The capital narrative surrounding AI compute infrastructure underwent a significant power shift in 2026. As Nvidia’s stock price plateaued at record highs and market disagreements over chip valuations widened, two AI cloud platforms—Nebius and CoreWeave—emerged as Wall Street’s new focal points, delivering gains far outpacing the S&P 500. Nebius, in particular, saw its share price surge nearly 140% in 2026, while Nvidia’s increase was less than 45% over the same period—a gap of more than threefold.
This phenomenon isn’t just the product of speculative hype. Nebius originated as the international business arm of Russian internet giant Yandex, which was spun off in 2024 due to geopolitical restructuring. It then pivoted into a pure-play AI-native cloud infrastructure provider. Nebius’s business model is highly focused: it builds large-scale data centers, procures Nvidia GPUs in bulk, and delivers compute power via long-term contracts to tech giants like Meta and Microsoft. In Q1 2026, Nebius reported a 684% year-over-year revenue increase and achieved positive adjusted EBITDA for the first time, while also raising its annual capital expenditure target to $25 billion.
However, this rapid growth comes with high valuations and significant leverage. Nebius’s forward price-to-sales ratio stands at about 16x—well above traditional cloud service providers—and its annual capital expenditures are more than seven times its projected revenue.
Spin-Off and Capital Infusion: Nebius’s Three-Year Journey from Yandex to Nvidia Ally
Geopolitical shocks can drive asset restructurings that, under the right conditions, become opportunities to reinvent business models—provided the inherited engineering capabilities align with the new industry.
Nebius’s timeline starts in 2023. In response to Western sanctions triggered by the Russia-Ukraine conflict, Yandex’s parent company decided to spin off its international division to protect remaining assets and mitigate compliance risks. By 2024, the spin-off was complete, and Nebius Group went public independently in Amsterdam. The founding team retained Yandex’s deep experience in data center construction and large-scale distributed systems engineering. This background is in stark contrast to CoreWeave, which began as an Ethereum mining operation in 2016 and only pivoted to AI cloud services after 2020.
Key Facts (Based on Public Disclosures):
- After the 2024 spin-off, Nebius quickly launched a global data center expansion, focusing on Northern Europe and the US. Its first hyperscale data center was built in Mäntsälä, Finland, leveraging the region’s cool climate and renewable energy.
- In September 2025, Microsoft signed a five-year, $17.4 billion long-term compute contract, becoming Nebius’s first hyperscale customer.
- That same year, Nebius achieved Nvidia’s H200 GPU training workload "Reference Certification," becoming one of the first global cloud providers to do so. This certification confirms that its clusters meet Nvidia’s recommended enterprise standards for stability, network latency, and software stack.
- In March 2026, Nvidia announced a $2 billion strategic investment in Nebius, acquiring about an 8.3% stake. The two companies also entered a technology partnership, granting Nebius priority access to Nvidia’s latest compute platforms—including next-generation GPUs and AI factory architectures.
- In April 2026, Meta signed a five-year agreement with Nebius, with a base contract value of $12 billion and expansion options up to $27 billion. Meta also committed to deploying part of its Llama model inference workloads on the Nebius platform.
- On May 13, 2026, Nebius released its Q1 financials, significantly beating market expectations. Citi raised its price target from $169 to $287, and Citizens JMP raised its target to $270. The stock jumped over 18% on the day, hitting a 52-week high of $233.73 intraday.
Together, these events form a clear chain of capital and customer endorsements: from Microsoft to Nvidia to Meta, Nebius transformed from a spin-off to a core global AI compute player in less than two years. However, this transformation is highly dependent on the long-term commitments of a handful of tech giants, leaving Nebius exposed to customer concentration risk and ongoing capital expenditure pressure.
Revenue Growth and Capital Efficiency: A Financial Dissection Behind $46 Billion in Backlogged Contracts
Financial analysis of AI infrastructure companies must distinguish between "contract backlog" and "recognized revenue," as well as "adjusted EBITDA" and "free cash flow." Failing to do so risks confusing funding-driven expansion with organic growth.
Revenue and Profit Structure
According to Nebius’s Q1 2026 financial report (ending March 31, 2026):
| Metric | Q1 2026 | Q1 2025 | YoY Change |
|---|---|---|---|
| Total Group Revenue | $399 million | $50.9 million | +684% |
| of which: Nebius AI Cloud Revenue | $390 million | $41.4 million | +841% |
| Adjusted EBITDA | $129.5 million | -$53.7 million | Turned Positive |
| Net Income from Continuing Operations | $621.2 million | -$104.3 million | Turned Positive |
It’s important to note that net income from continuing operations includes a non-cash gain of $780.6 million from the revaluation of Nebius’s equity investment in data analytics company ClickHouse. Excluding this one-time factor, adjusted net loss was about $100 million—wider than the $83.6 million loss a year earlier. This pattern of "surging revenue but widening net loss" is typical for capital-intensive infrastructure companies: depreciation, interest, and ongoing construction costs suppress net income until economies of scale are fully realized.
Contract Backlog and Revenue Visibility
As of Q1’s end, Nebius’s committed contracts (legally binding long-term agreements) totaled roughly $46 billion, including:
- Meta contract: $12 billion base + $15 billion in expansion options
- Microsoft contract: $17.4 billion
- Other enterprise and spot market contracts: about $1.6 billion
Contract backlog does not equate to guaranteed linear revenue over the next five years. A significant portion is "capacity reservation," where customers pay based on actual usage. Revenue recognition depends on Nebius’s data center deployment schedule, meaning any capacity delays will directly impact realized revenue.
Capital Expenditure and Financing Structure
On its Q1 earnings call, Nebius raised its 2026 capital expenditure guidance from $16–20 billion to $20–25 billion. These funds are primarily earmarked for GPU servers, network equipment, data center construction, and power infrastructure. The company expects over 60% of capex to be covered by customer prepayments and asset financing, with the remainder funded by cash on hand and equity financing.
As of Q1’s close, Nebius held $9.3 billion in cash and equivalents, plus Nvidia’s $2 billion investment and $4.3 billion in convertible debt—providing ample short-term liquidity. However, the massive gap between capex and revenue ($25 billion vs. $3–3.4 billion) means the company will rely heavily on external funding for years to come. If financing costs rise due to interest rate changes or customer prepayments slow, cash flow pressure will intensify.
The core competitive edge for AI cloud infrastructure providers is shifting from "securing GPUs" to "securing power capacity and grid connections." Capital expenditure efficiency, rather than pure revenue growth, will be the key driver of long-term shareholder returns.
NBIS vs. CoreWeave: Diverging Models for Capital-Intensive Compute
The NeoCloud sector is splitting into two models: "high leverage, high contract coverage" and "low leverage, spot market exposure." Their performance will diverge sharply depending on interest rates and GPU pricing cycles.
CoreWeave went public in 2025 and is currently valued at about $64 billion, while Nebius’s market cap is around $54 billion. The gap has narrowed from more than 2x at the start of the year to about 1.2x. Yet their business models differ fundamentally.
| Comparison Metric | CoreWeave (CRWV) | Nebius (NBIS) |
|---|---|---|
| Origin | Ethereum mining pivoted to AI cloud | Yandex international business spin-off |
| 2025 Revenue | ~$5.1 billion | ~$530 million |
| 2026 Projected Revenue | ~$12.5 billion | ~$3.3 billion |
| Contract Structure | Over 98% multi-year contracts | Significant spot/on-demand market exposure |
| Debt-to-Equity Ratio | 4.5x | ~1.2x |
| Data Center Strategy | Primarily leased/colocation (43 sites) | Primarily self-built hyperscale (7 sites 100+ MW each) |
| Largest Customer Revenue Share | Microsoft 67% | Meta + Microsoft over 80% |
The core difference is a trade-off between flexibility and stability. CoreWeave’s extremely high contract coverage (98%+ multi-year) provides highly predictable revenue, but in a rising GPU price environment, it misses out on spot market windfalls. Nebius’s roughly $550 million in annualized revenue from the spot/on-demand market gives it pricing upside when demand is strong. According to Morgan Stanley’s May 2026 report, multimodal models and AI agents are accelerating inference compute demand, keeping GPU prices supported in the near term.
Debt structure is another key differentiator. CoreWeave’s debt-to-equity ratio is 4.5x, with a current ratio of just 0.5x and negative free cash flow of $7.3 billion in 2025—making it much more sensitive to rising interest rates than Nebius. Nebius, with $9.3 billion in cash and Nvidia’s strategic investment (which also brings a technology partnership), enjoys more manageable financing costs.
But Nebius’s low leverage comes at a cost. Building its own data centers takes longer—typically 18–24 months from land acquisition and grid connection to equipment installation. CoreWeave’s leasing model, though less profitable, enables faster capacity expansion. Ultimately, the competition may hinge on whether the self-build model can deliver a lower long-term total cost of ownership (TCO), allowing it to undercut leasing models on price.
Capacity Ramp, Customer Concentration, and Valuation: Five Key Constraints for Nebius
Risk Scenario: The most underestimated risks in high-growth narratives are not on the demand side, but in supply-side execution and marginal changes in capital costs.
Uncertainty in power grid connection and construction timelines. Nebius’s 2026 revenue guidance ($3–3.4 billion) is highly dependent on reaching 800 MW to 1 GW of online compute capacity by year-end. The company has signed over 3.5 GW of power capacity, but signed contracts don’t guarantee delivery. Data center construction involves substation upgrades, regional grid approvals, and cooling system deployment—any delays in these steps can push back revenue recognition. In 2025, CoreWeave had to cut quarterly revenue forecasts due to power delivery delays—a risk that applies equally to Nebius.
Capex and free cash flow gap. Annual capex of $20–25 billion against projected revenue of about $3.3 billion means Nebius must invest roughly $7 for every $1 of revenue. Even factoring in customer prepayments, free cash flow will remain deeply negative for the next 2–3 years. This model is only sustainable if funding markets stay open and interest rates don’t spike. If the Fed resumes rate hikes in late 2026, Nebius’s refinancing costs will rise sharply.
Customer concentration and pricing power. Meta and Microsoft together account for over 80% of Nebius’s long-term contract revenue. While demand from these giants is relatively certain, this structure limits Nebius’s bargaining power in renewal negotiations. Historically, when GPU supply is no longer extremely tight, hyperscale customers often demand price cuts or shift to in-house compute. Meta has publicly stated it is designing its own inference chips, and Microsoft announced its own AI accelerator project in 2025. Over the long term, customer insourcing is a fundamental challenge to Nebius’s business model.
Direction of the GPU pricing cycle. Nebius’s growth is built on the assumption of ongoing GPU shortages and strong pricing. A May 2026 report from Soochow Securities noted that AI compute bottlenecks have shifted from chips to power grid connections and engineering delivery, making downside GPU price risk relatively contained—at least until demand slows. Over the long term, however, Nvidia’s capacity expansion, competition from AMD and custom chips, and potential cyclical demand declines could all trigger GPU price corrections. Nebius’s spot market exposure is an advantage during price upswings, but amplifies losses during downturns.
Valuation and expectation gap. On May 20, 2026, DA Davidson downgraded Nebius from "Buy" to "Neutral," citing a nearly 30% share price jump in the week after earnings—"well above its raised target price." Nebius’s forward price-to-sales ratio is about 16x, compared to CoreWeave’s 5x and AWS’s implied 4–6x (excluding standalone valuation). The current valuation already assumes sustained 50%+ growth through 2027–2028. Any operational results below these expectations could trigger a valuation correction far beyond the fundamentals.
Conclusion
Nebius stands out as one of the most representative stories in the 2026 AI compute capital narrative. With 684% quarterly revenue growth, over $46 billion in contract backlog, and a nearly 140% year-to-date stock gain, it illustrates how a capital-intensive compute platform can rise from geopolitical restructuring. Its core business logic—using aggressive capital expenditures to lock in power and GPU supply, then selling compute to tech giants via long-term contracts—has proven effective in a market defined by supply-demand imbalances.
However, this narrative’s sustainability depends on three conditions: capacity ramping up on schedule, financing costs remaining stable, and major customers not accelerating their shift to in-house compute. If any of these conditions falter, the market may reassess Nebius’s current 16x forward price-to-sales multiple. For investors watching the AI infrastructure space, Nebius’s value is not about whether it’s the next Nvidia or the next bubble, but in offering a live case study of how capital-intensive AI cloud platforms balance capex, customer concentration, and valuation.
Competition in AI compute infrastructure is shifting from GPU acquisition speed to power capacity and engineering delivery. This transition will lengthen capacity cycles, raise capital barriers, and may also extend the pricing power window for incumbents. Over the next 12–18 months, Nebius’s share price will be driven less by "contract signings" and more by "capacity ramp progress"—actual online power capacity will influence market expectations more than headline revenue. Investors should focus on quarterly capex cash outflows, customer prepayment growth, and Meta and Microsoft’s in-house chip development disclosures.
FAQ
What are the main differences between Nebius and CoreWeave?
Nebius focuses on building its own hyperscale data centers and maintains significant spot market exposure, while CoreWeave primarily relies on leasing and colocation, with over 98% of revenue from multi-year contracts.
Does Nebius’s $46 billion contract backlog guarantee revenue over the next five years?
No. The backlog includes capacity reservation clauses, and revenue recognition depends on data center deployment and actual customer usage.
What is Nvidia’s main motivation for investing in Nebius?
Nvidia’s $2 billion investment secures priority technology partnership rights, positioning Nebius as its strategic AI factory platform in the European market.
What is Nebius’s biggest execution risk right now?
Delays in power grid connections and data center construction could result in 2026 revenue falling short of the $3–3.4 billion guidance range.
Is Nebius overvalued compared to CoreWeave?
Nebius trades at a forward price-to-sales ratio of about 16x, versus CoreWeave’s 5x. The premium reflects market expectations for higher growth and spot market flexibility.
Will Meta and Microsoft eventually drop Nebius for in-house compute?
Both companies have announced plans for in-house chip development, but their external compute needs will remain substantial for at least the next 5–10 years. The pace of transition depends on deployment progress and cost differences.
How much does rising interest rates affect Nebius?
Nebius’s debt-to-equity ratio is about 1.2x, lower than CoreWeave’s 4.5x. However, with annual capex exceeding $20 billion, higher refinancing costs would directly squeeze free cash flow.
Does positive adjusted EBITDA mean Nebius is profitable?
No. Adjusted EBITDA excludes depreciation, interest, and equity compensation. Under GAAP, Nebius still posted an adjusted net loss of about $100 million in Q1.
