.S. AI lab, and a large U.S. quant-trading firm averaged more than $1B each. Midterm contracts were priced at yields of $20M to $25M per megawatt, while short-term capacity negotiations reached $40M to $50M per megawatt and sometimes higher. These figures show why allocation discipline matters: selling every megawatt under a long contract would create visibility, but retaining some capacity can capture higher short-term pricing.
Get AI research on any stock
Instant reports, daily intelligence, and an AI analyst in your pocket.
Nebius AI Cloud is the flagship product. It is designed to give AI developers access to GPU clusters, orchestration, storage, networking, managed Kubernetes, and production services through a single platform. The Q1 2025 investor presentation listed Blackwell early access, topology-aware training with Slurm, predictive failure notification, node auto-healing, managed MLflow, managed Postgres, JupyterLab, container registry, encryption, monitoring, and software development kits.
Token Factory extends the platform into inference and model-serving economics. Management said the service provides day-one support for frontier open models and cited a GLM 5.2 implementation that achieved a 100% quality score and leading performance in independent benchmarking. The strategic value is straightforward: infrastructure revenue creates scale, while inference tools can increase customer usage and add higher-margin revenue above raw GPU rental.
The product thesis depends on keeping pace with rapidly changing hardware and models. Nebius has highlighted support for NVIDIA GB300 systems, HGX B300, and Vera Rubin NVL72 systems. In this market, a platform that is technically elegant but late to the newest hardware is simply an expensive museum exhibit.
Nebius is building an advantage around flexibility rather than a single proprietary chip or model. Its multi-tenant platform can operate across Nebius-owned, colocated, and partner infrastructure. That design supports a broader capacity pool and allows management to shift between long-term contracts, premium short-term demand, and software-led services.
The asset-light partnership model is the most important structural innovation. Partners finance, build, and operate facilities, while Nebius supplies the platform and customer demand. Management described the resulting revenue as high margin with minimal balance-sheet capital. If the model works at scale, it can reduce the amount of equity and corporate debt required for each additional block of capacity.
The first capacity auction also provides evidence of pricing power. Management said the auction cleared at a price 15% above the highest price previously charged for Blackwell capacity. That result does not establish a permanent price floor, but it demonstrates that Nebius can use market-based allocation when demand exceeds immediately deployable supply.
Customer validation adds another layer. Nebius said the four landmark Q2 wins were competitive victories against existing suppliers, including hyperscalers, and that customers cited scale, performance, reliability, responsiveness, transparency, and white-glove support. Those attributes are practical advantages in an infrastructure market where a delayed cluster can delay a customer's product launch.
Nebius is pursuing a power-first operating strategy. Management raised its year-end 2026 contracted-power target to 5 gigawatts and said almost all of that capacity should come online over the following three to three and a half years. The company also said it can build more than 1 gigawatt of new capacity per year beginning in 2027.
The footprint includes owned and colocated sites across the United States and Europe, with projects in locations such as New Jersey, Kansas City, Iceland, Finland, the United Kingdom, Pennsylvania, and Spain. At the Vineland, New Jersey site, management said the building had been completed, engineering fit-out was progressing, and a switch to Bloom fuel cells was expected to have no significant effect on the project timeline.
Capital intensity is substantial. Q2 2026 capital expenditures were approximately $5.7B, primarily for GPUs, GPU-related hardware, and data-center expansion. Full-year capital-expenditure guidance is $20B to $25B. The quarterly cash-flow statement reported operating cash flow of $2.25B against capital expenditures of $5.66B, producing negative free cash flow of $3.41B.
Customer financing helps offset that burden. Approximately 70% of Q2 deals included upfront prepayments, and management expects more than $9B of customer prepayments during 2026. Nebius also completed a $775M asset-backed debt facility priced at SOFR plus 250 basis points and tied to deployed GPU infrastructure and contracted cash flows. The company cited more than $40B of additional customer commitments as a potential foundation for similar financing.
Nebius operates in a market where demand is moving from model training toward a broader mix of training, inference, and production workloads. An investor presentation cited an AI infrastructure market of approximately $33B in 2023 expanding to more than $260B by 2030. The same presentation projected inference to represent 64% of AI server end-user spending by 2027, compared with 34% in 2023.
Gartner forecast worldwide spending on AI models and platforms of $64B in 2026, up 63.4% year over year. Gartner also identified data-center systems and infrastructure-as-a-service as among the fastest-growing areas of IT spending. These figures support the growth runway for Nebius, while the company's $3.0B to $3.4B 2026 revenue guidance shows that even strong execution would represent a small share of the broader opportunity.
The market is also becoming more demanding. Gartner's 2026 compute outlook identifies agentic AI as a force reshaping compute infrastructure, while the AI platform forecast emphasizes cost control, latency, performance, reliability, and usage transparency. Nebius's Token Factory, managed services, and full-stack platform address those requirements more directly than a bare-metal GPU provider.
Like what you're reading?
Get full access to AI-powered research reports, market analysis, and portfolio tools.
Nebius targets AI-native companies, agentic AI leaders, new AI laboratories, quant-trading firms, and sophisticated enterprises. These customers need contiguous clusters, rapid deployment, technical support, and flexibility across training and inference. The Q2 landmark deals with Reflection, Cohere, a U.S. AI lab, and a U.S. quant-trading firm demonstrate that the customer base extends beyond a single startup category.
Large enterprise relationships are becoming more important. Nebius has disclosed capacity arrangements involving Microsoft and Meta, and its Q2 management commentary cited more than $40B of contracted backlog with investment-grade customers. A five-year Meta agreement announced in March 2026 included $12B of dedicated capacity and a commitment to purchase additional available compute capacity of up to $15B over five years, with delivery beginning in early 2027.
Prepayment behavior improves customer quality from a financing perspective. Management said 70% of Q2 deals included upfront payments, and the 20-F states that a substantial portion of revenue is collected on a prepaid basis. The tradeoff is customer concentration: the same 20-F identifies concentration of credit, customers, and suppliers as primary risk areas.
Nebius names CoreWeave, Crusoe, and Lambda as direct AI infrastructure competitors. CoreWeave offers infrastructure services, managed software services, and application software services for AI workloads. Lambda focuses on AI-only cloud infrastructure and single-tenant GPU clusters. Crusoe competes in AI infrastructure development and capacity deployment.
The larger competitive group includes Amazon Web Services, Google Cloud, Microsoft Azure, and Oracle. These hyperscalers have greater balance sheets, broader enterprise distribution, and more extensive product suites. Nebius counters with a narrower AI-first proposition, early access to NVIDIA platforms, large contiguous clusters, and a full-stack platform designed specifically for AI workloads.
Nebius's position is strongest when a customer values time-to-capacity, specialized support, and open infrastructure. Its position is weaker when a customer wants a broad enterprise cloud relationship or can accept slower deployment in exchange for the hyperscalers' global scale. The Q2 deals won from existing suppliers show that specialization has commercial value, but the competitive landscape still gives customers substantial choice.
Power availability is the central macro constraint for Nebius. The company's 5-gigawatt contracted-power target, its use of grid and behind-the-meter generation, and its Bloom partnership all show that energy access is as important as GPU procurement. Data-center approvals add another layer of execution risk, illustrated by the Vineland public hearing that adjourned without a vote.
Geographic diversification reduces dependence on a single site. Nebius operates across the United States, the European Union, and Israel, and management said cloud contracts provide flexibility to deliver within a region rather than relying on one location. The company's 2024 exit from Russia also materially changed its geographic and operating profile.
The supply chain remains tied to NVIDIA hardware and data-center construction. Nebius has a strategic alliance with NVIDIA and has highlighted early deployment of Blackwell and Vera Rubin systems. That relationship supports product relevance, but the $20B to $25B capital-expenditure plan means GPU pricing, delivery timing, financing costs, and site approvals can all affect returns.
FY2025 debt climbed to $4.21B against $3.68B of cash, while free cash flow was negative $3.68B, underscoring how capital-intensive the buildout remains.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →Q2 2026 revenue jumped 454% year over year to $582.3M, with adjusted EBITDA reaching $236M at a 41% margin.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →Management reaffirmed 2026 revenue guidance of $3.0B to $3.4B and annualized run-rate revenue of $7B to $9B, with adjusted EBITDA margin near 40%.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →NBIS trades at a 68.5x forward P/E and 57.3x enterprise-value-to-revenue multiple, a premium that assumes the capacity ramp stays on schedule.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →The report's fair value estimate is $276.46, which sits between the $240 Buy level and the $320 Sell level.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →Nebius has moved beyond the early proof-of-concept stage. Q2 2026 revenue of $582.3M, a $3.0B annualized run rate, $236M of adjusted EBITDA, four landmark AI cloud deals, and a 5-gigawatt contracted-power target show a company operating at a much larger scale than its 2025 revenue base.
The investment challenge is that infrastructure growth arrives before infrastructure returns. Negative free cash flow, large capital expenditures, reliance on customer prepayments, and an EPS forecast that remains negative through 2029 create real risk even alongside strong demand. NBIS is a credible AI infrastructure growth story, but the stock needs sustained execution to justify its premium. At $276.46, the evidence supports patience rather than aggressive accumulation.
Get AI-powered research reports, daily market intelligence, and a personal analyst in your pocket.
Get Full Access →Not ready to subscribe? ·
Cancel anytime
NBIS is growing fast, but its $5.0 billion convertible offering and $2.8 billion share sale have made financing the stock's immediate risk. The AI demand is real; the dilution burden is arriving faster than the profits.
Nebius Group N.V. (NBIS) jumps after reporting Q2 revenue above estimates, lifting the stock in after-hours trading. The move was also helped by strength across AI infrastructure names, but investors are still weighing a rich valuation, high volatility, and ambitious growth targets.
Nebius is rapidly scaling an AI-native cloud platform, with Q1 2026 revenue up 684% and adjusted EBITDA turning positive. The stock looks compelling for investors who can handle heavy capital intensity and valuation risk.