Nebius Group NBIS
Technology · Software - Infrastructure · Synthos Deep Dive · 2026-07-03
The Overview
Nebius rents out AI supercomputers. Companies that need massive banks of Nvidia chips to train and run AI models pay Nebius to use its data centers instead of building their own. It's the same idea as renting cloud space from Amazon, but purpose-built for AI. (It's the leftover "good half" of Russia's old Yandex, re-headquartered in Amsterdam and re-listed on Nasdaq.)
The business is growing at a blistering pace — revenue more than quadrupled last year and is running several times higher than a year ago. But it loses money today, because it's spending billions buying chips and building data centers faster than the revenue comes in. So this is a bet on the future, not a company you value on today's earnings (there aren't any).
Is the stock cheap or expensive? On today's numbers, very expensive — you're paying roughly 59 dollars for every 1 dollar of current sales. That only makes sense if the growth keeps compounding for years. Our verdict is Buy — but as a small "satellite" bet, money you can afford to see swing 50% either way.
Here's what our three scores mean in everyday terms:
- Downside Risk 8/10 (high). The company has plenty of cash on hand right now, but the stock is priced for perfection, burns cash, and moves violently — a bad quarter could cut it in half.
- Growth Quality 8/10 (strong). The growth is real and enormous — but it's still early, unprofitable, and unproven at large scale.
- Exponential Potential 9/10 (very high). Small company + growth that's speeding up + a gigantic market to grow into = one of the highest exponential scores we'll assign.
The one big worry: Nebius is in a capital arms race. It must keep spending billions on chips. If AI demand cools, or the money to fund that spending dries up, it may have to sell lots of new stock (diluting owners) or hit a wall.
Putting a number on it: our fair-value estimate is $235 against a current price of $209.18 — real upside if our numbers are right.
Our summary metrics
Net-cash balance sheet, but 59× EV/sales, −$3.7B FCF, beta 1.43 and a −25% drawdown make it high-volatility.
~100× forward revenue ramp ($530M→$53B by FY30E), but EPS stays negative to ~FY28 and margins unproven at scale.
Tiny base + accelerating triple-digit growth + trillion-dollar AI-infra TAM — a textbook accelerating exponential.
What does “fair value” mean?
Fair value is Synthos’s estimate of what one share is worth today, based on our model of the company’s future cash generation and the risks to it. It is not a price target or a prediction of where the stock trades next quarter — it’s the price where we think risk and reward are balanced. Above it, you’re paying for outcomes better than our base case; below it, the market is offering a margin of safety.
The model’s inputs and weightings are proprietary — they’re the product. What we publish is the output, plus an independent, public-math cross-check further down the page (“Check our number”) so you can judge it for yourself.
What do the 5 tiers mean? (Core · Tactical · Watch · Hold · Avoid)
Exponential Potential
Tiny base + accelerating triple-digit growth + trillion-dollar AI-infra TAM — a textbook accelerating exponential.
What could take this further than the base case — and where the market may be underpricing it. Full forward-growth and acceleration math in Deeper analysis, §4 below.
Deeper analysis
Technicals, fundamentals, valuation, and the full expert-claim evidence panel — the detail behind the numbers above.
Reference table
| Street consensus | $196 (high $255 / low $129; 7 Buy · 1 Hold · 0 Sell) — context, not our anchor |
| Valuation | Loss-making on EPS · 59× EV/sales TTM · ~15× FY26E sales · ~5× FY27E sales · P/B 7.7× |
| Technicals | Consolidating after a huge run — $215.62, −25% off 52-wk high, at 50-DMA, far above 200-DMA, RSI 48, +329% 12-mo (SPY +21%) |
| Conviction | Quant-only — 0 expert voices, 0 claims in the Synthos KB; nothing to reconcile |
| Position sizing | Satellite / high-risk sleeve, ~1–2.5% — a volatile moonshot, not a core holding |
What the experts actually said 4 traceable claims on NBIS · showing the highest-conviction voices
“Neo-cloud category (CoreWeave, Crusoe, Nebius, Lambda) is durable — running a GPU cluster is like driving an F1 car; their GPUs are utilized 2–3x more per hour, a rare and durable operational advantage.”
“Small AI clouds Nvidia's full-stack enabled — CoreWeave (now worth $50-70B) and Nebius — are growing incredibly fast.”
“At the ~$12-13M/MW five-year IaaS/self-build pricing across CoreWeave, Oracle and Nebius, these providers make only modest double-digit margins — the real premium comes only from on-demand emergency capacity.”
Every claim reconciles to a real claim_id in the Synthos knowledge base — this is the evidence the verdict is built on, not vibes. Management (the company itself) is shown but half-weighted; one cautionary voice is included on purpose.
Price & moving averages 12 months · 50 & 200-day averages · 52-week range
Solid line = price · dashed line = 50-day average · dotted line = 200-day average · the two thin horizontal lines mark the 52-week high and low. Price above both averages is an uptrend.
Data summary: last close $209.18, 5% below the 50-day average ($219), 38% above the 200-day average ($152) — a mixed trend. 27% below the 52-week high of $287, 227% above the 52-week low of $64.
Bollinger Bands 20-day average ± 2 standard deviations
The shaded band widens when the stock gets more volatile. Riding the upper edge = strong momentum (sometimes stretched); the lower edge = weak / potentially oversold.
Data summary: price $209.18 is currently inside the band (band $171–$275).
RSI (14) momentum gauge · 0–100
Above 70 (overbought zone, shaded) = overbought, below 30 (oversold zone, shaded) = oversold. Currently 47.
MACD 12 / 26 / 9 · trend & momentum
The MACD line crossing above the signal line (bars flip to the up color) = momentum turning up; crossing below (bars flip to the down color) = turning down. Bar height = the size of that gap.
Data summary: MACD is currently below its signal line by 2.50, negative momentum.
Relative performance vs S&P 500 & its sector (XLK (sector)), set to 100 a year ago
Solid = NBIS · dashed = S&P 500 · dotted = XLK (sector). A rising line means it is beating that benchmark — the sector line shows whether it is a leader or laggard within its own group.
Forward revenue & earnings actual → estimate · "FY" = fiscal year, "E" = estimate
Darker bars = actual results, brighter = analyst estimates. Taller bars to the right = expected growth.
Key stats an RIA wants
1. What it is
Nebius Group N.V. (NASDAQ: NBIS) is an Amsterdam-headquartered technology company building full-stack infrastructure for the AI industry. It is the corporate successor to Yandex N.V. — after divesting its Russian operations in 2024, the remaining international assets were rebranded to Nebius and re-listed on Nasdaq (IPO/relisting October 2024). CEO Arkady Volozh is the original Yandex founder. Fiscal year ends December 31.
The group runs four businesses, anchored by the flagship:
- Nebius — the core AI-focused cloud platform: large GPU compute clusters, cloud tooling, and developer services for training and inference. This is the growth engine.
- Toloka AI — data-labeling / data solutions for generative-AI development.
- TripleTen — an ed-tech venture reskilling people for tech careers.
- Avride — autonomous-driving and delivery-robot technology.
Revenue mix (FY2025, from filings):
- By segment: FMP does not provide a clean product-segment breakout for FY25 (
seg_prodis empty); management commentary attributes the overwhelming majority of the growth to the core Nebius AI-cloud platform, with Toloka, TripleTen and Avride still small. - By geography: the only geographic line FMP carries is United States $340.1M of the $529.8M FY25 total (~64%), confirming the AI-cloud demand is heavily US-driven; the remainder is Europe / rest-of-world. This is a US-customer-concentration flag (§11).
The strategic story is singular: be a Western, independent "neocloud" that supplies GPU capacity to AI labs and enterprises at the exact moment demand for training/inference compute is exploding — funded by heavy, front-loaded capex.
2. The expert thesis — why the panel is bullish (no KB coverage)
There is no expert coverage of NBIS in the Synthos knowledge base. total_claims = 0, net_bullish_voices = 0, and the top array is empty. There are no claim_ids to cite — so, per house standard, we do not manufacture any.
That means this verdict is entirely fundamentals- and quant-driven: it rests on the reported financials, the FMP analyst-estimate consensus (labeled as estimates throughout), the balance sheet, and the technical/positioning read below — not on any distilled expert conviction. Readers should weight this note accordingly: it carries less independent-voice corroboration than a conviction-track name like LLY, and the bull case leans harder on a single macro bet (AI-infrastructure demand) that no vetted expert in our panel has underwritten.
Where the Street sits (context, not our anchor): 7 Buy, 1 Hold, 0 Sell, consensus price target $196 (high $255, low $129). FMP's own letter rating is "C" (overall score 2/5) — reflecting the negative DCF, weak profitability and rich multiples that the quant model penalizes. So the sell-side is constructive while the mechanical quant screen is cautious; that tension is the NBIS debate.
3. Synthos scores & the Bull / Base / Bear cases
The one-glance judgment — three scores, 0–10, each anchored to real metrics (not probabilities we can't honestly calibrate):
| Score | 0–10 | The read |
|---|---|---|
| Downside Risk (lower = safer) | 8 · High | Net cash is a cushion (net debt just $1.3B, net-debt/EBITDA 0.14×, current ratio 8.3×), but 59× EV/sales, −$3.7B FCF, beta 1.43 and a −25% drawdown mean the equity is priced for perfection and moves violently. |
| Growth Quality | 8 · High | Revenue +351% FY25 and +624% YoY in Q1'26; consensus models a ~100× revenue ramp to ~$53B by FY30E. Docked from 9 because EPS stays negative into ~FY28E, margins at scale are unproven, and returns on capital are still negative. |
| Exponential Potential | 9 · Very High | Tiny $530M base × growth that is still accelerating × a trillion-dollar AI-infrastructure TAM vs a $52B cap. Textbook forward exponential — exactly what the flagship philosophy hunts for. |
The three cases (our own scenario model — assumptions shown; each target is a ~12–18-month fair value). We deliberately do not attach probabilities: the base case is the expected path, so a weighted blend would just restate it with false precision. The cases bound the range; the scores summarize them. Because NBIS is loss-making, we value it on forward EV/sales (the only honest lens today), not P/E.
| Case | Key assumptions | Fair value |
|---|---|---|
| Bull | AI-cloud demand stays red-hot; Nebius hits/beats the high end of FY27E revenue (~$13.4B) and proves a credible path to positive EBITDA; capacity fills as fast as it's built. Market pays a premium ~9× FY27E sales. | ~$430 (+99%) |
| Base (our anchor) | Consensus roughly holds — FY26E revenue ~$3.5B, FY27E ~$11.6B; execution is good but still cash-burning. Market pays ~5× FY27E sales on the growth, discounted for dilution/risk. | ~$235 (+9%) |
| Bear | GPU demand softens, financing tightens, or utilization disappoints; growth decelerates and the capex bill forces dilution. FY27E revenue misses toward ~$9B; multiple compresses to ~2.5× sales. | ~$95 (−56%) |
Synthos fair value = the base case, ~$235 (+9%), with the full $95–$430 span as the honest — and deliberately wide — range. That width is the message: at 59× sales the outcomes are genuinely bimodal. Our base sits above the Street's $196 consensus (we give the FY27 revenue ramp more credit) while our bear is below the Street's $129 low (we take the funding/dilution risk seriously). This is a tracked call — the Forecaster Scorecard grades it once it matures.
4. Exponential Potential
Synthos separates compounders (durable high returns on capital) from exponentials (accelerating, multi-baggers-from-here). NBIS is a near-pure exponential — small, accelerating, and early:
- Forward growth: consensus revenue ~$3.5B FY26E → ~$11.6B FY27E → ~$21.8B FY28E → ~$36.2B FY29E → ~$53.0B FY30E (estimates). That is roughly a 100× revenue expansion off the FY25 $530M base over five years — among the steepest ramps in the entire Nasdaq-100 pool.
- Acceleration (the 2nd derivative) is positive right now: revenue growth ran +351% FY25, and Q1'26 revenue of $399M was +624% YoY — the growth rate is still rising, not fading. This is the rare case where the second derivative points up, which is exactly what the Exponential score rewards.
- Room to run: a $52B market cap against a trillion-dollar-plus AI-infrastructure/compute TAM leaves enormous headroom. Unlike a mega-cap, the law of large numbers is not the binding constraint here — execution and funding are.
- Reinvestment runway: capex was $4.07B in FY25 (8.6× depreciation) — the company is plowing every dollar and more into GPU capacity. The reinvestment story is intact and aggressive; the question is whether it converts to durable, profitable revenue.
Exponential Potential: Very High (9/10). This is the profile the flagship framework is built to find — forward, accelerating, tiny-relative-to-TAM. The honest counterweight lives entirely in the Risk score: exponentials that burn cash can also de-rate fastest. Own it for the asymmetry, not for safety.
5. Financials (real numbers — FMP annual/quarterly)
- Revenue: FY25 $529.8M, +351% (FY24 $117.5M; FY23 $20.9M). The ramp is real and steep.
- Quarterly trajectory (real acceleration): Q1'25 $55.3M → Q2 $105.1M → Q3 $146.1M → Q4 $227.7M → Q1'26 $399.0M (+624% YoY, +75% QoQ). The second derivative is positive through the latest print.
- Margins: gross margin ~48% TTM (a note of caution: quarterly gross margins are volatile as capacity is absorbed — Q1'26 gross margin was only ~21% as new GPU depreciation hit). Operating margin is negative (operating loss −$596M FY25). The headline TTM "EBITDA margin" and "net margin" in the ratios file are distorted by large one-off, non-operating items (asset disposals / discontinued Yandex-legacy gains) — do not read them as core profitability.
- Earnings: reported net income is noisy and one-off-driven (FY25 $101.7M bottom line, Q1'26 $621M net income) but core operations lose money; consensus EPS is −$2.38 FY26E, −$2.13 FY27E, turning slightly positive only around FY28E (+$0.25) and materially positive only by FY30E (~$8.44, wide dispersion, just 2 analysts). Treat the far-out EPS as low-confidence.
- Cash flow: operating CF +$385M FY25, but capex −$4.07B → FCF −$3.68B. This is a cash-consuming build-out; FCF yield is −4.8%.
- Balance sheet: cash $3.68B, total debt $4.97B, net debt just $1.30B, net-debt/EBITDA 0.14×, current ratio 8.3×. Nebius funded the build with a large FY25 debt raise (+$4.16B issuance) and sits on a strong liquidity position today — but at −$3.7B annual FCF, that runway is finite and future capex likely needs fresh capital (debt or equity).
6. Valuation — priced in or room?
There is no honest way to call NBIS cheap. It is loss-making on EPS, so P/E is meaningless; on sales it trades at ~59× EV/sales TTM and ~7.7× book. The entire bull case is that revenue grows into the multiple: on consensus, EV/sales falls to roughly ~15× on FY26E sales (~$3.5B) and ~5× on FY27E sales (~$11.6B) — i.e. the valuation compresses fast even at a flat price if the estimates hit. That is the same "grow into it" logic as a hyper-growth SaaS name, but with a heavier, capex-driven cost base and no profits yet.
A reverse read: at $215.62 and ~$52B EV, the market is already pricing several years of near-flawless triple-digit-then-double-digit revenue execution. Any wobble in AI-cloud demand, utilization, or financing terms de-rates it hard — which is why the bear case is −56%, not −15%. Street targets (context): consensus $196, high $255, low $129 — our $235 base is above consensus because we credit the FY27 ramp, but we flag that consensus itself embeds heroic growth. Not a value buy; a priced-for-perfection exponential buy.
7. Technicals (from the tech block)
- Trend: up but consolidating. $215.62 sits right at the 50-DMA ($215.09) and far above the 200-DMA ($132.94) — the longer-term uptrend is firmly intact (200-DMA well below price), but the stock has paused after a parabolic run.
- Location: −24.8% off the 52-week high ($286.69) and a staggering +387% off the 52-week low ($44.30). The −25% drawdown from peak is the current risk tell — this is a high-amplitude name.
- Momentum: RSI(14) 48.4 — neutral, neither overbought nor oversold. MACD +7.8 (still positive). No stretched-entry signal; the froth has partly worked off.
- Relative strength (the tell): NBIS +329% 12-mo vs SPY +20.6% and QQQ +30.3%; +111% 3-mo vs SPY +14% / QQQ +22%. Extraordinary outperformance — and a reminder that a name up this much can give a lot back.
- Read: the primary uptrend is intact but the stock is digesting a huge move at its 50-DMA. Beta 1.43 plus a −25% drawdown means scale in — a break and hold below the 200-DMA (~$133) would be a serious thesis-damage signal.
8. Moat & competitive position
Nebius's edge is specialization and independence: a Western, neutral "neocloud" purpose-built for AI workloads, run by an experienced ex-Yandex engineering team, with early access to large GPU allocations and a full-stack software layer (Toloka data, developer tooling) around the raw compute. In a market where AI labs want capacity fast and don't want to be captive to a hyperscaler, that positioning has real pull.
But the moat is contested and capital-intensive, not structural: the true competitors are the hyperscalers (AWS, Azure, Google Cloud), Nvidia-backed neoclouds (CoreWeave and peers), and every enterprise weighing build-vs-rent. Barriers are money and GPU access, both of which large rivals have in abundance. Switching costs are moderate. This is a land-grab, and Nebius is a fast, credible, but sub-scale participant.
Peer set (as supplied by FMP — note the mismatch): the file lists Charter ($19B), Fox ($25B), Pinterest ($15B), Reddit ($37B), Telefónica, Tencent Music, Twilio ($32B), Vodafone, Zillow. These are "Communication Services" classification artifacts, not true operating comparables — none is an AI-cloud provider. The economically relevant comps are neoclouds (e.g. CoreWeave) and hyperscaler AI-infra units, which are not in the supplied peer list; readers should mentally substitute those.
9. Management, capital allocation & guidance
- Capital allocation: all-in reinvestment — $4.07B capex in FY25 (8.6× depreciation) into GPU capacity, funded by a $4.16B debt raise, no dividend, no buyback. This is the right posture for a land-grab if utilization follows the spend; it is also what makes the equity fragile if it doesn't.
- Leadership: CEO Arkady Volozh, the founding Yandex entrepreneur — deep technical/operating pedigree, but also the executive whose company had to unwind its Russian roots, a reputational/geopolitical wrinkle worth noting.
- Insider activity: the only insider transactions in the sampled window are director John Wilson Boynton IV selling small Class-A lots (a few hundred shares each) across 2026-06-15 at $246–$262 — modest, scheduled-looking disposition near the highs, not a red-flag cluster of executive selling, but not the insider buying one would love to see at these prices.
- Guidance: management guides to continued rapid revenue scaling and a path to profitability as capacity fills; there are no vetted expert or management claims in the Synthos KB to weight here, so we take guidance at face value and flag it as the company's own book.
10. Catalysts & what to watch
- Next earnings: 2026-08-06 (Q2'26; Street EPS −$0.56, revenue ~$583M). The key lines: revenue growth rate (is +YoY still triple-digit?), gross margin trend (is capacity being absorbed profitably?), and any annualized-run-rate / ARR disclosure.
- Capex & funding: the size of the next capex step-up and how it's financed — more debt vs an equity raise (dilution) is the swing factor for the bear case.
- Utilization / contract wins: signed capacity commitments from named AI labs/enterprises would materially de-risk the demand side.
- Path to positive EBITDA: the first clean, non-one-off operating-profit inflection is the single most important fundamental tell.
- Macro / AI-capex cycle: any sign the broad AI-infrastructure build is slowing hits neoclouds first and hardest.
Thesis tripwires (what would change the call): revenue growth decelerating below ~50% YoY; gross margin failing to trend up as capacity fills; an unexpected large equity raise; or the stock breaking and holding below its 200-DMA (~$133).
11. Key risks
- Cash burn / funding wall (structural): −$3.7B FCF against a finite cash pile means Nebius depends on continued access to cheap capital. A tighter market forces dilution or slows the build — the core bear driver.
- Valuation / de-rating: 59× sales with no profits leaves zero margin for a demand or execution miss; the equity can halve on a single soft quarter.
- Competition from deeper pockets: hyperscalers and Nvidia-backed neoclouds can outspend Nebius on GPUs and data centers.
- Demand cyclicality: AI-infrastructure capex could prove lumpy or overbuilt; neoclouds are the most exposed link.
- Customer / geographic concentration: ~64% of FY25 revenue is US; early-stage revenue likely leans on a handful of large AI customers (concentration risk).
- Governance / legacy: ex-Yandex origins carry residual geopolitical and reputational considerations; dual-class Class-A structure and founder control.
- No expert corroboration: unlike conviction-track names, no vetted Synthos voice underwrites this thesis — the entire bull case rests on one macro bet.
12. Verdict, position sizing & monitoring
Buy — Tactical. Nebius is a genuine, high-conviction exponential: revenue is compounding at triple-digit rates off a tiny base (FY25 $530M, +351%; Q1'26 +624% YoY) into one of the largest TAMs in technology, with a strong current liquidity position (net debt just $1.3B, current ratio 8.3×). But it burns $3.7B of free cash a year, has no profits, trades at 59× sales, carries a 1.43 beta and just fell 25% from its high — so the same asymmetry that makes the Exponential score a 9 makes the Risk score an 8. That combination belongs in the high-risk tactical sleeve, never the core.
- Sizing: satellite, ~1–2.5% of the flagship — money you can watch swing 50%. Scale in (starter now, adds on pullbacks toward the 200-DMA) rather than a lump at the 50-DMA after a parabolic run.
- Monitoring: re-underwrite on the §10 tripwires; formal re-score each earnings print, with special weight on the gross-margin trend and the first clean EBITDA inflection. This verdict is logged as a tracked Synthos call as of 2026-07-03 at $215.62.
- Single biggest risk: the capex/funding treadmill — if AI-cloud demand or cheap financing falters, dilution or a funding wall follows.
Provenance & disclosures
- Traceability: 0 KB claims, breadth 0 — there is no expert coverage of NBIS in the Synthos knowledge base, so no
claim_ids are cited and none are fabricated. This note is explicitly fundamentals- and quant-driven. - Data as-of: fundamentals 2026-03-31 (Q1'26) · estimates & prices 2026-07-02/03 · no expert claims. Forward figures (revenue $3.5B–$53B, EPS −$2.38 → +$8.44, price targets) are analyst consensus (FMP), labeled as estimates; far-out FY30E EPS rests on just 2 analysts and is low-confidence.
- Reported-profit caveat: TTM net/EBITDA margins are distorted by large non-operating, one-off items (asset disposals / Yandex-legacy gains); core operations are loss-making. Do not read the headline margins as underlying profitability.
- Peer caveat: FMP's supplied peer set (Charter, Fox, Pinterest, Reddit, etc.) is a sector-classification artifact, not a set of true AI-cloud comparables.
- Not investment advice. Independent research, educational and informational only, never personalized. Hypothetical/forward figures are labeled; the only performance numbers Synthos will headline are the live, real-money Flagship's.
- Version: 2026-07-03. Prior versions available via the deep-dive version dropdown ("based on the info at the time").