N/A
OverallTickerSpark AI RatingN/A
Nebius Group N.V. (NBIS) is one of the market’s clearest pure-play ways to invest in AI infrastructure, but it is not a simple growth story. The company finished 2025 with revenue of $529.8M, up 479% YoY, and Q4 revenue of $227.7M, up 547% YoY. Its core AI cloud business generated $214.2M of Q4 revenue, or about 94% of group revenue, and produced $51.8M of adjusted EBITDA at a 24% margin. Management also said annualized run-rate revenue reached $1.2B in December 2025, above prior guidance, while 2026 revenue guidance stands at $3.0B to $3.4B with group adjusted EBITDA margin targeted at about 40%.
That is the bull case in plain English: Nebius has real demand, real large-customer validation, and a product stack built for AI workloads rather than generic cloud plumbing. Management said the company sold out capacity in Q3 and Q4 2025 and was already sold out in early 2026. It also said average contract duration for new cloud customers grew 50%, Meta capacity was fully deployed in early February 2026, and Microsoft’s first tranche was delivered on time in November 2025. In this market, capacity is destiny, and Nebius is racing to secure it.
The bear case is just as real. Nebius plans $16B to $20B of 2026 CapEx after spending about $4.066B on property and equipment in 2025. EBIT remains negative, EBITDA is only recently positive at the group level, and the valuation already prices in a large amount of future success. Core valuation data shows a market cap of about $45.48B, trailing P/E of 4477.75, forward P/E of 68.49, and EV/revenue of 87.29. That is a premium multiple even by AI infrastructure standards, and it leaves little room for execution mistakes.
For a balanced, moderate-risk investor with a medium-term horizon, NBIS looks like a high-quality but high-volatility growth name. The business momentum is strong enough to justify constructive coverage, but the stock is best treated as a selective buy on pullbacks rather than a chase-at-any-price compounder. The central judgment in this report is straightforward: Nebius has the ingredients to grow into a much larger AI infrastructure platform, but the current valuation already assumes a lot of that future arrives on schedule.
Nebius Group N.V. is a technology company building full-stack infrastructure for the global AI industry. The company operates in the United States, the United Kingdom, and internationally, and its core offer includes large-scale GPU clusters, cloud platforms, and tools and services for developers. It also owns TripleTen, an edtech platform focused on reskilling, and Avride, which develops autonomous driving technology for self-driving cars and delivery robotics. The company has a strategic alliance with Nvidia(NVDA) to support expansion of its AI cloud infrastructure.
The current Nebius is effectively a new company built out of a major strategic reset. The business was formerly Yandex N.V. and changed its name to Nebius Group N.V. in August 2024. Its 2026 20-F states that the company exited Russia through a 2024 divestment that accounted for more than 95% of the group’s 2023 revenue and about 95% of consolidated assets and employees immediately before the transaction. That matters because historical comparisons across multiple years can look distorted. Investors are not looking at a mature software company with steady continuity. They are looking at a rebuilt AI infrastructure platform with a very different revenue base and risk profile.
Management’s strategic focus is now explicit. CEO Arkady Volozh said on the Q4 2025 call, “Our main strategic focus is to scale our core AI cloud business, which is our multi-tenant AI cloud.” The company ended 2025 with 1,543 employees and is led by founder and CEO Arkady Volozh, CFO Maria del Dado Alonso Sanchez, COO Ophir Nave, and CTO Danila Shtan. The headquarters is in Schiphol, the Netherlands, while the stock trades on Nasdaq under NBIS.
The investment case therefore rests far less on legacy internet assets and far more on whether Nebius can become a scaled AI compute and platform provider. The company’s own description, its capital allocation, and its customer wins all point in the same direction. This is an AI cloud infrastructure business first, with adjacent assets that may add optionality but do not drive the main thesis.
Nebius reports consolidated results that include the core AI infrastructure business, Avride, and TripleTen, while Toloka is no longer consolidated after a 2025 investment transaction. In Q4 2025, the core AI cloud business generated $214.2M of revenue, up more than 800% YoY, and represented about 94% of total group revenue. That concentration tells the story. Whatever else Nebius owns, the engine is AI cloud.
The core AI cloud segment is also where the economics are strongest. In Q4 2025, it delivered $51.8M of adjusted EBITDA at a 24% adjusted EBITDA margin. Management attributed that performance to high utilization, strong pricing, and strong execution. The company said it sold out capacity again in Q4, and management repeated that “everything we build, we sell.” That is exactly what investors want to hear in a market where unused GPU capacity can turn a growth plan into a very expensive science project.
TripleTen is much smaller. The shareholder materials said TripleTen contributed about 6% of group revenue in Q4 2025. That makes it relevant as a diversification asset, but not central to the valuation. It also recorded an adjusted EBITDA loss in the quarter, according to management commentary. In other words, TripleTen is currently a side business, not a margin driver.
Avride contributed immaterial revenue in Q4 2025 and FY 2025. It remains an option on autonomous driving and robotics rather than an earnings contributor. Management also highlighted that Avride and TripleTen both recorded adjusted EBITDA losses in Q4, while the group still turned positive on adjusted EBITDA because the core AI cloud business was strong enough to carry them.
Toloka moved out of consolidation after a strategic investment transaction announced on May 7, 2025. The 20-F says Nebius continued to hold a significant equity stake but ceased to hold majority voting power, so Toloka is now reported as a combination of equity method investment and investment in non-marketable equity securities. This matters because it reduces noise in the core operating picture. The market can now evaluate Nebius more cleanly as an AI infrastructure company rather than a collection of unrelated assets.
The segment read-through is simple. NBIS is increasingly a one-segment story with side pockets of optionality. That is usually a good thing for valuation clarity, but it also means the stock’s fate is tied tightly to AI cloud execution.
Get AI research on any stock
Instant reports, daily intelligence, and an AI analyst in your pocket.
Nebius’ flagship product is its full-stack AI cloud platform. This is not marketed as a general-purpose cloud competing feature-for-feature with Amazon(AWS is not a ticker), Microsoft(MSFT), or Alphabet(GOOGL). Instead, Nebius is positioning itself as purpose-built infrastructure for AI workloads, including GPU clusters, storage, cloud services, and developer tools. That specialization is the product.
The strongest evidence that the product is resonating comes from usage and monetization data. Core AI cloud revenue reached $214.2M in Q4 2025, up more than 800% YoY. Management said the business operated at peak utilization and sold out capacity in Q4. On the call, CFO Dado Alonso said core AI cloud revenue grew 830% YoY and 63% QoQ, driven by “high utilization, strong pricing, and strong execution.” Those are not vanity metrics. They show customers are buying, using, and renewing capacity at scale.
Nebius is also moving up the stack. The company launched AI Studio and an inference platform in Q4 2024, launched Token Factory in November 2025 for production inference at scale, and said the recent Tavily acquisition adds agentic search capabilities while bringing almost 700,000 developers to the platform. That matters because raw compute can become commoditized over time, while software layers, workflow tools, and developer ecosystems can improve retention and pricing power.
The hardware roadmap is another product strength. Investor materials say Nebius started deploying Nvidia B300 and GB300 NVL72 systems in 2025 and expects in 2026 to be among the first AI cloud providers to deploy Nvidia Vera Rubin NVL72 systems in the US and EMEA. In AI infrastructure, being early on leading hardware is not cosmetic. It can determine who wins the best customers, the longest contracts, and the highest-value workloads.
That quote from CFO Dado Alonso is useful because it explains how management wants the product judged. Revenue is influenced by deployment timing, but ARR captures the earning power of capacity once it is online. Nebius ended December 2025 with ARR above $1.2B, exceeding the high end of prior guidance. For a product business in hypergrowth mode, that is a strong sign of product-market fit.
Nebius’ edge is not scale. Against hyperscalers, it plainly does not have that. Its edge is specialization, speed, and access. The company says it builds a full-stack AI cloud from the ground up, and management repeatedly emphasized speed and reliability in scaling capacity. In a market where many customers care less about broad cloud menus and more about getting high-performance AI compute now, that focus matters.
The first competitive advantage is capacity discipline. Management said Nebius sold out capacity in Q3 and Q4 2025 and was already sold out in early 2026. It also said average contract duration of new cloud customers grew 50% and that pricing for GPUs did not fall even on previous generations. Those facts point to a favorable supply-demand balance. When a provider can fill capacity before it comes online, it is operating with the wind at its back.
The second advantage is customer validation. Nebius announced a multi-billion-dollar agreement with Microsoft(MSFT) in September 2025 and a long-term supply agreement with Meta Platforms(META) in March 2026. Management said Meta’s capacity was fully deployed by early February 2026 and that the first Microsoft tranche was delivered on time in November 2025. Large customers do not eliminate risk, but they do validate technical credibility and delivery capability.
The third advantage is ecosystem alignment with Nvidia(NVDA). Nebius has a strategic alliance with Nvidia and highlighted deployment of advanced Nvidia systems. In AI infrastructure, access to leading hardware and close alignment with the dominant chip ecosystem can be the difference between premium demand and second-tier demand. It is not a moat in the classic sense, but it is a real competitive asset.
The fourth advantage is product breadth inside a focused lane. Nebius is adding Token Factory, agentic search through Tavily, and software layers around its compute platform. Gartner’s industry framing cited Nebius alongside CoreWeave, Crusoe, and Nscale as part of the full-stack, compliance-oriented GPU cloud cohort. That is a useful place to be. It means Nebius is not just renting metal. It is trying to become an operating environment for AI workloads.
That line from Arkady Volozh captures the current advantage and the current risk. It shows demand strength today. It also means the company must keep building without losing discipline. This is a race car with a very large fuel bill.
Operations are the heart of the Nebius story because this business rises or falls on power, data centers, GPUs, and deployment timing. Management announced nine new data centers globally and said the company had already secured more than 2 gigawatts of power by February 2026, prompting it to raise its 2026 forecast to more than 3 gigawatts. It also said it is on track to deliver 800 megawatts to 1 gigawatt as available data center capacity.
That scale-up is material. Starting in Q2 2026, management expects to begin bringing some new sites online, with the majority of planned capacity deployed in the second half of the year. This timing also explains why 2026 revenue guidance of $3.0B to $3.4B sits below the year-end ARR target of $7B to $9B. Capacity comes online progressively, so the revenue run rate at year-end can be much higher than the average revenue recognized during the year.
The company’s supply chain and capital plan are tightly linked. Management broke 2026 CapEx into three parts: less than 10% for securing power, about 20% for building data centers, and the remaining majority for deploying GPUs. That is a useful detail because it shows Nebius is not just buying chips. It is building a vertically integrated AI infrastructure stack, from power to facility to compute.
Execution on major customer deployments has been solid so far. Management said the first Microsoft tranche was delivered on time in November 2025, the remaining tranches are scheduled through 2026 with more than half arriving in the second half, and Meta’s contracted tranches were delivered on time and entered servicing stage in early February 2026. The New Jersey site was also described as on track, with management saying partners had secured components in the supply chain and built safety margin into project timelines.
The operational risk is obvious. Nebius plans $16B to $20B of CapEx in 2026, after spending roughly $2.1B in Q4 2025 alone and $4.066B for full-year 2025 property and equipment purchases. This is a business where a delay in power, a bottleneck in hardware, or a customer deployment shift can create a very large mismatch between spending and revenue recognition. The company has momentum, but this is still heavy industrial execution wearing software-market clothing.
Nebius operates in one of the market’s most attractive pockets: AI infrastructure. Deloitte reported that 86% of respondents expect AI infrastructure budgets to increase over the next three years, with budgets expected to more than triple on average. Deloitte also said the share of respondents with 31 or more production-ready AI use cases is expected to rise from 44% in 2025 to 67% by 2028. That is the broad demand backdrop behind Nebius’ growth.
The market is also shifting from experimentation to production. Deloitte’s framing of “AI factories” as purpose-built infrastructure combining compute, network, storage, and AI-optimized software fits Nebius almost perfectly. The company is not selling a generic cloud seat. It is selling throughput, reliability, and deployment speed for AI workloads. That is a market where specialized providers can carve out meaningful share even without matching hyperscaler breadth.
Supply constraints remain central. Deloitte highlighted HBM memory costs, longer procurement times, and power and grid interconnection constraints as key bottlenecks. Nebius’ own strategy around securing gigawatts of power and expanding data center capacity directly addresses that bottleneck. In other words, the company is not just chasing demand. It is trying to own the scarce inputs that determine who can serve demand.
The software layer around infrastructure is also expanding. Gartner’s 2025 technology trends include agentic AI, AI governance platforms, and GenAI platform engineering. Nebius’ moves into Token Factory and Tavily fit that direction. The company is trying to capture a larger slice of the AI stack, not just the compute rental line item.
The market opportunity is large, but it is not frictionless. Specialized GPU clouds are competing in a fast-growing field where customers increasingly want full-stack offerings, compliance, and cost-effective compute. That creates room for Nebius, but it also means the company must keep proving that its platform is differentiated enough to win repeat business once capacity becomes less scarce.
Like what you're reading?
Get full access to AI-powered research reports, market analysis, and portfolio tools.
Nebius’ customer base spans hyperscalers, AI-native startups, ISVs, enterprises, and vertical-specific users. Management said the company is seeing traction across healthcare and life sciences, media and entertainment, physical AI, and retail. It also said AI-native customers are evolving from ordering hundreds or thousands of GPUs to ordering tens of thousands, while enterprise customers are using AI for more vital business processes and signing longer contracts.
The customer profile has two attractive features. First, Nebius serves both very large strategic accounts and a broader multi-tenant cloud base. Management said its 2026 ARR target is not dependent on any new mega deals, which implies the broader customer pipeline matters alongside Microsoft and Meta. Second, a substantial portion of group revenue is collected on a prepaid basis, according to the 20-F, which can improve cash flow visibility.
There is also concentration risk. The 20-F states that customer concentration is a primary area of potential risk, and the company’s own narrative makes clear that large contracts play a meaningful role in the ramp. Microsoft and Meta are excellent proof points, but they also mean a small number of customers can influence near-term revenue timing in a big way. In this business, one delayed tranche can move a quarter.
From a commercial standpoint, Nebius looks strongest with customers that value speed, performance, and access to compute rather than the broadest possible cloud catalog. That is a good niche today. The challenge is keeping that niche profitable as the market matures.
Nebius’ direct competitors in core AI infrastructure include specialized cloud service providers focused on AI, including CoreWeave, Crusoe, and Lambda, according to industry context and the company’s 20-F. Gartner also grouped Nebius with CoreWeave, Crusoe, and Nscale in the bucket of full-stack, compliance-oriented, scalable GPU clouds. That is the right peer set conceptually, even though detailed peer multiple data is not available here.
The company also competes indirectly with hyperscalers such as Microsoft(MSFT), Alphabet(GOOGL), and Amazon(AMZN), because customers can choose to run AI workloads on broader cloud stacks. That is the awkward part of the Nebius story: one of its largest customers is also part of the broader competitive universe. Markets can be funny that way. Sometimes your customer is your rival, your supplier is your strategic partner, and your moat is basically execution speed.
Relative to hyperscalers, Nebius is smaller but more focused. Relative to niche GPU rental providers, it is trying to offer more software and platform depth. Relative to peers like CoreWeave, its differentiation rests on full-stack architecture, Nvidia alignment, marquee customer wins, and a rapidly expanding power footprint. The market is attractive enough for several winners, but it is also crowded enough that no one gets to coast.
The practical conclusion is that Nebius does not need to beat hyperscalers at being hyperscalers. It needs to remain the specialist provider that customers call when they need AI capacity, deployment speed, and a more purpose-built stack. So far, the revenue growth and capacity utilization data say it is doing that.
Macro matters here mostly through capital markets, enterprise AI spending, and infrastructure bottlenecks. On the demand side, rising AI infrastructure budgets are a tailwind. Deloitte’s survey showing 86% of respondents expect higher AI infrastructure budgets over the next three years supports the idea that Nebius is riding a real spending cycle rather than a short-lived fad.
On the cost side, this is a capital-intensive business exposed to financing conditions. Nebius raised significant capital during 2025, and management said it is exploring corporate debt and asset-backed financing in 2026 while remaining mindful of shareholder dilution if it uses equity. When a company plans $16B to $20B of annual CapEx, the cost of capital is not a footnote. It is part of the product.
Geopolitically, the biggest company-specific issue is already in the rearview mirror but still shapes investor perception. The 2024 divestment of Russian businesses was a strategic shift that removed more than 95% of the prior revenue base and marked the company’s exit from Russia. That reduces one major overhang, but it also means Nebius is still building a new identity in public markets.
There is also a sovereignty and regional infrastructure angle. Gartner’s cloud trends point to digital sovereignty as a shaping force, and Nebius expects to deploy advanced Nvidia systems in the US and EMEA. A geographically diversified AI infrastructure footprint can become more valuable as customers care more about compliance, latency, and jurisdictional control over compute.
The macro read is favorable for demand but unforgiving for execution. If AI spending stays strong and capital remains available, Nebius has room to scale fast. If financing tightens or infrastructure bottlenecks worsen, the company’s growth plan becomes harder and more expensive.
Nebius spent about $4.066B on property and equipment in 2025 and is planning $16B to $20B of 2026 CapEx, underscoring how capital-intensive the AI buildout has become.
Unlock the full analysis
Subscribers get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full AccessRevenue reached $529.8M in 2025, up 479% YoY, while Q4 core AI cloud revenue of $214.2M produced $51.8M of adjusted EBITDA at a 24% margin.
Unlock the full analysis
Subscribers get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full AccessManagement guided 2026 revenue to $3.0B-$3.4B and said annualized run-rate revenue hit $1.2B in December 2025, above prior guidance.
Unlock the full analysis
Subscribers get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full AccessWith a market cap near $45.48B, forward P/E of 68.49, and EV/revenue of 87.29, Nebius already prices in a lot of future AI cloud success.
Unlock the full analysis
Subscribers get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full AccessThe report frames NBIS as a selective buy on pullbacks rather than a chase-at-any-price compounder because the current valuation leaves little room for execution mistakes.
Unlock the full analysis
Subscribers get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full AccessNebius is one of the more interesting companies in public markets because it sits right at the intersection of AI demand, infrastructure scarcity, and capital intensity. The company has already shown enough to be taken seriously. Revenue growth is explosive, core margins are improving, marquee customers are onboard, and management is securing the power and capacity needed to scale. This is not a concept stock pretending to be an industrial platform. It is an industrial platform trying to become a software-quality business.
That said, the stock is not offering a wide margin of safety today. The company’s own numbers show a business still in heavy build mode, with negative EBIT and a huge 2026 CapEx plan. Analyst targets near the current share price and recent insider selling reinforce the idea that enthusiasm is already well represented in the stock.
The balanced conclusion is to respect both sides of the tape. Nebius has the ingredients to become a major AI infrastructure winner over the medium term, and the operating data supports that possibility. But at current levels, the better stance is disciplined optimism rather than blind momentum. For investors who want exposure, patience still looks like a feature, not a bug.
NBIS is a high-quality AI infrastructure name, but the report does not treat it as a straightforward buy at current levels. It is better viewed as a selective opportunity on weakness because growth is exceptional while valuation and capex demands remain very high.
NBIS's fair value is not provided in the grades block you shared, so I can't state a dollar estimate from this report. The analysis instead emphasizes that the stock already trades at a premium multiple, with forward P/E of 68.49 and EV/revenue of 87.29, so the key issue is whether execution can justify that valuation.
Nebius is one of the clearest pure-play ways to invest in AI infrastructure, and its core AI cloud business is scaling fast. Q4 2025 revenue was $227.7M, up 547% YoY, and management said capacity was sold out again while major customers like Microsoft and Meta were being onboarded.
The biggest risks are capital intensity and valuation. Nebius plans $16B to $20B of 2026 CapEx after spending about $4.066B on property and equipment in 2025, and the stock already reflects a lot of future growth with a market cap around $45.48B.
Very strong: the core AI cloud business generated $214.2M of Q4 revenue, about 94% of group revenue, and produced $51.8M of adjusted EBITDA at a 24% margin. Management also said new cloud customer contract duration increased 50%, which supports the durability of demand.
Get AI-powered research reports, daily market intelligence, and a personal analyst in your pocket.
Get Full Access
Nebius Group N.V. (NBIS) jumps after-hours after reporting Q1 2026 results that topped expectations. Better-than-expected revenue and a smaller loss fueled the rally, pushing the AI infrastructure stock above its prior 52-week high and keeping investors focused on rapid growth, heavy spending, and premium valuation.
No, Perplexity is not publicly traded. Retail investors can’t buy Perplexity stock directly today, so the practical paths are waiting for a future IPO or using public proxies like Alphabet, Microsoft, and Baidu.
No, Temu is not publicly traded. The investable public parent is PDD Holdings (NASDAQ: PDD), while Temu remains a business line inside that company. If you want exposure now, the realistic paths are PDD shares or public marketplace comps like Amazon, Alibaba, and eBay.