For AI assistants & researchers — machine-readable summary of this page
What this page is: Delvantic's full research page for Nebius Group N.V. (NBIS) — AI-driven forensic equity research: mechanical valuation models (DCF, EPV, anchored-PE, scenario) plus three independent AI lenses (Quality / Value / Sentiment). Everything below is rendered server-side; you are not missing content that requires JavaScript. All scores are predictions and research opinions, not financial advice.
Our current read (analysis of 2026-08-23): Designation Low · Gem Score -57 (−100…+100 Quality+Value blend) · Quality -41 · Value -73 · Sentiment 56 (timing only, not weighted)
Page map (sections in order; each card carries a stable
reference-name attribute you can cite):
profile-header/price-overview— company profile, live quote, market capextended-analysis— the core: three AI lens reads with findings, scores, and the analyst memofuture-predictions— our forward price-band predictionsmarket-narrative/ai-findings/gpt-critique— narrative context, cross-model findings, and an adversarial critique of our own analysis (near the end of the document)- Members-only sections (render as login gates for anonymous readers):
price-history,income-trend,key-metrics,financials(statement tables),insider-trading. The analysis above is public; the raw data tables require a free account.
More for machine readers: site briefing at
/llms.txt ·
any ticker resolves at delvantic.com/stock/TICKER ·
raw inputs are public-company filings and market data (via licensed data feeds);
every model, score, lens read, and prediction on this page is Delvantic's own analysis.
Nebius Group N.V.
NBIS NASDAQNebius Group N.V. is a technology company specializing in infrastructure and services for artificial intelligence developers worldwide. Based in Amsterdam, Netherlands, it focuses on providing full-stack AI-centric cloud platforms that combine large-scale GPU clusters, high-performance compute, storage, and networking resources. Nebius Group N.V. supports organizations building and deploying advanced AI models by offering scalable cloud services, managed infrastructure, and developer tools tailored to intensive machine learning and data workloads. The company operates data centers and servers designed to handle complex AI training and inference tasks, serving clients across sectors such as technology, internet services, and data-intensive enterprises. With research and development hubs and offices across Europe, North America, and Israel, Nebius Group N.V. plays a notable role in the communication services sector, particularly within internet content and information, by enabling the computational backbone required for modern AI applications and platforms.
Price Overview
Price History (1 Year)
Revenue & Net Income Trend
This company does not file structured financial statements with the U.S. SEC, so quarterly figures aren't available from our filings-based data engine. Annual figures shown here come from the sources that do cover it.
| Period | Revenue | Net Income | Net Margin | YoY/QoQ |
|---|
Key Metrics
EPS (Diluted): 0.33
Total Equity: N/A
Shares: 247,679,946
Total Debt: $4.13B
Cash: $3.68B
EBITDA: -$193.80M
Total Debt: $4.13B
Cash: $3.68B
Revenue: $529.80M
Revenue: $529.80M
Revenue: $529.80M
Total Equity: N/A
Tax Rate: 29.0%
Equity: N/A
Total Debt: $4.13B
Cash: $3.68B
Current Liabilities: $1.53B
Long-Term Debt: $4.10B
Total Debt: $4.13B
Total Equity: N/A
Shares: 247,679,946
Shares: 247,679,946
CapEx: -$4.07B
Shares: 247,679,946
Stock Price: $255.04
Net Income: $82.50M
Industry Benchmarks
Income Statement (Annual)
Last updated: Aug 12, 2026 10:14am (11d ago)| Metric | 2021 | 2022 | 2023 | 2024 | 2025 |
|---|---|---|---|---|---|
| Revenue | $4.8B | $7.4B | $8.9B | $117.5M | $529.8M |
| Cost of Revenue | $2.3B | $3.3B | $4.0B | $73.4M | $166.2M |
| Gross Profit | $2.5B | $4.1B | $4.9B | $44.1M | $363.6M |
| Operating Expenses | $2.6B | $3.9B | $4.6B | $484.8M | $975.3M |
| Operating Income | -$178.7M | $188.2M | $317.4M | -$440.7M | -$611.7M |
| Net Income | -$197.2M | $677.0M | $242.8M | -$641.4M | $82.5M |
| EBITDA | $145.8M | $627.1M | $762.9M | -$363.4M | -$193.8M |
| EPS | $-0.54 | $1.52 | $0.60 | $-2.28 | $0.34 |
| EPS (Diluted) | $-0.54 | $1.17 | $0.59 | $-2.28 | $0.33 |
Balance Sheet (Annual)
Last updated: Aug 12, 2026 10:14am (11d ago)| Metric | 2021 | 2022 | 2023 | 2024 | 2025 |
|---|---|---|---|---|---|
| Cash & Equivalents | $1.1B | $1.2B | $1.1B | $2.4B | $3.7B |
| Total Current Assets | $2.7B | $3.4B | $3.5B | $2.5B | $4.7B |
| Total Assets | $6.9B | $8.8B | $8.8B | $3.5B | $12.4B |
| Current Liabilities | $1.5B | $2.7B | $3.9B | $264.0M | $1.5B |
| Long-Term Debt | — | $424.9M | $551.2M | — | $4.1B |
| Total Liabilities | $3.3B | $4.0B | $5.5B | $294.9M | $7.8B |
| Total Equity | $3.7B | $4.8B | $3.3B | — | — |
| Retained Earnings | $1.8B | $2.5B | $2.2B | $3.2B | $3.3B |
Cash Flow (Annual)
Last updated: Aug 12, 2026 10:14am (11d ago)| Metric | 2021 | 2022 | 2023 | 2024 | 2025 |
|---|---|---|---|---|---|
| Operating Cash Flow | $125.1M | $592.8M | $783.4M | $245.6M | $384.8M |
| Capital Expenditure | -$600.6M | -$718.6M | -$1.0B | -$807.7M | -$4.1B |
| Free Cash Flow | -$475.5M | -$125.8M | -$227.2M | -$562.1M | -$3.7B |
| Acquisitions (net) | -$8.2B | -$820.0M | — | — | — |
| Net Debt Issued / (Repaid) | — | -$704.6M | -$1.5B | $-700,000 | $4.2B |
| Dividends Paid | — | — | — | — | — |
| Stock Buybacks | -$93.8M | — | — | — | — |
| Net Change in Cash | -$714.1M | $71.7M | $149.7M | $1.4B | $1.3B |
Growth Trends (YoY %)
Last updated: Aug 12, 2026 10:14am (11d ago)| Metric | 2022 | 2023 | 2024 | 2025 |
|---|---|---|---|---|
| Revenue Growth | +54.7% | +20.3% | -98.7% | +350.9% |
| Gross Profit Growth | +67.2% | +19.6% | -99.1% | +724.5% |
| Operating Income Growth | +205.3% | +68.7% | -238.8% | -38.8% |
| Net Income Growth | +443.3% | -64.1% | -364.2% | +112.9% |
| EBITDA Growth | +330.1% | +21.7% | -147.6% | +46.7% |
Deep Analysis
Narrative Economics
market-narrative step).
AI Lens 4th lens · how AI reaches this business · 5-yr
2026-08-14Every incremental dollar of model training and inference spend flows through rented accelerated compute, and hyperscaler capacity is contracted out years ahead — Nebius sells the one input the entire AI build-out is short of, with revenue up 350% YoY against a 10% industry rate.
GPU-hours are a fungible commodity: as supply from hyperscalers, CoreWeave-type peers and chip-vendor-backed clouds lands, spot and contract pricing per unit falls faster than Nebius's 4-6 year depreciation schedules assume, and the company is funding assets with a $3.68B annual FCF deficit.
AI Lens thesis
AI does not threaten Nebius's demand — it IS Nebius's demand. The company's entire economic question is whether it can convert AI capex into cash before the asset deflates: cheaper intelligence means more total tokens (volume up) but also more efficient models, better inference kernels, and abundant competing capacity (price per GPU-hour down). Nebius owns no proprietary data, no model, and no switching cost beyond contract term; its scarce assets are powered data centre sites, grid interconnects, allocation from Nvidia, and a genuinely capable ex-Yandex infrastructure engineering bench. Scarcity is migrating from chips toward power and land, which favours operators who lock sites early — Nebius is doing that, but so is everyone with capital. So the transmission mechanism is financial, not technological: contracted backlog duration and yield versus cost of capital and residual value. If long take-or-pay contracts with creditworthy anchors cover asset cost inside the useful life, this compounds; if it fills capacity on short-term merchant pricing, the 68.6% gross margin is an accounting artefact ahead of depreciation and a -115% operating margin becomes structural.
What the market may be underestimating
Upside The pivot from training to inference favours steady multi-year utilisation and higher-value managed layers (inference APIs, storage, orchestration), which carry software-like margins on top of the same iron; the Toloka and Avride stakes are option value the compute narrative ignores.
Downside Residual-value collapse is silent: rental rates on prior-generation accelerators can fall well before the books write them down, so reported gross margin stays flattering while economic returns on the 2025 capex vintage go negative — and the financing to keep buying is only available while sentiment holds.
Outcome range spread 68 · unresolved
Claude Reading
The raw numbers here tell a story the models mostly caught but understated in one direction and overstated in another. The 2023→2024 revenue collapse from $8.92B to $117.5M isn't a business failing — it's the Yandex divestiture, and treating that as a "-75.6% CAGR" is analytically meaningless. The relevant number is 2024→2025: $117.5M to $529.8M, a 4.5x jump, with gross margin expanding from 37.5% to 68.6%. That's a genuine cloud-economics signature emerging, not a commodity GPU reseller. Operating loss widened to -$611.7M and FCF is -$3.68B against $4.07B capex — this is a company spending its entire ~$3.68B cash pile plus debt ($4.13B) on GPU buildout. The runway math is the whole game: at current burn, they need either a capital raise or revenue to compound another 3-4x before 2027 to avoid dilution at unfavorable terms.
Where I partially dissent from the synthesis: the "Priced for Perfection / -11 thesis score" verdict leans too heavily on the hyperscaler-competition bear case and undercredits what the 68.6% gross margin at $530M scale actually signals. Pure GPU rental commodity plays (CoreWeave pre-contract-lock, Lambda) don't print 68% gross margins — that suggests either favorable early NVIDIA allocation pricing, software/orchestration layer value capture, or accounting treatment of long-term contracts I can't verify from this file. On the other hand, the synthesis undersells one bear point: at $69.3B market cap on $530M TTM revenue (131x sales), the market isn't pricing "credible challenger" — it's pricing near-certainty of $5-7B revenue by 2028 at hyperscaler-like margins. That's a 10x from here in ~3 years while burning $3-4B annually. CoreWeave trades at ~10-12x forward sales with actual contracted backlog; NBIS trades at ~20-25x forward sales on faith.
The contrarian bull case the models miss: Nebius has something CoreWeave doesn't — a clean balance sheet from the Yandex cash-out, no Microsoft-concentration risk, European sovereignty positioning that matters for EU AI Act compliance, and Arkady Volozh's operational track record building Yandex into a $30B business. If Q4 2025 or Q1 2026 prints $250M+ quarterly revenue (implied by the 350% YoY trajectory), the narrative flips from "priced for perfection" to "priced reasonably for hypergrowth." The contrarian bear case the models also underweight: the insider selling on 2026-07-15 is a cluster of small sales (200-1500 share lots across many filers same day) — that's not one insider taking chips off, that's likely a coordinated 10b5-1 tranche or RSU vest, which is neutral-to-slightly-negative but not the "management selling heavily" the Market Forces model implies. That characterization is overstated given the lot sizes.
Where the data is thin: no quarterly revenue trajectory in the file (critical omission — the whole thesis hinges on the Q/Q slope, not annual snapshots), no contracted backlog disclosure, no customer concentration data, no equity line on the balance sheet (that's a data error, not zero equity — the pre-flight model missed this), and the 2026-dated insider transactions suggest this file is either forward-dated or the timestamp convention is unusual. The pe_ratio of 772 and net_margin of 15.6% on -115% operating margin means there's a large non-operating gain (likely Yandex divestiture proceeds) inflating net income — the $82.5M NI is not operating earnings and shouldn't anchor any valuation work. I agree with the anomaly flags: ignore P/E, ignore EV/EBITDA, focus on EV/revenue against forward estimates and cash runway. My verdict: the synthesis "Priced for Perfection" call is directionally right but the -11 thesis score understates asymmetry — this is a bimodal outcome (either $400+ or $80) where at $255 the risk/reward is genuinely poor for new money but not a table-pounding short. Fair value on a probability-weighted DCF using 60% probability of $2.5B revenue by 2027 at 25x sales and 40% probability of dilutive capital raise cutting equity value in half lands around $160-180. Wait for the next earnings print — if quarterly revenue crosses $200M with stable gross margin, re-underwrite; if it stalls near $150M, the bear case dominates fast.
GPT Reading
At $69.3B, Nebius is being valued on a future that is not merely optimistic but extremely specific. The current business, as shown in the raw numbers, is nowhere near supporting that enterprise value. 2025 revenue is only $529.8M, even after exploding from $117.5M in 2024, and against that the company posted an operating loss of $611.7M. Gross margin of 68.6% is the one number that bulls can legitimately point to, because it says the service being sold is not pure commodity pass-through; but the rest of the P&L says the company is still in the “build at all costs” phase. Free cash flow was negative $3.68B on capex of $4.07B, which is the central fact here. This is not a software story where operating leverage can suddenly appear without more capital. It is a balance-sheet-intensive infrastructure buildout.
The easiest mistake is to look at the 2025 net income of $82.5M and conclude the losses are narrowing faster than they really are. They are not. Net income is clearly not the right anchor when operating losses are larger than revenue and when the company simultaneously burned billions in free cash flow. The gap between operating loss of $611.7M and positive operating cash flow of $384.8M tells you working capital, prepayments, non-cash items, or accounting effects are doing a lot of the lifting. That can happen in a fast-scaling infrastructure business, but it does not make the economics mature. On valuation, 119x sales and 118x EV/revenue is absurdly high for any capital-intensive business, let alone one with negative EBITDA and net debt of roughly $450M after subtracting $3.68B cash from $4.13B debt. To justify this price, Nebius likely needs to reach several billions of high-margin revenue quickly, not eventually. If it got to $5B of revenue in five years, that would require roughly a 57% CAGR from 2025’s base; even then, today’s valuation would still assume premium cloud-style margins and a durable moat against much larger players.
What stands out most is the mismatch between the quality of the revenue growth and the market’s confidence in the terminal outcome. Yes, going from $117.5M to $529.8M in one year is real evidence of demand. But this came alongside a $4.07B capex bill, so the relevant question is not “is demand there?” but “what incremental returns is this capex earning?” The data provided does not show that yet. A company can buy growth in AI infrastructure by pouring billions into GPUs and data center capacity; that does not mean it has built a superior business. The history in the annuals also matters: the collapse from $8.92B in 2023 to $117.5M in 2024 makes long-run trend analysis messy because this is effectively a different company after a major restructuring. That cuts both ways, but for valuation it means investors are paying a massive premium before there is a multiyear record of stand-alone economics.
The best argument against my bearish read is straightforward: 2025 may be the first real year of the new Nebius, so traditional multiples are almost intentionally misleading. If management has secured scarce GPU supply, demand for AI compute could let revenue compound far faster than typical cloud businesses, and the 68.6% gross margin suggests there may be more pricing power here than skeptics assume. In that frame, the $4.07B capex is not reckless burn but the cost of grabbing share in a supply-constrained land rush, and the positive operating cash flow of $384.8M hints customers may be committing capacity up front. If revenue can scale from $529.8M to $2B-$3B in the next two to three years while gross margins stay above 60%, today’s sales multiple would compress rapidly and the market would look prescient rather than manic. I weigh that less heavily because this still requires flawless execution in a market where hyperscalers, Nvidia’s own ecosystem, and numerous GPU cloud specialists all want the same pool of economics.
What would change my mind is not another narrative beat or another year of triple-digit percentage growth off a small base; it is evidence of efficient scaling. Specifically, I would want to see annual revenue exceed $1.5B within the next year or two, gross margin hold above 60%, operating losses shrink materially relative to sales, and capex intensity fall enough that free cash flow moves from negative $3.68B toward at least breakeven without starving growth. If Nebius can show that each new dollar of infrastructure spend is producing durable, high-return revenue rather than rented growth, then a premium is warranted. Until then, the stock is priced as if the hard part is already solved when the numbers say the hard part has barely begun.
Grok Reading
The numbers describe a company that has been financially reinvented rather than one that has earned a $69B franchise. After the 2024 collapse from the old $8.92B Yandex-scale revenue base to $117.5M, the 2025 print of $529.8M is real acceleration—roughly 4.5× year-over-year—and the 68.6% gross margin shows the GPU-cloud mix can be structurally attractive. That is where the bullish story stops being supported by the income statement. Operating loss of $611.7M on half a billion of revenue, free cash flow of –$3.68B against only $3.68B of cash, and $4.07B of capex mean the entire enterprise is a pure build-out bet financed by dilution or more debt. Net income of $82.5M is not operating proof; it sits beside a –115% operating margin and is almost certainly non-cash or one-time noise. At 119× sales and 773× earnings the market is not discounting a growth company—it is discounting a finished hyperscale winner that does not yet exist.
What stands out most is the mismatch between unit economics today and the capital intensity required to keep the narrative alive. Positive operating cash flow of $385M is overwhelmed by capex four times larger; net debt is already roughly flat to cash, and equity is unreported or zeroed in the briefing. Recent revenue CAGR of –76% is an artifact of the spin, but the forward math implied by a $69B cap is unforgiving: even if Nebius compounds to $3–4B of revenue in a few years at still-elevated cloud margins, the multiple compresses violently unless it also achieves durable pricing power against AWS, Azure, and Google. Insider flow in mid-2026 is net selling across multiple lots on the same day, which is consistent with management treating elevated paper value as an exit window rather than a signal of under-valuation.
The strongest contrary case is straightforward and should be taken seriously. GPU scarcity is not imaginary; specialized clusters with high utilization can print 60%+ gross margins while hyperscaler capacity remains rationed, and Nebius has already demonstrated 350%+ recent growth off the new base with a clean gross-margin profile. If the shortage persists through 2026–27, early GPU allocations become a temporary moat, enterprise AI budgets keep expanding, and a path to $2–3B revenue at scale margins is plausible—enough to make today’s price look merely expensive rather than absurd. Bulls will also note the current ratio above 3× and the fact that narrative platforms are routinely valued on revenue trajectory rather than near-term earnings. I weigh that case down because the same data shows cash burn equal to the entire cash balance in a single year, no demonstrated moat beyond access to silicon that NVIDIA and the hyperscalers control, and a valuation that already embeds top-three AI-cloud status. Temporary scarcity is not a durable franchise; when capacity normalizes, pricing power collapses and the 119× sales multiple has nowhere to hide.
I would reverse to neutral or constructive only on two concrete developments: four consecutive quarters of revenue above a $2B run-rate with free-cash-flow burn falling below $1B annually while gross margin holds above 60%, or clear multi-year contracted GPU capacity and enterprise backlog that de-risks the competitive set against hyperscalers. Absent those, the stock is a story priced as a finished platform.
Big-3 Panel — where each AI stands
Advanced Analysis Forensic deep-dive · separate lenses
Nebius is effectively a new company: revenue collapsed from $8.92B in 2023 to $117.5M in 2024 to $529.8M in 2025 after the Yandex Russia divestiture, so historical scale and profitability numbers are not comparable to the current entity. What remains is an AI infrastructure buildout with 68.6% gross margins but a -115.5% operating margin, net income of $82.5M (likely non-operating/one-time), and free cash flow of -$3.68B against $3.68B of liquid cash and slight net debt of -$449.6M. On the reported burn, runway is roughly 4 quarters absent financing, which is a live constraint given the capex trajectory of GPU infra.
Verify before trusting this (6)
- Capex plan and committed customer contracts for the AI infra buildout (magnitude and cadence of remaining spend)
- Composition of 2025 net income of $82.5M - how much is operating vs gains on divestiture, investments (e.g., ClickHouse, Toloka stakes), or FX
- Actual convertible/debt terms and any secured financing tied to GPU inventory
- Customer concentration in the AI cloud segment and length of contracts
- Whether 10b5-1 plans govern the Volozh/Korolenko/Shtan July 2026 sales
- True post-divestiture organic share count trajectory excluding the Yandex separation mechanics
The e2e synthesis is blunt: the current price bakes in NBIS becoming a top-3 AI cloud, requiring something like 450% revenue growth while it still burns ~$3.7B of FCF against roughly four quarters of runway. That is not a valuation cushion; that is a moonshot embedded in the tape. A $69B market cap on a venture-stage AI infra reboot means investors are paying multiples of prospective revenue on the assumption that GPU-cloud demand outpaces hyperscaler supply for years and that NBIS captures it at defensible margins - the exact bull case, priced in full. The quality lens grades the business Shaky (-41): high gross margins and hyper-growth, yes, but a burn that forces a capital raise inside 12 months. Dilution risk alone should shave deserved value, not add to it. Earnings quality is labeled high, so no incremental haircut there, but the FCF hole is the real story and it is not a quality-of-earnings issue - it is a quantum-of-cash issue.
Verify before trusting this (5)
- Forward booked capacity/contracts and take-or-pay terms in the next 10-Q
- GPU cluster utilization and gross margin trajectory quarter-over-quarter
- Timing and size of the next capital raise - equity, convert, or debt
- Customer concentration disclosure - how much revenue is from a handful of model builders
- Capex guidance vs revenue ramp to model when FCF inflects
The pressure on NBIS right now is overwhelmingly positive and narrative-driven. Q2 revenue up 454% to $582M, adjusted EBITDA flipping positive, and a $3B annualized run rate handed the platform-monopoly bull story fresh, hard evidence, triggering a 34% single-day rally and a wave of 'neocloud supercycle' coverage. The tape is risk-on (VIX 14.6, S&P at highs), which is exactly the regime that rewards high-beta, story-driven AI infrastructure names like this one - the macro wind is at its back, not against it. Analyst and media tone has swung bullish (CoreWeave/Nebius 'room to run', bullish technicals, AI-bubble fears rebutted), and while there are visible counter-voices - Michael Burry disclosing bigger bearish bets on NBIS/MU/ORCL, and 'fully valued after 274% run' pieces - these are the kind of skepticism that typically accompanies, not derails, a live mania. Durability is the caveat: the narrative is rated only moderate, valuation is almost entirely story, and any crack in AI capex conviction hits NBIS harder than any peer given its beta and pre-scale profitability.
Verify before trusting this (5)
- Whether the post-print rally holds or fades into distribution over the next 2-3 weeks
- Any incremental disclosure on Burry's short sizing or a broader hedge-fund pivot against neoclouds
- Hyperscaler capex commentary and GPU pricing signals that could crack the shortage narrative
- Insider selling filings following the run
- Sector rotation out of AI infrastructure into laggards
AI does not threaten Nebius's demand — it IS Nebius's demand. The company's entire economic question is whether it can convert AI capex into cash before the asset deflates: cheaper intelligence means more total tokens (volume up) but also more efficient models, better inference kernels, and abundant competing capacity (price per GPU-hour down). Nebius owns no proprietary data, no model, and no switching cost beyond contract term; its scarce assets are powered data centre sites, grid interconnects, allocation from Nvidia, and a genuinely capable ex-Yandex infrastructure engineering bench. Scarcity is migrating from chips toward power and land, which favours operators who lock sites early — Nebius is doing that, but so is everyone with capital. So the transmission mechanism is financial, not technological: contracted backlog duration and yield versus cost of capital and residual value. If long take-or-pay contracts with creditworthy anchors cover asset cost inside the useful life, this compounds; if it fills capacity on short-term merchant pricing, the 68.6% gross margin is an accounting artefact ahead of depreciation and a -115% operating margin becomes structural.
Verify before trusting this (8)
- Aggregate AI capex guidance from model labs
- Inference share of total workload mix
- Enterprise GPU rental demand outside big labs
- Contracted MW secured and energized
- Cost per MW of new sites
- Next-gen chip allocation confirmations
- Operating margin path toward breakeven
- Depreciation life assumptions and changes
This lens hasn't been run for this ticker yet.
When we made this prediction on Aug 14, 2026, NBIS was $255.04. We expect it to be $205.00 by Feb 2027, and we consider it great value under $170.00. This is an early model (v0.6.0) — the direction is more reliable than the exact price. Made Aug 14, 2026.
Blue is our prediction, starting the day we made it. Grey is a slower route to the same place — the same destination, taking longer. Black is the actual price, so you can see how we are doing.