ANALYSIS, NOT A POSITION. Quality Criteria PASS on the INFLECTION standard (74% gross margin from the Q1'26 6-K; the us-gaap:GrossProfit tag does not exist, so the automated check correctly returns INDETERMINATE rather than FAIL - this name is calibration item D1's named case). Valuation Criteria FAIL: today's $43.1bn enterprise value requires a 143.8% revenue CAGR for five years, and the growth-matched exit multiple is UNIDENTIFIED because the only comparators bracketing its (artifactual) 239.8% growth are four sub-$100m shells. The required path stays above 67%/yr at every exit multiple up to 30x. Demonstrated CAGR is NOT ESTIMABLE - revenue ran $4,794m (2021, the old Yandex N.V.) to $14m (2022) after the Russian divestment, so the margin (demonstrated - required) cannot be computed and is reported as such rather than fabricated. 12-month target $118-$227 on consensus FY2027E revenue at 2.87x-5.50x EV/revenue; NBIS trades at 4.11x, a 43% premium to CoreWeave. Downside Criteria (MEASURED, blocks nothing): -94.9% at p=0.27, cause named as GPU residual value against a useful life lengthened from four years to five in Q1'26.
How to read this
This is an analysis, not a position. The memo scores every Criteria and blocks on none of them. Whether an analysis justifies a position is a question about a particular book, and two books answer it differently — so this page carries no Long, Short, Watchlist or Avoid verdict. The previous verdict has been retired.
Every Criteria returns PASS / FAIL / INDETERMINATE, and carries a type. BINDING criteria are admission tests for a long-only absolute-return strategy. MEASURED criteria are always scored and stored, and never block — they inform timing, sizing or a future strategy. A missing input is INDETERMINATE, never FAIL.
Two valuation outputs, over two horizons. The implied-path test (reverse DCF) asks what today's price requires over five years and whether the business has demonstrated it; the 12-month target asks what the name is likely to trade at, on near-term consensus and the name's own multiple history. Neither replaces the other. Sensitivity is run over the exit multiple, never over scenario probabilities.
Momentum is entry timing only. It governs when to enter a position the thesis already justifies, never whether to own one.
Key findings
- Fundamentals are improving fast, on evidence rather than narrative: revenue +684% y/y to $399.0m in Q1'26; AI cloud ARR $1.25bn -> $1.92bn in one quarter; gross margin 51% -> 74%; adjusted EBITDA positive at +$129.5m group and a 45% margin in the AI cloud unit; RPO $21.3bn -> $33.6bn; $4.78bn of customer cash collected in advance; a first asset-backed facility priced at SOFR+250 covering more than 100% of the capex it funds. Every cost line fell sharply as a share of revenue.
- Against that: GAAP net income is uninformative (a $780.6m non-cash ClickHouse revaluation); the server useful life was extended from 4 to 5 years in the SAME quarter, flattering EBIT; customer concentration is material and undisclosed; and the FY2026 plan requires spending $20-25bn against $3.3bn of revenue.
- The market already expects it. House Base FY2026 revenue of $3,300m sits 2.9% BELOW the Street's $3,400m, and FY2027 $9,800m sits 6.7% BELOW ~$10,500m. Street rating is Buy across 18 analysts with targets RISING (Goldman $267 -> $286 on 2026-07-01; Northland to $410 on 2026-07-20). Delta-E is negative and small — no positive gap to monetise on the long side, and far too small and pointed the wrong way to build a short on.
- Fairly valued at the centre, extraordinary variance around it: Base DCF $174.88 (-5.4%); reverse DCF says today's $184.92 implies a 9.31x exit on Base 2031 EBITDA against the 9.0x the model uses — a 3% gap; probability-weighted value $190.56; scenario range $9 / $175 / $436. Valuation cannot originate a thesis here in either direction.
- Gate 2B was tested for the first time under v1.4.0 against all four evidence legs and FAILS on leg 4 — the price already embeds 93% of contracted power.
- Gate 4 fails: E[R] +5.7% gross / +3.1% dilution-adjusted / +2.8% net against a 4.7% cash hurdle, still 1.9pp short. P(bear) was cut 0.30 -> 0.27 on new transcript evidence of re-lease pricing, and the gate still fails. Simple framing: risking ~29% to make ~-5% to the Base target.
- Gate 5 passes with a hard constraint — 105% realised volatility against 161% implied — which is why the vehicle on conversion is outright equity, small, and NOT options.
- Conversion parameters: entry trigger $155, target $175 (Base DCF), invalidation $132, horizon 18 months anchored on the FY2026 exit-ARR print (Feb 2027) rather than the Q2 print. Size 0% today, 1.5% on conversion.
- Invalidation triggers are specific: a weekly close below the 200-day moving average ($140.29 and rising) kills the long-side watch entirely; any ATM equity raise at or below $180 signals the asset-backed financing template failed; Q2'26 ARR below $2.8bn or FY26 revenue guidance cut below $3.0bn means the Bear case is running.
- Screen provenance carried honestly: the 45-of-47 rank and the -2.87 composite are stated, and the two factors driving them (asset growth, gross profitability) are adjudicated with sourced evidence rather than repeated or waved away. The adjudication removes a false negative; it does not create a positive case.
- Model verification found and fixed two real bugs: the Scenarios-tab sensitivity grid referenced the SBC row instead of the Adj. EBITDA row when computing terminal value, and a column-A label beginning with '=' — very likely the reason Excel refused to open the first build.
Sections
Disclosed limitations
- Customer concentration (top-two share of the $33.6bn RPO) is NOT KNOWABLE. Corpora attempted: 20-F and 6-K segment notes (not disaggregated), transcripts (Meta 23 and Microsoft 9-12 mentions but no percentages), trade press. As a foreign private issuer Nebius is not required to give the US quarterly disclosure that would settle it.
- Interconnection-queue positions for PA / MO / FI were NOT RETRIEVED in this pass (PJM, MISO, Fingrid public queues and county permit records attempted). Disclosed as an incomplete leg rather than as 'not disclosed by the company'. This is the single highest-value follow-up for the next refresh.
- Realised revenue per MW and its trend are derivable only as an ESTIMATE: $8.9m per MW-year from mid-guidance ($8bn exit ARR / 900 MW), range $7.0-11.25m. The company has the number and does not publish it.
- Re-lease economics on first-generation GPUs are NOT KNOWABLE. Transcripts return 're-lease' 0/6 and 'churn' 0 since Q4-2024; no issuer in this sector publishes a cohort re-lease curve. The one qualitative datapoint is G-3's CRO assertion.
- The basis for the 4 -> 5 year server useful-life extension is disclosed as a change but unexplained as a judgement; 'useful life' appears 0/6 times in transcripts and no analyst has asked in six calls.
- Alpha Vantage EARNINGS_ESTIMATES returns an EMPTY ARRAY for NBIS (no coverage for this foreign private issuer), so estimate-revision counts are NOT COMPUTABLE. Consensus figures are web-aggregator-sourced, not Alpha Vantage-sourced. Flagged as a genuine gap rather than omitted.
- Six of seven earnings transcripts were retrieved; Q3-2024 could not be obtained from any accessible archive, and Alpha Vantage's EARNINGS_CALL_TRANSCRIPT quota was exhausted before the task began.
- The equity risk premium of 5.00% is an assumption — stated, not sourced.
- Excel would not open a workbook under automation on this machine, so the workbook's formula graph was evaluated independently with the `formulas` library and cross-checked against a separate Python implementation; both agree to the cent. The Python implementation additionally charges a financing-dilution adjustment that the Excel model of record excludes (Bull $416.72, probability-weighted $185.71, E[R] +0.4%); both figures fail the 4.7% hurdle, so the conclusion is unchanged.