AI cloud & computing infrastructure

Nebius Group

NBIS

The lower USD 81.31 entry price raises price attractiveness to 6.9/10 at unchanged modeled prices of USD 233.04 / 43.30. Company strength remains 6.6/10; any equity funding at the lower price would dilute shareholders more, and its actual extent remains unverified.

Analysis and ratings: ChatGPT·Business analysis 11/21/2026·Ratings & valuation 11/21/2026
THESIS STATUSActive growth thesis · return on capital remains unproven
CONFIDENCEMediumMedium confidence in business strength; low confidence in the three-year valuation because of financing, construction and dilution.
HORIZON3 years and longer
WHAT MATTERS MOSTConversion of capacity into revenue, contract economics, returns on capital and per-share value.

Business strength & price attractiveness

How AI assesses them

Business strength

Medium confidence
6.6/ 10

Previously 6.6 ·
Unchanged · Assessed

Business strength, independent of the stock price.

No quantified evidence since the post-Q3 review changes the five dimension scores. Price does not change business strength; funding and infrastructure replacement remain the chief weakness.

Five areas & supporting evidence

Price Attractiveness

Low confidence
6.9/ 10

Previously 6.3 ·
↑ +0.6 · Assessed

At 81.31 USD · close
Three-year horizon, through .

The lower USD 81.31 entry price raises price attractiveness to 6.9/10 at unchanged modeled prices of USD 233.04 / 43.30. Company strength remains 6.6/10; any equity funding at the lower price would dilute shareholders more, and its actual extent remains unverified.

Scenarios & supporting evidence

What changed in the latest update

New information: The lower USD 81.31 entry price raises price attractiveness to 6.9/10 at unchanged modeled prices of USD 233.04 / 43.30. Company strength remains 6.6/10; any equity funding at the lower price would dilute shareholders more, and its actual extent remains unverified.

Strength 6.6 → 6.6Attractiveness 6.3 → 6.9Scores changed

The numerical comparison does not by itself confirm or weaken the investment thesis.

Dated prices & model comparison
Input / modelPreviousCurrent
Reference price115.88 USD81.31 USD
Base modelled return+101.1 %+186.6 %
Adverse modelled return-62.6 %-46.7 %

Quote dates: 11/13/2026 → 11/20/2026. Model horizons end: 11/15/2029 → 11/21/2029. Cumulative returns include modelled dividends; they are scenarios, not guarantees.

Terminal-price and cumulative-dividend assumptions are unchanged; reference-price inputs differ as shown above.

View rating history →

Assessed · Methodology 1.2. Analytical judgments by ChatGPT; price changes do not automatically update these scores.

MARKET OVERVIEW

NBIS · Price chart

Delayed data · English controls.TradingView
Analysis contents · 13 sections
01

Company Snapshot

Analysis and scenario assumptions: ChatGPT.

ItemCurrent assessment
CompanyNebius Group N.V.
Ticker and share classNasdaq NBIS, Class A common shares, quoted in USD
SectorAI cloud, compute infrastructure and managed inference
Thesis statusActive growth thesis; long-term return on capital remains unproven
ConfidenceMedium for the business, low for three-year valuation
HorizonThree years for valuation; longer for platform development
Current viewPositive on product and demand; selective about entry valuation
Company strength6.6/10, assessed November 21, 2026, methodology 1.2
Price attractiveness6.9/10 at USD 81.31, regular close November 20, 2026 at 16:00 EST (22:00 CET)
Three-year scenariosBase USD 233.04 (+186.6%); adverse USD 43.30 (−46.7%)
Next quarterly resultsQ4 2026 date not yet researched; fewer than two months have passed since the last results

Investment thesis: Nebius can use the shortage of high-quality AI capacity to build a broader customer base and sell more software, inference and related services. Customer prepayments and long-term contracts help fund that expansion. Value creation depends on returns on infrastructure exceeding the real costs of GPU replacement and financing, with timely billing, healthy contract economics and rising value per share. The test is whether Nebius remains differentiated once scarce compute becomes more readily available.

Nasdaq NBIS Class A common shares closed the November 20 regular session at USD 81.31, 16:00 EST / 22:00 CET. This is 23.3% below the USD 106.04 input used in the previous assessment. Verified November 21. Price history

02

Investment Thesis

Commercial demand is the first pillar. On the Q2 earnings call Nebius announced four significant new contracts averaging more than USD 1 billion in total contract value. Named customers included Reflection and Cohere. Management attributed wins in competitive evaluations and technical tests to performance, reliability and support; signed contracts carry weight, while the explanation comes from the vendor. From the start of Q3 through the earnings call, Nebius cited more than USD 12 billion of new total contract value, including multiyear expansions. The letter named new or expanded relationships with Databricks, Datadog and Snowflake. September 30 accounting RPO was USD 38.7 billion versus USD 37.5 billion in June; these measures have different reporting windows and definitions. [S1, S9–S11]

Moving from infrastructure into services is the second pillar. Token Factory, Tavily and inference technology extend the offering around model training and deployment. Better utilization of the same hardware could raise returns on capital; measurable service revenue and gross profit would strengthen the thesis. Enterprise distribution is the third pillar: the Palantir partnership offers access to commercial customers and customer-controlled open-model deployments. Preferred status is neither exclusivity nor guaranteed revenue. Nebius may have more software and operating value than a replaceable GPU lessor; current scarcity and pricing are not permanent, and rapid revenue growth can coexist with weak per-share returns.

The November letter says Token Factory is approaching an annualized run rate in the hundreds of millions of dollars. Separate profit and cash flow are not disclosed. [S10]

03

The Business

Nebius integrates clusters, networking, storage, orchestration and model services for training, fine-tuning and running AI. Token Factory offers managed inference; Tavily provides search and external information retrieval. [S1, S2]

Contract typeManagement descriptionEconomics and risk
Medium-term AI cloudMainly 1–3 years; core marketHigher monetization and customer relationships; renewals matter
Short-term capacityTypically 3–6 months for intensive trainingPremium pricing, lower long-term visibility
Strategic long-term contractsFive or more years, mainly large creditworthy customersSupports secured funding; concentration and delivery obligations
Infrastructure partnershipsPartners provide capital and infrastructure; Nebius provides the platform and customer accessLower direct capital requirements; partner dependency and limited track record

Management’s earlier annual figures of USD 20–25 million per MW for medium-term deals and USD 40–50 million for premium short deals are neither quarterly sales nor rates for all contracted power. The Q3 weighted annual figure for new 1–5-year deals exceeded USD 25 million per MW and average GPU-hour pricing was over 20% above Q2. For comparable 1–3-year deals, reported payback improved from 22 to 19 months on management’s revenue-based measure, which accounts for operating and capital costs before prepayments. Actual returns after interest, GPU replacement and utilization beyond contract expiry are unproven. Live spot pricing was announced October 21. [S10, S13]

Avride, TripleTen and other stakes are non-core assets and receive separate model value without assuming rapid sales. AI cloud accounted for approximately 99% of Q3 group revenue. [S9, S10]

04

Competitive Advantages

Hardware/software integration and efficient, reliable cluster operation can reduce customer cost, but advantages need to show up in returns on invested capital. Designing and operating distributed systems is harder to copy than renting GPUs; technical wins are encouraging while the renewal record remains short. Multiple sites, contract lengths and partnerships offer flexibility, but building ahead of firm demand exposes capacity utilization to a turn in the cycle. Using infrastructure and inference together can increase switching costs; open models also improve portability and limit pricing power. Dependence on NVIDIA and well-capitalized competitors mean the technically credible moat remains unproven over a full investment cycle. [S1, S2]

05

Growth Drivers

Growth driverEvidenceWhat must follow
CapacityFour new revenue-generating sites in Q3; some colocations moved to H1 2027Cluster acceptance and billing, especially remaining Microsoft phases in Q2 2027
Larger customersOver USD 12 billion of contracts from the start of Q3 to the November callDelivery, collection and expansion at healthy returns
Inference and agentsToken Factory, Tavily, Eigen AI and Clarifai technologyUtilization and commercial service contribution
Enterprise distributionSeptember 2026 Palantir partnershipIntegration and paid production deployments
Partner capacityApproximately USD 325 million of one-off Q3 deployment revenueRecurring revenue and scale from 2027
Inference optimizationOctober 1, 2026 Inferize acquisitionFaster deployment and improved GPU utilization

Q3 set a sequential record for active capacity additions. Management expects to deliver more than 50% of Microsoft’s New Jersey capacity by year-end and the remainder in Q2 2027. A move from temporary to more permanent power changed DataOne’s timetable. Nebius says owned projects are on schedule or ahead and should account for most newly operational capacity from 2027. The year-end 2026 target remains 5 GW of contracted power, up from over 4 GW in November. Contracted power, accepted clusters and billable operation are distinct. September management comments pointed to demand extending into 2028 and strong use of older chips; neither guarantees permanent scarcity or current GPU economic life. [S2, S10, S11]

06

Economics & Financial Profile

MetricQ2 2026 / June 30Q3 2026 / September 30
Group revenueUSD 582.3mUSD 980.8m (+571% YoY, +68% QoQ)
AI cloud revenueApprox. USD 575mUSD 973.8m (+643% YoY)
Group adjusted EBITDAUSD 236.2m (40.6%)USD 465.8m (47.5%)
GAAP net loss from continuing operationsUSD 190.4mUSD 37.5m
Operating cash flow+USD 2,246.1m−USD 273.0m
Cash purchases of tangible and intangible assetsUSD 5,657.4mUSD 7,300.4m
CFO after these purchases, analyst calculation−USD 3,411.3m−USD 7,573.4m
Cash and equivalents excluding restricted cashUSD 8,042.1mUSD 7,425.6m
Book financial debt excluding operating leasesUSD 8,545.7mUSD 14,190.8m
Net financial debt, analyst calculationUSD 503.6mUSD 6,765.2m
Stock-based compensationUSD 102.5mUSD 58.7m
Period-end shares outstanding (A+B)Approx. 271.9m292.9m

Subtracting roughly USD 325 million of first-stage partner revenue from USD 980.8 million leaves about USD 656 million, roughly 13% above unadjusted Q2. The margin on that item was not disclosed, so the reported 47.5% does not reliably show normalized group margin. Adjusted EBITDA excludes depreciation, financing and other costs, and GPU renewal is a real economic cost. [S3, S9–S11]

Deferred revenue movements added USD 3.43 billion to Q3 operating cash flow, while receivables absorbed USD 3.76 billion. Nine-month cash outflow after those purchases was USD 11.20 billion. September deferred revenue was USD 9.41 billion and represents obligations to deliver services. Prepayment and collection timing can materially change quarterly cash flow. [S9]

November 12 guidance. Management moved 2026 revenue toward the low end of USD 3.0–3.4 billion and specified USD 3.0–3.2 billion on the call. Adjusted EBITDA margin remains around 40%; capex rises from USD 20–25 billion to about USD 27 billion. The former December ARR goal of USD 7–9 billion was replaced by a February 2027 run-rate above USD 7 billion. Initial 2027 guidance is USD 12–13 billion revenue, USD 40–45 billion capex, over 1 GW of new capacity and Q4 pretax break-even using five-year accounting lives for compute equipment. Pretax profitability after financing costs and cash flow after investment are distinct tests. [S10, S11]

Share-price dilution sensitivity, analyst calculation as of November 21. Raising USD 1bn at USD 81.31 would require approximately 12.3m shares versus 9.4m at USD 106.04, roughly 30% more shares for the same proceeds. September's unused ATM authorization of 7.64m shares would raise approximately USD 621m gross at USD 81.31. These are hypothetical calculations before issuance costs, based on September's authorization. The actual remaining authorization and post-Q3 issuance have not been verified. [S9, S18]

NVIDIA, November 17. NVIDIA reported aggregate maximum contingent neocloud commitments of USD 32bn as of October 25. Under the disclosed terms, operators must first secure funding for infrastructure purchases and make capacity operational for NVIDIA acceptance. Commitments may decline or expire. This is a sector aggregate without a Nebius-specific amount. It therefore does not alter September NBIS cash/funding or the modeled debt and shares. NVIDIA prepaid warrants arise from a separate March investment. [S9, S15, S16]

07

Key KPIs

KPIQ3 positionWhat to watch
Revenue run rateUSD 6.4bn at September end vs USD 3.0bn in JuneOver USD 7bn in February guidance; then focus on revenue, pretax profit and capacity guidance
Active and billable capacityFour new revenue sites, some colocation delaysAcceptance and billing, Microsoft New Jersey Q2 2027
New-contract economicsOver USD 25m per MW annually for weighted 1–5-year mix; claimed 19-month payback for comparable 1–3-year contractsRenewals, residual value and returns after interest
Prepayments relative to capexOver 100% of associated capex for disclosed new Q3 deals excluding sub-one-year contracts versus 50–60% in Q2Actual collections and group-wide capex coverage
RPOUSD 38.7bn; approximately 40% expected to be recognized within 24 months from September 30, 2026Concentration and timing; RPO excludes contracts originally one year or shorter
CFO after capex−USD 7.57bn in Q3Sustainability after prepayments, growth and replacement investment
Share count and converts292.9m outstanding and approximately 21.1m prefunded NVIDIA warrantsATM, compensation, conversions or refinancing
Partnership and inferenceApprox. USD 325m of one-off deployment revenue; company says Token Factory is approaching a run rate in the hundreds of millions of dollarsRecurring sales, margins and cash separately from one-offs

Sources: [S9–S11]. RPO, contract value, cash prepayments, annualized monthly revenue and recognized sales measure different things.

08

Management & Capital Allocation

Arkady Volozh leads a technically ambitious business willing to invest ahead of demand. Growth and contracts support execution, while weaker pricing or funding would test discipline.

Citi bridge. The secured USD 3.5 billion facility announced October 22 matures October 16, 2027. The margin over Term SOFR starts at 2.75 percentage points and rises by 0.25 points every 90 days to 3.50 points. It requires minimum liquidity and extensive asset collateral. It expands short-term funding options; actual drawings are unverified. [S12]

August notes had initial conversion prices of approximately USD 313.46 and 324.65. The first series’ low coupon does not eliminate the cost because principal accretes. Weak share performance leaves debt outstanding; success may dilute shareholders. Q3 filings show the USD 5.75 billion issuance and exchange of USD 800 million of older principal for 15.77 million shares. Undiscounted contractual cash payments on all convertibles at September 30 were about USD 17.35 billion including accretion and coupons without prepayment or conversion, distinct from book debt. [S4, S9]

ATM and warrants. In Q3 Nebius sold 4.63 million treasury shares at USD 229.14 on average for USD 1.05 billion net. About 7.64 million ATM shares remained at September 30. The 292.9 million outstanding shares and roughly 21.1 million prefunded NVIDIA warrants give approximately 314.0 million economic shares before additional compensation or conversions. [S9]

2027 funding. Management expects USD 17–20 billion of prepayments and 75–85% capex coverage with asset- or creditworthy-contract-backed financing included. At USD 42.5 billion midpoint capex, about USD 6.4–10.6 billion still needs balance-sheet liquidity, operating collections or additional debt/equity, assuming guidance is met and funding and payments align. Secured loans are still debt. The July USD 775 million facility at SOFR plus 2.50 percentage points is a precedent. [S9–S11]

09

Risks / Bear Case

RiskShareholder impactWarning sign
Costly or unavailable fundingHigher interest, delayed growth or equity issued at a low priceFunding gap for 2027 or worse Citi refinancing
Excess supplyWeaker renewals and GPU revenueDiscounts alongside lower utilization
Chip obsolescenceCostly replacement and lower residual valueReplacement sooner than the five-year accounting life
Construction delaysDebt and costs before revenueFurther delay to remaining New Jersey phases beyond Q2 2027
Customer concentrationLarge-customer loss or schedule changeRestructuring or lower prepayments
Weak customer economicsHigher credit riskPayment problems despite rising use
NVIDIA and partner dependenceLess bargaining power and fewer delivery optionsWorse chip access or terms
Weak software contributionInfrastructure rather than platform valuationService growth fails to lift returns on capital

The adverse case only requires capacity supply to catch up with demand when investment and debt are already large. Sales can grow while value per original share falls.

10

Valuation

ChatGPT produces the calculations and valuation scenarios. Share-price inputs are dated to the analysis; they are not continuously updated market data.

CHATGPT SCENARIOS · November 21, 2026 → November 21, 2029

Different outcomes. Different values.

Historical analysis input
81.31 USD · November 20, 2026 · 16:00 EST · Nasdaq close
Bear
43.3 USD
Base
233 USD
Conditional scenarios from the analysis. The dashed line marks the historical share price used in the calculation. Values are neither current intrinsic value nor a share-price forecast.

Method and horizon. ChatGPT models AI operations with EV/adjusted EBITDA and an explicit equity bridge. GPU replacement, capex and financing risk constrain the multiple; growth alone does not warrant a software multiple. The horizon is November 21, 2026 to November 21, 2029; full-year 2029 assumptions approximate the exit rate. These are analyst scenarios, not guidance.

ItemBaseAdverse
Operating developmentDelayed colocations come online, owned sites ramp and pricing normalizesSlower 2027 ramp, further delays, lower prices, costlier funds and lower utilization
Annual 2029 AI revenueUSD 22.0bnUSD 14.0bn
Adjusted EBITDA margin43%32%
Adjusted EBITDAUSD 9.46bnUSD 4.48bn
Terminal EV/EBITDA11×9×
Enterprise valueUSD 104.06bnUSD 40.32bn
Terminal net financial debtUSD 22.0bnUSD 25.0bn
Net value of non-core assetsUSD 3.0bnUSD 2.0bn
Equity valueUSD 85.06bnUSD 17.32bn
Modeled shares365m400m
Three-year priceUSD 233.04USD 43.30
Three-year dividendsUSD 0USD 0
Total return from USD 81.31+186.6%−46.7%
Compound annual return+42.0%−18.9%

Base USD 22 billion in 2029 implies roughly 32.7% annual growth over two years from the midpoint of 2027 group guidance, with most sales in AI cloud. The adverse case assumes a 2027 revenue miss and slower renewals. Both multiples are 1× below October because of higher investment needs, ramp risk and GPU replacement. The base margin reflects scale and normalized pricing; it does not extrapolate Q3’s reported margin.

September net book debt was USD 6.77 billion. The model assumes roughly USD 15 billion of net additions by 2029 in the base case after cash and conversions, versus USD 18 billion adverse. These are scenario balance sheets for a larger asset base funded through senior secured borrowing, prepayments and capital. Adverse revenue is lower but weaker internal funding and refinancing needs increase projected debt and equity issuance. Public information does not support a complete funding schedule to 2029.

The share-count starting point is approximately 314.0 million economic shares including NVIDIA warrants. Another roughly 51 million base or 86 million adverse allow for new equity, compensation and possible conversion of older series, with greater issuance at lower prices in the adverse case. August notes with initial conversion prices above USD 313 remain debt in both cases. Other instruments are counted either as shares or as remaining debt under the scenario.

Price equals (revenue × EBITDA margin × multiple − net debt + non-core assets) / shares. Operating leases are treated as expense and not added again as financial debt. Non-core assets are valued outside AI operations, so their earnings are not counted in terminal EBITDA. Net debt is after assumed conversions; debt and dilution are not double-counted. Customer prepayments reduce external funding depending on timing and are reflected in net debt. Return equals (terminal price + dividends) / 81.31 − 1, in USD before taxes, fees and USD/CZK moves.

Change from October. Base value fell from USD 246.29 to 233.04: higher debt, higher share count and a lower multiple outweighed higher revenue. Adverse value fell from USD 63.16 to 43.30 as delays and funding costs hit equity harder.

Change from baseTerminal price impact
Multiple 1× EBITDA higher or lowerApprox. ±USD 25.92 per share
Net debt USD 5bn higherApprox. −USD 13.70 per share
400m rather than 365m sharesUSD 212.65 versus USD 233.04
500m rather than 400m shares in the adverse case, at unchanged equity valueUSD 34.64, a −57.4% return from USD 81.31

Discounting base USD 233.04 at an 18–22% annual requirement indicates USD 128–142 before weighting the adverse case. Review levels of USD 135 and USD 105 imply approximately 20% and 30% annual base returns. The November 20 close of USD 81.31 is below both review levels and triggered this reassessment. The adverse case still implies roughly 59% loss at USD 105.

Maintaining 365m/400m shares in the November 21 revision is a model assumption based on multiple funding sources. The lower quote does not establish that a new issuance occurred. Longer-term equity funding at lower prices could require more shares, as reflected in the added sensitivity.

Share-price dilution sensitivity, analyst calculation as of November 21. Raising USD 1bn at USD 81.31 would require approximately 12.3m shares versus 9.4m at USD 106.04, roughly 30% more shares for the same proceeds. September's unused ATM authorization of 7.64m shares would raise approximately USD 621m gross at USD 81.31. These are hypothetical calculations before issuance costs, based on September's authorization. The actual remaining authorization and post-Q3 issuance have not been verified. [S9, S18]

11

What Would Change the Assessment?

Positive evidence: repeated on-time capacity starts, renewal pricing holding, measurable high-return services and fewer equity issues than modeled. Paid Palantir deployments would help. Near-term checks include a February run rate above USD 7 billion, remaining Microsoft delivery in Q2 2027 and confirmed draws to finance the 2027 investment plan.

Negative evidence: further delays while capital is spent, rapid older-GPU price declines, weaker collections, financing terms harmful to existing shareholders or growth without per-share gains. Watch for another New Jersey delay, prepayments materially below guidance or Citi refinancing that materially increases costs or equity needs.

Questions for future reviews

  • Does connected capacity become billable in a reasonable time?
  • Are new deals profitable after replacement capex and interest?
  • Are services increasingly bought for quality rather than GPU availability alone?
  • Can signed deliveries be funded in a weaker equity market?
  • Does value per share keep pace with the company?
  • Does older-chip demand hold without large discounts?
  • Does Palantir bring paying customers?
  • Can recurring cash generation be separated from new prepayments?

These are open questions, not a list of conditions already met.

12

Current View

The growth thesis remains active. Q3 showed capacity monetization, higher prices on new contracts and more customer funding. Per-share value now hinges on turning expanded investments into recurring revenue and cash after capex. At USD 81.31 risk/reward improved, but ramp timing, refinancing and dilution remain decisive. The base annual return is 42.0%; the adverse three-year loss is 46.7%.

Company strength

6.6/10 as of November 21, 2026, methodology 1.2. Each area has equal 20% weight.

DimensionScoreReason
Competitive advantage7.0/10Infrastructure, software and inference integration; supplier dependency and limited renewal history
Demand durability8.8/10Higher prices and longer contracts, backed by USD 38.7bn RPO
Management execution7.1/10Four new revenue sites, but colocation delays moved deliveries and affected guidance
Business economics5.8/10Better new-deal unit economics per management and higher EBITDA; renewal investment and normalized cash conversion unproven
Financial resilience4.3/10Prepayments and secured financing help; net debt, capex, refinancing and dilution risk rose

(7.0 + 8.8 + 7.1 + 5.8 + 4.3) / 5 = 6.6. Stronger demand and contract economics are offset by delivery delays and capital needs. The chief weakness is having to finance and renew expensive infrastructure in weaker markets. Confidence is medium: operating growth is documented, but full-cycle history is short and service economics lack transparency. Demonstrated higher returns on capital, repeatable funding without excessive dilution and partner-model revenue would lift the score; repeated delivery misses, weaker GPU renewals or funding would reduce it.

Price attractiveness

Price attractiveness: 6.9/10 at USD 81.31, the November 20, 2026 regular-session close, assessed November 21 under methodology 1.2, with a three-year horizon. Lower entry price improves model returns to +186.6% base and −46.7% adverse at unchanged terminal prices of USD 233.04 / 43.30 and zero dividends. The chief weakness remains terminal debt, dilution and GPU returns. Confidence is low: future funding and construction through 2029 remain uncertain. Timely billing, advance collections and longer-term financing on acceptable per-share terms would support a higher score; further delays, more shares or weaker renewal economics would require reassessment. Sources: Q3 filings, letter, call, Citi terms, NVIDIA commentary and the November 20 close [S9–S12, S15, S16, S18].

November 21 rationale: No newly quantified change in paid capacity, contract economics, investment guidance or financing was found in the reviewed November 18–21 sources. The Physical AI Awards announcement supports ecosystem development without disclosing new contracted revenue. All five business scores remain unchanged. The decline from USD 106.04 to USD 81.31 increases base potential from +119.8% to +186.6%; the adverse case still loses 46.7%. Attractiveness rises less than mechanically implied upside because funding for the full investment plan is not documented and lower-price issuance may require more shares. EBITDA valuation remains sensitive to GPU replacement. Timely billing of delayed capacity, advance collections and longer-term funding with acceptable per-share effects would support a higher score; further delays, more shares or weaker renewal economics would require reassessment.

13

Thesis History / Updates

DateEventEffect
August 12, 2026Q2 results, large deals and reiterated guidanceDemand support alongside exceptional investment needs
August 24, 2026USD 5.75bn offering closed and older debt exchangedGrowth funding, additional liabilities and dilution
September 8, 2026Palantir partnership and conferenceStronger enterprise access; economics still to be demonstrated
October 1, 2026Inferize acquisitionInference capabilities; price undisclosed
October 18, 2026First full assessmentStrength 6.6/10, attractiveness 5.4/10 at USD 165.47, methodology 1.2
October 22, 2026USD 3.5bn Citi facilityShort-term liquidity; matures October 2027
November 12, 2026Q3 and initial 2027 guidanceBetter deal pricing and prepayments, some delays and higher capex
November 15, 2026Post-Q3 review, methodology 1.2Strength unchanged at 6.6/10; attractiveness 6.3/10 from 5.4 at USD 115.88 November 13 close; base USD 233.04 from 246.29 and adverse USD 43.30 from 63.16

The historical events are not retroactively assigned scores. The first dated assessment is October 18, 2026.

Next review follows Q4 or a material change in capacity, financing, contracts or price.

Sources

S1 — NBIS Q2 2026 earnings call, August 12, 2026, for the period through June 30; comments on contracts, capacity and outlook. Where transcript accounting figures conflict, official filings take precedence.

S2 — Goldman Sachs Communacopia and Technology Conference, September 8, 2026. Volozh and Boroditsky on platform strategy, contracts and demand; management comments are not independent market verification.

S3 — Official Q2 release, August 12, 2026 and financial report, quarter through June 30.

S4 — Convertible issuance closing, August 24 and August 19 pricing; final and originally priced amounts differ.

S5 — Palantir and Nebius, September 8; no guaranteed revenue.

S6 — Inferize acquisition, October 1; consideration undisclosed.

S7 — NBIS price history, USD 165.47 October 16 close, verified October 18.

S8 — Investment Analysis Rules, methodology 1.2 effective October 16, 2026.

S9 — Q3, November 12: Nebius results, SEC earnings exhibit, SEC management commentary, full financial statements.

S10 — Q3 shareholder letter, November 12: guidance, commercial metrics and capacity.

S11 — Q3 earnings call, November 12: webcast and transcript.

S12 — Citi bridge announced October 22: SEC 6-K terms and Nebius announcement.

S13 — Live GPU spot pricing announcement, October 21.

S14 — November 13 regular close verified November 15: Stock Analysis and Morningstar.

Actual Citi drawings, first partner revenue margins and project-by-project funding through 2029 remain unverified. Modeled exit debt, share count and non-core value have low confidence.

November 21, 2026 — methodology 1.2: The lower USD 81.31 entry price raises price attractiveness to 6.9/10 at unchanged modeled prices of USD 233.04 / 43.30. Company strength remains 6.6/10; any equity funding at the lower price would dilute shareholders more, and its actual extent remains unverified.

S15 — NVIDIA Q3 FY2027, November 17, 2026. Confirmation of operational Rubin systems at Nebius.

S16 — NVIDIA CFO commentary, November 17, 2026. Conditional neocloud commitments without a Nebius-specific amount.

S18 — NBIS regular close, November 20, 2026. USD 81.31; checked November 21.

S19 — Physical AI Awards announcement, November 18, 2026. Ecosystem development without disclosed new contracted revenue.

S19 — Physical AI Awards · Nebius IR.

Was this analysis useful to you?

Have a suggestion about this analysis?

Suggestions are not displayed publicly. They help inform future updates. No name or email is required.

Sources and scenarios reflect the date stated in the analysis.Back to all companies