Nebius Group is a cloud infrastructure company that builds and rents out the massive computing clusters needed to run artificial intelligence. It generated $400 million in revenue in its most recent quarter, representing nearly 566% growth over the same period last year. Following its formal separation from the Russian internet giant Yandex, the company has repositioned itself as a pure-play provider of AI-centric computing power for global developers.
The investment thesis on Nebius Group is that it can carve out a specialized niche as the primary European provider of high-end GPU capacity, competing on hardware design and software optimization rather than just scale. While it faces giants like Amazon and Microsoft, Nebius designs its own servers and racks to extract more performance from the chips it buys. If it can successfully scale its data centers in Finland and France while maintaining its partnership with Nvidia, the business could become a critical infrastructure layer for the AI industry.
We view Nebius as a high-potential turnaround story that has already proven it can build world-class infrastructure, but the stock price currently reflects a significant amount of that future success. The business is scaling exceptionally fast, yet the execution risk of building billion-dollar data centers remains high.
Nebius stock has soared lately after the company completely reinvented itself. The shares took off because the business stopped being part of a Russian internet giant and started renting out the massive computer power needed to build artificial intelligence. This shift has led to huge growth, though investors remain cautious about the risks involved.
What does it do?
Nebius Group is a hypergrowth business that earns money by renting out high-performance computing power to companies building and running artificial intelligence models. The company builds massive data centers filled with thousands of specialized chips, primarily Nvidia GPUs, and allows customers to access this power over the internet. Instead of buying millions of dollars in hardware themselves, AI developers pay Nebius a recurring fee based on the amount of computing time and storage they use. This "GPU-as-a-Service" model is enhanced by a proprietary software layer that helps developers train their models more efficiently.
Where does revenue come from?
The vast majority of revenue comes from its AI-centric cloud platform, which now accounts for approximately two-thirds of total company sales. The remaining revenue is generated by a handful of legacy and adjacent technology businesses, including autonomous driving development and edtech services. Geographically, the company is focused on the European market, with a major data center hub in Finland and a new cluster in Paris, while maintaining R&D centers in North America and Israel.
Who are its customers?
Nebius Group serves more than 40 managed cloud clients, including Fortune 500 companies and specialized AI startups. The customer count has grown quickly from 30 clients in the prior quarter as the company brings new GPU clusters online. While the specific names of all clients are not disclosed, management has noted that the base includes significant AI labs that require intensive computing power for model training. The company reported $400 million in revenue in the most recent quarter, indicating that while the customer count is still small, each client represents a substantial spending commitment.
What gives it staying power?
Staying power comes from the company's proprietary hardware design and deep integration with Nvidia. Nebius does not just buy standard servers: it designs its own racks and cooling systems to handle the extreme heat of AI workloads. This hardware specialization, combined with early access to the latest chips like the H200, makes it difficult for generic cloud providers to match its efficiency.
Where is it headed?
The company is making a massive strategic bet on becoming Europe's largest independent AI cloud provider by investing over $1 billion in infrastructure. Management is tripling the capacity of its Finland data center to 75 megawatts, which could eventually house 60,000 GPUs. If successful, this expansion could turn the current $400 million quarterly revenue into a multibillion-dollar annual business by 2026.
The most important trend is the massive revenue acceleration, with sales jumping from $60 million to $400 million in just one year. This growth is not organic expansion but the result of a total business pivot that is now seeing full market demand for GPU capacity.
Cash quality is currently poor as the company burns through capital to build the data centers required for its new model. Free cash flow was negative $3.68 billion in the last fiscal year, reflecting the massive CapEx needed to buy Nvidia chips and build high-tech cooling infrastructure.
The balance sheet is recently fortified following a $700 million capital raise from investors including Nvidia and Accel. While the company carries a debt-to-equity ratio of 1.31x, the recent influx of cash provides the runway needed to complete its current data center expansions in Europe.
Nebius Group is a financially transforming business whose massive growth is currently being funded by external capital rather than internal profits. Nebius Group is a financially transforming business whose massive growth is currently being funded by external capital rather than internal profits.
The core AI cloud business is scaling at a rare pace, with the annualized revenue run-rate surpassing $1.6 billion based on the latest quarter. This suggests that as soon as Nebius brings new GPU clusters online, they are immediately being filled by paying customers.
The primary risk is the massive cash burn required to stay competitive in the GPU arms race. If capital markets tighten or Nvidia chip availability fluctuates, Nebius could find its data centers half-finished and its cash reserves depleted before it reaches self-sustaining profitability.
The AI cloud infrastructure market is currently valued at roughly $100 billion and is growing at over 30% annually as companies shift from general computing to specialized AI workloads. This is an excellent industry for specialized players because demand for high-end GPUs currently far exceeds supply, granting providers significant pricing power for the time being. Nebius Group is a challenger in this market, aiming to occupy the "Sovereign Cloud" niche in Europe where local data residency and specialized support are competitive advantages over US giants.
The competitive dynamic is defined by a desperate race for hardware allocation and energy capacity rather than price competition. While entry barriers are theoretically low for anyone with capital, the true bottleneck is the relationship with Nvidia and the specialized engineering required to run thousands of GPUs in sync. The industry is currently fragmenting as specialized "GPU clouds" take share from general providers who cannot move as fast.
The most dangerous threats are the US hyperscalers (Amazon, Microsoft, and Google) who have the capital to outspend Nebius ten-to-one on hardware. They can bundle AI compute with their existing software suites, making it difficult for an independent provider to win long-term loyalty. Nvidia-backed specialized clouds like CoreWeave represent the most direct threat to Nebius's technical edge.
Nebius is currently gaining share by moving faster than the giants to deploy the latest H200 chips in European markets.
The primary source of protection is the company's proprietary technology stack, which includes in-house designed servers, racks, and cooling systems. This hardware specialization allows Nebius to achieve higher GPU utilization and lower latency than providers using off-the-shelf equipment. The company's ability to extract 10% to 15% more performance from the same chip creates a narrow but real efficiency moat.
The financial metrics currently show a business in flux: the 95% net margin is a temporary result of divestiture gains rather than core operations. However, the 47.9% gross margin in a capital-intensive business proves that there is significant value being added beyond just reselling hardware. These numbers suggest a business that could develop a real moat if it can scale its software layer to match its hardware performance.
The moat is currently strengthening as the company deepens its design partnership with Nvidia and scales its unique European footprint.
Scaled quarterly revenue from $60M to $400M in one year.
Raised $700M from Nvidia and Accel to fund data center expansion.
Founder Arkady Volozh retains significant ownership and leadership control.
Capital Allocation Track Record
Arkady Volozh is a proven technology founder who built one of the world's most successful internet companies outside the US, and his return to leadership signals high strategic caliber. He has demonstrated an exceptional ability to attract top-tier global talent and secure critical partnerships with players like Nvidia during a period of extreme corporate transition. His decision to pivot entirely to AI infrastructure shows a clear vision and the willingness to abandon legacy models in favor of the highest-growth opportunity available.
The primary governance risk is the high degree of key-person dependence on Volozh, whose history and dual-class control structure give him nearly total authority over the company's direction. While his interests are clearly aligned with long-term value creation, there is a limited public bench of executives who could maintain the company's momentum if he were to leave. Investors are essentially backing Volozh's ability to repeat his prior success on a new, more specialized global stage.
We expect revenue to grow from $3.5B in FY2026 to $61.8B in FY2031 (~77% CAGR), with EPS growing from $-2.11 to $11.90. Revenue scales rapidly as the company brings massive new GPU clusters online to meet the global demand for AI compute power. Profitability improves as the heavy initial investment in data centers and hardware is spread across a rapidly growing base of cloud customers. EPS grows Operating margin expected to reach ~35% by FY2031.
European sovereign cloud demand drives massive regional adoption. As EU regulations tighten, European companies will increasingly prefer local AI infrastructure over US hyperscalers to ensure data residency and compliance.
Proprietary hardware design lowers operating costs versus competitors. Custom racks and cooling systems allow Nebius to run GPUs more efficiently, leading to higher margins as the network scales.
Expansion into AI inference software creates recurring revenue lock-in. Launching software platforms like AI Studio helps developers manage models, turning a hardware rental business into a sticky software ecosystem.
GPU supply constraints limit the pace of data center expansion. If Nvidia shifts allocation priority away from smaller providers, Nebius may be unable to fill its planned data center capacity on schedule.
Hyperscaler price wars compress margins in the GPU rental market. If Amazon or Microsoft aggressively lower prices to win AI startups, Nebius's specialized hardware edge may not be enough to maintain pricing power.
Massive capital requirements force highly dilutive secondary stock offerings. The $1 billion expansion plan may require more cash than currently available, potentially hurting existing shareholders through repeated capital raises.
Below is our estimate of current and future fair value, with detailed reasoning and assumptions. Fair value is a judgment, not a fact, and other analysts will likely land on different numbers. Use it as one data point in your research, and apply your own discretion in any investing decision.
We use a Sum-of-the-Parts (SOTP) approach to value Nebius Group. This framework fits the business because it accurately separates the hyper-growth AI cloud infrastructure—which commands a high revenue multiple—from the valuable but distinct portfolio of non-core equity stakes like ClickHouse and Avride.
The calculation adds a $50.4B valuation for the Core AI segment to $9.2B in non-core assets, yielding an equity value of $59.4B after adjusting for $200M in net debt. Our 4.5x EV/Revenue multiple for the Core AI segment sits between the 8-10x range of private specialized peers like CoreWeave and the 3-5x range of mature providers like Oracle; this premium is justified by the NVIDIA partnership and 225% projected revenue growth. We used the consensus FY2027 revenue estimate of $11.2 billion as the basis for the core business valuation.
Cross-checked with the deterministic 5-year DCF, which produces a fair value of $207, our SOTP result of $234 is within 12% and confirms the valuation. The slight premium in our SOTP model is due to its more explicit capturing of the private market valuations for ClickHouse and other non-core assets, which may not be fully reflected in near-term free cash flow projections. Both models strongly agree that the current market price of $222.24 is roughly in line with intrinsic value, assuming the massive infrastructure ramp-up remains on schedule.
We're assuming the Core AI Infrastructure segment generates $11.2 billion in revenue by FY2027. This matches the consensus analyst estimate and is supported by management's target of reaching 800MW to 1GW of connected power by the end of 2026, which provides the physical foundation for this revenue ramp.
We're assuming the portfolio of non-core business units, including ClickHouse and Avride, is worth approximately $9.2 billion. This is a base-case valuation for these venture-stage assets, anchored by ClickHouse's recent $15 billion private valuation (where Nebius holds a 25% stake) and established funding rounds for the other segments.
We're assuming the company maintains access to capital to fund its $2.5 billion annual CapEx requirements without massive shareholder dilution. The recent $2 billion investment from NVIDIA and a current cash balance of $9.3 billion provide a significant multi-year runway, reducing the immediate risk of value-destructive equity raises.
The biggest risk is a "GPU glut" or a sharp slowdown in AI model training demand that leaves Nebius with massive underutilized capacity and high debt service costs. This would likely force the market to value the Core AI business at a fire-sale revenue multiple of 1.5x rather than 4.5x, knocking over $120 off the per-share fair value. Watch for quarterly capital expenditure (CapEx) trends outstripping revenue growth for more than two consecutive quarters.
Bear case ($100): AI infrastructure utilization rates fall below 60% as competition from CoreWeave and Lambda intensifies; or Non-core assets like ClickHouse or Avride face funding rounds at significantly lower valuations than previous Series D marks.
Bull case ($470): FY2027 revenue exceeds $18 billion driven by the NVIDIA strategic partnership and rapid expansion to 1GW of power; or A successful IPO or sale of ClickHouse at a valuation exceeding $35 billion provides a massive cash windfall.
Clearthesis wrote this report from 39 sources, including SEC filings, industry research, and recent news.
How did you like this thesis?
Your feedback helps us make reports better for you
© 2026 Clearthesis.ai · Report generated on June 23, 2026
This is an AI-generated analysis for informational purposes only and does not constitute financial advice. Data and analysis may not reflect recent developments if viewed significantly after the generation date. Always conduct your own due diligence before making any investment decisions.
The market is leaning bullish because Nebius is capturing massive demand as a specialized provider of high-end computing power for artificial intelligence developers. The company grew revenue by 566 percent last quarter by successfully separating from its former Russian operations. It is now building a dedicated infrastructure business focused on renting specialized hardware to global AI teams.
Skeptics think that the company carries significant risks linked to its complex corporate history and the difficulty of maintaining a specialized niche. Investors worry that the rapid growth might be hard to sustain as competition from larger cloud giants increases and the company works to shed its past associations.