Nebius Group has gone from a sanctioned Yandex shell company to the most aggressively expanding AI infrastructure player on the planet in less than two years. The stock is up roughly 400% over the past twelve months. It has signed deals worth a combined $46 billion with Meta and Microsoft. Nvidia just invested $2 billion in it. And the company is planning to spend $16 billion to $20 billion on data centers in 2026 alone.
The question for investors is no longer whether Nebius can grow. The question is whether the growth justifies the valuation, whether the capital structure can sustain the buildout, and whether this stock has more in common with early Amazon Web Services or with the debt-fueled infrastructure bets that imploded in the early 2000s.
From Yandex’s Ashes to AI Neocloud
Understanding Nebius requires understanding where it came from. The company’s predecessor, Yandex N.V., was the Dutch parent of Russia’s dominant search engine, often called “Russia’s Google.” Yandex went public on the NASDAQ in 2011, raising $1.3 billion in one of the largest tech IPOs of that era.
Then Russia invaded Ukraine. NASDAQ suspended Yandex trading in February 2022. By July 2024, the company had sold all Russian assets to a consortium of domestic investors and retained only its non-Russian businesses: a Finnish data center, the Nebius AI cloud unit, Toloka AI (a data labeling platform), TripleTen (an edtech company), and Avride (autonomous driving).
The company renamed itself Nebius Group N.V., headquartered in Amsterdam, and resumed NASDAQ trading in October 2024. Founder Arkady Volozh, who co-founded Yandex in 1997, returned as CEO after renouncing his Russian citizenship in February 2026. Volozh had immigrated to Israel before the pandemic and built Nebius into a full-stack AI infrastructure company from what was essentially an orphaned data center and a handful of spinoff businesses.
That origin story matters because it explains both the company’s strengths (deep technical DNA from building Russia’s largest search infrastructure) and its risks (governance questions, geopolitical overhang, and a management team that had to rebuild credibility from scratch).
The Business: Full-Stack AI Cloud Infrastructure
Nebius positions itself as a “neocloud,” a purpose-built cloud platform optimized for AI workloads rather than general-purpose computing. Unlike AWS, Azure, or Google Cloud, which bolt AI capabilities onto existing infrastructure, Nebius designs its entire stack around GPU-accelerated compute from day one.
The core offering is Nebius AI Cloud, which provides GPU clusters, managed Kubernetes for AI training and inference, and a storage layer optimized for the massive datasets that foundation model development requires. The platform runs primarily on Nvidia hardware, which is why the Nvidia investment is strategically significant. It gives Nebius preferred access to next-generation GPU architectures including Blackwell Ultra and Vera Rubin.
Revenue Breakdown
| Segment | Q4 2025 Revenue | % of Total | Notes |
|---|---|---|---|
| AI Cloud (Core) | $214.2M | ~94% | GPU cloud infrastructure; 547% YoY growth |
| Toloka AI | ~$7M | ~3% | Data labeling and annotation platform |
| TripleTen | ~$5M | ~2% | Tech bootcamp / edtech |
| Avride | ~$2M | ~1% | Autonomous driving (early stage) |
| Total | $227.7M | 100% | ~504% YoY growth (total); AI Cloud segment grew 547% |
The AI Cloud segment is the entire investment thesis. Everything else is either a future optionality play (Avride) or a modest contributor (Toloka, TripleTen). Investors are buying this stock for the GPU cloud buildout.
The Numbers That Matter
Nebius reported full-year 2025 revenue of $529.8 million, up 479% year over year. Q4 alone contributed $227.7 million, with quarter-over-quarter growth of 56%. Exit annualized recurring revenue hit $1.25 billion at year-end, exceeding the high end of management’s $1.1 billion guidance range.
The AI Cloud segment delivered $51.8 million in adjusted EBITDA at a 24% margin in Q4, a meaningful improvement from earlier quarters when margins were negligible. However, the company posted a net loss of $173 million for Q4, driven by heavy infrastructure spending and share-based compensation.
Financial Summary
| Metric | FY2024 | FY2025 | FY2026 Guidance |
|---|---|---|---|
| Revenue | $91.3M | $529.8M (+479%) | $3.0B–$3.4B |
| Exit ARR | ~$250M | $1.25B | $7B–$9B |
| Adj. EBITDA Margin | Negative | ~20% (Q4: 24%) | ~40% (target) |
| Net Income | ($340M) | ($518M) | N/A |
| CapEx | ~$1.5B | ~$4B | $16B–$20B |
| Cash & Equivalents | $2.5B | $8.3B | Raising additional capital |
| Connected Power | ~50 MW | 170 MW | 800 MW–1 GW |
The growth trajectory is extraordinary. But the gap between revenue ($3.0 to $3.4 billion guided) and planned capital expenditure ($16 to $20 billion) is the number that should keep investors awake at night. Nebius plans to fund approximately 60% from existing cash, operating cash flow, and committed financing, with the remainder coming from debt, asset-backed financing, and opportunistic equity raises.
The $46 Billion Contract Stack
What separates Nebius from the pack of aspiring AI infrastructure companies is the quality and scale of its customer commitments.
Major Contracts
| Customer | Deal Value | Duration | Details |
|---|---|---|---|
| Meta Platforms | Up to $27B | 5 years | Dedicated AI cloud capacity on Vera Rubin platform; delivery from early 2027 |
| Microsoft | Up to $19.4B | Multi-year | GPU compute from New Jersey data center |
| Nvidia | $2B equity investment | Strategic | Partnership for 5+ GW of Nvidia systems by 2030; preferred GPU access |
The Meta deal, announced March 16, 2026, is the crown jewel. A five-year, up to $27 billion commitment for Nebius to deploy dedicated AI cloud capacity powered by Nvidia’s Vera Rubin platform, with delivery beginning in early 2027. When Meta, a company building its own custom silicon and operating some of the world’s largest data centers, outsources AI infrastructure to you, it is a profound validation of your technical capabilities and cost structure.
The Microsoft deal, worth up to $19.4 billion, provides GPU compute from a New Jersey facility. This is the same Microsoft that owns Azure, the world’s second-largest cloud platform. The fact that Microsoft needs to lease external GPU capacity tells you everything about the current state of AI compute demand: even the hyperscalers cannot build fast enough.
The Data Center Empire
Nebius ended 2025 with 170 MW of connected data center capacity. The company is targeting 800 MW to 1 GW connected by end of 2026, and has contracted more than 3 GW of power. That is a 5x to 6x capacity expansion in a single year.
Global Data Center Footprint
| Location | Capacity | Status | Notes |
|---|---|---|---|
| Mantsala, Finland | 75 MW (tripled from 25 MW) | Operational | Original Yandex facility; fully owned |
| Lappeenranta, Finland | 310 MW | Under construction | One of Europe’s largest AI factories; online 2027 |
| Independence, Missouri | Up to 1.2 GW | Approved | 2.5M sq ft campus on 400 acres; power from H2 2026 |
| Kansas City | 5–40 MW | Operational | Colocation; ~35K GPU capacity at full build |
| New Jersey | Up to 300 MW | Expanding | Microsoft deal anchor site |
| Lille, France | 240 MW | Under development | European expansion; EMEA total >750 MW secured |
| Paris, France | Leased | Operational | Equinix colocation |
| London, UK | Leased | Operational | European presence |
| Keflavik, Iceland | Leased | Operational | Low-cost power; cold-climate cooling |
The Missouri campus is the headline project. A 1.2 GW facility on 400 acres outside Kansas City, with approval already secured and power delivery starting in the second half of 2026. If fully built, this single campus would have more power capacity than many mid-sized cities.
In Europe, Nebius has secured more than 750 MW across owned and leased facilities in the EMEA region, anchored by the Mantsala site in Finland that dates back to the Yandex era. The 310 MW Lappeenranta facility positions Nebius as a major player in Europe’s scramble for sovereign AI compute capacity.
Nebius vs. CoreWeave: The Neocloud Showdown
Every investor comparing AI infrastructure stocks eventually lands on the same question: Nebius or CoreWeave? Both companies operate GPU-accelerated cloud platforms. Both have massive customer contracts. Both are burning capital to build data centers. The differences are in the details. For a deeper breakdown, see our full Nebius vs. CoreWeave comparison.
| Metric | Nebius (NBIS) | CoreWeave (CRWV) |
|---|---|---|
| Market Cap | ~$25.2B | ~$40.7B |
| Stock Performance (12 months) | ~400% | ~109% |
| FY2025 Revenue | $529.8M | ~$5.1B |
| Revenue Growth (YoY) | 479% | ~168% |
| Key Customers | Meta ($27B), Microsoft ($19.4B) | Microsoft, Nvidia, various enterprises |
| Connected Power (End 2025) | 170 MW | ~500 MW |
| 2026 Power Target | 800 MW–1 GW | ~1.5 GW |
| Nvidia Relationship | $2B equity investment | Long-standing GPU customer |
| Geographic Focus | US + Europe (Finland, France, UK, Iceland) | Primarily US |
| Origin | Yandex spinoff (Amsterdam) | US-founded startup |
Nebius is smaller but growing faster and trading at a lower market cap despite comparable deal flow. Its European presence is a genuine differentiator as EU data sovereignty regulations drive demand for on-continent AI compute. CoreWeave is more established in the US market and has a larger existing infrastructure base, but Nebius’s Meta and Microsoft contracts close that gap rapidly.
Both stocks carry similar risks: massive capital requirements, customer concentration, and GPU supply dependence. The key differentiator for Nebius is geographic diversification and the Meta relationship; for CoreWeave, it is a larger existing revenue base and more mature operations.
The Bull Case
Demand visibility is unmatched. $46 billion in committed contracts from Meta and Microsoft provide revenue visibility that most growth companies can only dream of. These are not letters of intent or memorandums of understanding. They are binding multi-year commitments from two of the three largest technology companies on Earth.
Nvidia‘s $2 billion investment is a strategic moat. Nvidia does not invest in companies for financial returns alone. The investment gives Nebius preferred access to next-generation GPU architectures, which means Nebius can offer customers the latest hardware before competitors who are stuck in the general ordering queue. In a market where GPU availability determines revenue, this access is worth more than the $2 billion cash.
The European advantage is structural. EU data sovereignty regulations, particularly the AI Act’s data residency requirements, are creating massive demand for on-continent AI compute. Nebius is the only neocloud with significant owned infrastructure in Europe. AWS and Azure are expanding European GPU capacity, but they face regulatory scrutiny that homegrown alternatives do not.
Margin expansion is baked into the model. AI cloud infrastructure has high fixed costs and low marginal costs. As Nebius fills its data centers, margins expand dramatically. The jump from negative EBITDA margins in 2024 to 24% in Q4 2025 to a 40% target for 2026 reflects this operating leverage. At $3 billion in revenue and 40% EBITDA margins, Nebius would generate $1.2 billion in adjusted EBITDA, changing the valuation math entirely.
The Bear Case
The capital structure is aggressive by any standard. Planning $16 to $20 billion in CapEx against $3 to $3.4 billion in revenue means Nebius is spending roughly $5 to $6 for every $1 it earns. Even with $8.3 billion in cash and committed financing covering 60% of the buildout, the remaining 40% requires additional debt, asset-backed financing, or equity raises. The company has already raised $4.6 billion via convertible notes and announced an ATM equity program for up to 25 million Class A shares. Dilution is not a hypothetical risk; it is the funding plan.
Customer concentration is extreme. Two customers, Meta and Microsoft, account for the vast majority of contracted revenue. If either deal gets restructured, delayed, or partially canceled, the financial model breaks. Meta has its own custom silicon program and could reduce external GPU leasing if its in-house chips deliver. Microsoft has Azure and could reallocate workloads internally as its own capacity expands.
Execution risk at this scale is not trivial. Scaling from 170 MW to 800 MW to 1 GW of connected power in a single year requires securing real estate, procuring tens of thousands of GPUs, hiring hundreds of data center engineers, and managing construction timelines across four countries. Any GPU allocation delay from Nvidia, any permitting setback in Missouri or France, any construction timeline slip could push revenue recognition out of 2026 and into 2027.
Valuation leaves no margin for error. At $25.2 billion market cap on $530 million of trailing revenue, Nebius trades at approximately 47x trailing sales. Even on 2026 guided revenue of $3 to $3.4 billion, the stock trades at 7 to 8x forward revenue, which sounds reasonable until you remember the company is deeply unprofitable with a $518 million net loss in 2025. If AI infrastructure spending decelerates, or if hyperscalers shift workloads to in-house silicon faster than expected, the multiple compresses quickly.
Geopolitical overhang persists. Despite the Yandex separation and Volozh’s renunciation of Russian citizenship, the company’s origins continue to create perception risk among institutional allocators and ESG-focused funds. This may cap the investor base and limit multiple expansion relative to CoreWeave or other neoclouds without similar history.
Analyst Consensus
| Metric | Value |
|---|---|
| Consensus Rating | Strong Buy (8 analysts) |
| Average Price Target | $169 |
| Highest Target | $291 |
| Lowest Target | $108 |
| Implied Upside (from $112.54) | ~50% |
Wedbush analyst Dan Ives named Nebius his top AI infrastructure pick for 2026, arguing the company’s positioning and technology stack make it a potential acquisition target for hyperscalers like Microsoft, Alphabet, or Amazon. Goldman Sachs maintains a Buy rating with a $155 price target. Nebius also features in our roundup of the best AI stocks to watch this year.
Key Catalysts and Risks Timeline
| Timeframe | Catalyst / Risk | Impact |
|---|---|---|
| Q1 2026 | Q4 2025 earnings reaction; guidance for $3B+ revenue | Set expectations for full year; stock moved 15% on Meta deal |
| H1 2026 | Missouri data center power delivery begins | Validates largest single-site buildout; capacity drives revenue |
| H2 2026 | 800 MW–1 GW connected power milestone | Confirms execution on 5x capacity expansion |
| H2 2026 | Additional capital raises (debt/equity) | Dilution risk from funding 40% of $16–$20B CapEx |
| Early 2027 | Meta deal deliveries begin on Vera Rubin platform | $27B contract starts generating recognized revenue |
| 2027+ | Lappeenranta Finland 310 MW facility online | European capacity flywheel; data sovereignty demand driver |
| Ongoing | Nvidia GPU allocation and next-gen chip access | Supply chain determines how fast Nebius can fill capacity |
The Verdict
Nebius is the most compelling AI infrastructure growth story in the public markets right now. The contract stack is real: $46 billion from Meta and Microsoft, a $2 billion Nvidia strategic investment, and a data center buildout that spans three continents. The revenue trajectory from $91 million in 2024 to guided $3 to $3.4 billion in 2026 is the kind of hypergrowth that creates generational wealth for early shareholders.
But the risks are equally real. The capital plan is aggressive to the point of fragility. The company needs to spend $16 to $20 billion to capture the revenue it has contracted, and roughly 40% of that funding has not been secured. Customer concentration on two hyperscalers means any contract renegotiation hits the entire business model. And the Yandex heritage, while now legally and operationally severed, still creates an institutional perception discount that the stock has not fully shaken.
At $112.54, Wall Street sees roughly 50% upside to a $169 average price target, with the most bullish analyst at $291 implying potential to nearly triple from here. Those targets assume near-perfect execution on the buildout, no material contract revisions, and continued AI infrastructure spending growth. You can track the latest NBIS price action on Yahoo Finance.
For investors with a two-to-three-year horizon and tolerance for volatility, Nebius offers exposure to the AI infrastructure buildout through a company that has already won the customers that matter most. For those who need capital preservation or predictable returns, the execution risk and capital structure make this a stock to watch rather than own until the 2026 buildout milestones are confirmed.
Either way, what Arkady Volozh has built from the wreckage of Yandex’s international business is one of the most remarkable pivots in tech history. Whether Nebius becomes the next AWS or the next cautionary tale depends on what happens in the next twelve months.
What is Nebius Group and how did it start?
Nebius Group N.V. is an Amsterdam-based AI infrastructure company that emerged from the international assets of Russia’s Yandex. After Yandex sold its Russian operations in July 2024 due to sanctions, the remaining non-Russian businesses, including a Finnish data center and the Nebius AI cloud unit, were renamed Nebius Group. The company resumed NASDAQ trading in October 2024 under ticker NBIS, led by Yandex co-founder Arkady Volozh who returned as CEO.
What are Nebius’s biggest customer contracts?
Nebius has two landmark deals. Meta Platforms signed a five-year commitment worth up to $27 billion for dedicated AI cloud capacity powered by Nvidia’s Vera Rubin platform, with deliveries starting in early 2027. Microsoft committed up to $19.4 billion for GPU compute from Nebius’s New Jersey data center. Combined, these contracts represent approximately $46 billion in committed revenue.
How does Nebius compare to CoreWeave?
Both are GPU-accelerated cloud platforms (neoclouds) competing for AI infrastructure contracts. CoreWeave is larger by market cap (~$40.7B vs. ~$25.2B) and current revenue (~$5.1B vs. $530M), but Nebius is growing faster (479% vs. ~168% YoY) and has stronger geographic diversification with significant European operations. Nebius also secured a direct $2 billion equity investment from Nvidia, giving it preferred GPU access.
What is the biggest risk for Nebius investors?
The most significant risk is the capital structure. Nebius plans $16 to $20 billion in 2026 CapEx against $3 to $3.4 billion in guided revenue. Approximately 40% of the funding still needs to be raised through additional debt, asset-backed financing, or equity offerings. The company has already issued $4.6 billion in convertible notes and authorized an ATM equity program, meaning dilution is built into the growth plan. If AI demand softens or contracts get delayed, the financing stack could become strained.
What do analysts predict for NBIS stock price?
The 8 analysts covering Nebius have a consensus Strong Buy rating with an average price target of $169, implying approximately 50% upside from the current $112.54 price. The highest target is $291, while the lowest is $108. Wedbush analyst Dan Ives named NBIS his top AI infrastructure pick for 2026, and Goldman Sachs maintains a Buy rating with a $155 target.
Disclosure: This article is for informational purposes only and does not constitute investment advice. The author and TECHi may hold positions in securities mentioned. Always conduct your own research or consult a licensed financial advisor before making investment decisions. Stock prices and analyst targets are as of the date published and subject to change.
