NBIS · Forward model · Bull case
The Bull case, 20 quarters out
Each segment is projected from its own operating driver, rolled up into consolidated cash flow, and discounted back to a fair value per share. The assumptions below are editable — change them and every number on this page moves with them.
Nebius reports three operating segments plus eliminations, and quarterly segment revenue is disclosed or cleanly derivable back to 2024 Q3, so the split here is reported rather than invented. The two verticals sum to filed consolidated revenue in all eight historical quarters: $32.1M, $35.2M, $50.9M, $105.1M, $146.1M, $227.7M, $399.0M and $582.3M. Toloka was deconsolidated in Q2 2025 and every prior period was restated to continuing operations, so these figures are on the restated base and no growth rate quoted before that reclassification is comparable. What is NOT disclosed is any quarterly capacity series: connected and active megawatts exist only as year-end actuals - 25 MW connected at the end of 2024, about 170 MW active at the end of 2025 - and as guidance of 800 MW to 1 GW connected at the end of 2026. There is no megawatt figure at all for 2026 Q1 or Q2 and no utilisation figure anywhere, so a capacity driver would have had to invent its own installed base; the megawatt ladder is used as scenario evidence only. What is ours: the growth path and its decay, the margin and capex glides, the tax rate, the discount rate and the exit multiple. Corporate overhead is set to zero deliberately, because from Q2 2026 corporate-function expense is already inside the cloud segment's costs and charging it again would double count. Share count is the economic-basic 308.7M - 271.9M outstanding at 2026-06-30 plus the 21.1M NVIDIA pre-funded warrants the company itself counts as outstanding plus the ~15.8M Class A issued in the August 24 note exchange - and is then held flat, so further dilution from the 65.4M anti-dilutive convertible shares, the new 2030 and 2034 notes and the remaining ATM capacity is charged nowhere. Net debt is the basis-date $503.6M; the July secured facility and the August $5.75B of convertibles are not rolled in, because the July and August capex and interest that would offset them are undisclosed. FY2026 revenue guidance, the ~40% adjusted EBITDA margin and the $20-25B capex range exist only in the earnings call - no SEC filing contains them - and adjusted EBITDA adds back the depreciation that all that capex creates, which is why this model charges capex in cash rather than valuing a multiple of EBITDA.
The high end of every guided range plus the pricing evidence management put on the record: an auction that cleared 15% above the best previous Blackwell price, more than 30% higher pricing on older-generation GPUs than in Q1, and short-term capacity priced at $40-50M per megawatt against the $20-25M on the landmark deals. Payback of one year and ten months on Q2 deals, down from a two-to-three-year range, is the single number that changes the economics of the fleet, because it is what turns a capex machine into a compounding one inside this horizon.
Latest: $26.93B (2031Q2E)
| Period | Value |
|---|---|
| 2024Q3 | $32M |
| 2024Q4 | $35M |
| 2025Q1 | $51M |
| 2025Q2 | $105M |
| 2025Q3 | $146M |
| 2025Q4 | $228M |
| 2026Q1 | $399M |
| 2026Q2 | $582M |
| 2026Q3E | $935M |
| 2026Q4E | $1.42B |
| 2027Q1E | $2.06B |
| 2027Q2E | $2.84B |
| 2027Q3E | $3.79B |
| 2027Q4E | $4.87B |
| 2028Q1E | $6.08B |
| 2028Q2E | $7.39B |
| 2028Q3E | $8.80B |
| 2028Q4E | $10.28B |
| 2029Q1E | $11.81B |
| 2029Q2E | $13.38B |
| 2029Q3E | $14.99B |
| 2029Q4E | $16.62B |
| 2030Q1E | $18.28B |
| 2030Q2E | $19.96B |
| 2030Q3E | $21.66B |
| 2030Q4E | $23.39B |
| 2031Q1E | $25.14B |
| 2031Q2E | $26.93B |
What drives each segment
Nebius AI cloud
Growth pathA GPU cloud whose revenue is the delivery of connected, GPU-equipped power under increasingly prepaid multi-year contracts. Every quarter of growth so far has come from capacity coming online at rising prices, and the forward book is disclosed: $37.5B of remaining performance obligations with a published runoff, $3.0B of ARR at the end of June, and guidance to exit 2026 between $7B and $9B of ARR. Capacity is the stated constraint - the company says it could sell its whole 2027 book today and is choosing not to - but no quarterly megawatt or utilisation series exists, so this line is a growth path anchored on the ARR ladder and the RPO schedule rather than a capacity build.
Latest: $26.92B (2031Q2E)
| Period | Value |
|---|---|
| 2024Q3 | $27M |
| 2024Q4 | $24M |
| 2025Q1 | $41M |
| 2025Q2 | $94M |
| 2025Q3 | $131M |
| 2025Q4 | $214M |
| 2026Q1 | $390M |
| 2026Q2 | $575M |
| 2026Q3E | $928M |
| 2026Q4E | $1.41B |
| 2027Q1E | $2.05B |
| 2027Q2E | $2.84B |
| 2027Q3E | $3.78B |
| 2027Q4E | $4.86B |
| 2028Q1E | $6.07B |
| 2028Q2E | $7.38B |
| 2028Q3E | $8.79B |
| 2028Q4E | $10.27B |
| 2029Q1E | $11.80B |
| 2029Q2E | $13.37B |
| 2029Q3E | $14.98B |
| 2029Q4E | $16.61B |
| 2030Q1E | $18.27B |
| 2030Q2E | $19.95B |
| 2030Q3E | $21.65B |
| 2030Q4E | $23.38B |
| 2031Q1E | $25.13B |
| 2031Q2E | $26.92B |
Assumptions & reasoning
- Segment revenue includes intersegment transactions; the company states it is impracticable to separate external from intersegment revenue for any segment, so this line is the segment as filed, not an external-customer line.
- From Q2 2026 corporate-function expense is no longer shown separately and is folded into this segment's other costs, so the 49.7% adjusted EBITDA margin is not a clean facility-level contribution margin and corporate overhead is deliberately set to zero above it.
- Segment assets are not reviewed by the chief operating decision maker, so no segment capex exists. The 900% opening capex intensity is group capex applied to this line, which is defensible only because Avride and TripleTen are immaterial to the $5.66B spent in the quarter.
- Adjusted EBITDA adds back the depreciation created by the guided $20-25B of capex, and the server life was raised from four years to five from 2026, which lowers D&A as a share of revenue independently of anything operational. This margin is a pre-depreciation margin.
- The $12M, $20-25M and $40-50M per-megawatt figures management quotes are new-deal prices, not a fleet average, and no megawatt count was disclosed for 2026 Q1 or Q2. That is why this vertical is a growth driver and not a capacity driver.
- The 2024 Q4 figure is FY2024 segment revenue of $68.3M less the nine months to 2024-09-30 of $44.5M, so it is marked estimated; the other seven quarters are printed segment figures.
Avride, TripleTen and eliminations
Growth pathGroup revenue less the Nebius AI cloud segment: Avride's autonomous driving business, TripleTen's reskilling bootcamps, and the elimination of the other segments' own use of the cloud platform. Together they are 1.3% of group revenue and cost about $49.5M of adjusted EBITDA a quarter, which is the entire gap between the $285.7M the cloud segment earned and the $236.2M the group reported. They are carried here so the two verticals sum exactly to reported consolidated revenue.
Latest: $14M (2031Q2E)
| Period | Value |
|---|---|
| 2024Q3 | $5M |
| 2024Q4 | $11M |
| 2025Q1 | $10M |
| 2025Q2 | $11M |
| 2025Q3 | $15M |
| 2025Q4 | $14M |
| 2026Q1 | $9M |
| 2026Q2 | $7M |
| 2026Q3E | $7M |
| 2026Q4E | $7M |
| 2027Q1E | $7M |
| 2027Q2E | $8M |
| 2027Q3E | $8M |
| 2027Q4E | $8M |
| 2028Q1E | $8M |
| 2028Q2E | $9M |
| 2028Q3E | $9M |
| 2028Q4E | $9M |
| 2029Q1E | $10M |
| 2029Q2E | $10M |
| 2029Q3E | $10M |
| 2029Q4E | $11M |
| 2030Q1E | $11M |
| 2030Q2E | $12M |
| 2030Q3E | $12M |
| 2030Q4E | $13M |
| 2031Q1E | $13M |
| 2031Q2E | $14M |
Assumptions & reasoning
- Every quarter here is derived - reported group revenue less the reported Nebius AI cloud segment - so all eight points are marked estimated even though both inputs are printed figures. The subtraction is exact; it is the split that is not separately disclosed.
- Avride volumes are published - more than 200 AV-capable vehicles and over one million autonomous miles in 2026 to date - but there is no revenue per mile, ride or vehicle, so nothing here can be driven by units.
- TripleTen's real driver is student enrolment times average check, and neither series is published quarterly; the disclosed fact is that U.S. student volumes are falling as it moves off legacy offerings.
- The -668.9% margin is the combined Avride and TripleTen adjusted EBITDA of -$49.5M over $7.4M of residual revenue. It glides toward -200% on an assumption that Avride's losses narrow, which no published unit economics support.
- This line can fall while Avride and TripleTen both grow, because eliminations scale with how much cloud the other segments consume.
Where each case comes from
Bear case — primary sources
The primary sources this case is built from — filings, calls and posts. Where they are claims by an interested party rather than disclosures, every number in the Bear column is what happens if they are taken at face value.
Q2 2026 shareholder letter, deal timing and pricing
Q2 2026 interim financial statements, the cost of waiting
Bull case — primary sources
The primary sources this case is built from — filings, calls and posts. Where they are claims by an interested party rather than disclosures, every number in the Bull column is what happens if they are taken at face value.
Q2 2026 shareholder letter, pricing and payback
Q2 2026 earnings call, capacity pace
Volozh case — primary sources
The primary sources this case is built from — filings, calls and posts. Where they are claims by an interested party rather than disclosures, every number in the Volozh column is what happens if they are taken at face value.
Q2 2026 earnings call, the inventory decision
Q2 2026 shareholder letter, contracted power raised again
From cash flow to fair value
| Present value of free cash flow, 20 quarters | -$65.78B |
| Terminal-year revenue | $97.13B |
| Terminal-year EBITDA | $48.87B |
| Exit multiple, on revenue | 4.5x |
| Terminal value | $437.07B |
| Discounted at 15.0% a year, terminal value becomes | $217.30B |
| Enterprise value | $151.52B |
| Net cash | -$504M |
| Equity value | $151.01B |
| Shares | 0.31B |
| Fair value per share | $489.16 |
| Against the current price of $218.48 | +124% |
15% matches the rate the r40 CoreWeave model uses for a levered, negative-free-cash-flow neocloud; NBIS carries less net debt at the basis date but a heavier dilution stack - 65.4M anti-dilutive convertible shares at Q2, $5.75B of new 2030 and 2034 notes struck at $313.46 and $324.65, and 12.3M of ATM capacity left. The 3.5x exit sits above CoreWeave's 3.0x because the cloud segment earns a 49.7% adjusted EBITDA margin against CoreWeave's 59% on far more debt, and because at 2031 Q2 this model still spends 60% of revenue on capex and free cash flow is still negative. The exit multiple is therefore pricing the steady state after the build, not the terminal quarter. The shares trade near 19.7x FY2026 guided revenue and 5.6x FY2027 consensus.
Read the other way round: at $218.48 the market is paying 2.8x terminal-year revenue, holding every other assumption on this page fixed. That is the number to argue about.
The projected path
| Quarter | Nebius AI cloud | Avride, TripleTen and eliminations | Revenue | YoY | EBITDA | Capex | FCF | R40 | PV of FCF |
|---|---|---|---|---|---|---|---|---|---|
| 2026 Q3E | $928M | $7M | $935M | — | $450M | $6.17B | -$5.72B | — | -$5.52B |
| 2026 Q4E | $1.41B | $7M | $1.42B | — | $709M | $7.01B | -$6.30B | — | -$5.88B |
| 2027 Q1E | $2.05B | $7M | $2.06B | +415% | $1.04B | $7.65B | -$6.61B | +94 | -$5.95B |
| 2027 Q2E | $2.84B | $8M | $2.84B | +389% | $1.46B | $8.11B | -$6.65B | +155 | -$5.78B |
| 2027 Q3E | $3.78B | $8M | $3.79B | +305% | $1.94B | $8.41B | -$6.46B | +134 | -$5.43B |
| 2027 Q4E | $4.86B | $8M | $4.87B | +242% | $2.50B | $8.60B | -$6.10B | +117 | -$4.95B |
| 2028 Q1E | $6.07B | $8M | $6.08B | +195% | $3.12B | $8.75B | -$5.63B | +103 | -$4.41B |
| 2028 Q2E | $7.38B | $9M | $7.39B | +160% | $3.80B | $8.91B | -$5.12B | +91 | -$3.87B |
| 2028 Q3E | $8.79B | $9M | $8.80B | +132% | $4.51B | $9.11B | -$4.60B | +80 | -$3.36B |
| 2028 Q4E | $10.27B | $9M | $10.28B | +111% | $5.26B | $9.39B | -$4.13B | +71 | -$2.91B |
| 2029 Q1E | $11.80B | $10M | $11.81B | +94% | $6.03B | $9.75B | -$3.72B | +63 | -$2.53B |
| 2029 Q2E | $13.37B | $10M | $13.38B | +81% | $6.82B | $10.20B | -$3.39B | +56 | -$2.23B |
| 2029 Q3E | $14.98B | $10M | $14.99B | +70% | $7.62B | $10.75B | -$3.13B | +50 | -$1.98B |
| 2029 Q4E | $16.61B | $11M | $16.62B | +62% | $8.44B | $11.37B | -$2.94B | +44 | -$1.80B |
| 2030 Q1E | $18.27B | $11M | $18.28B | +55% | $9.26B | $12.08B | -$2.82B | +39 | -$1.67B |
| 2030 Q2E | $19.95B | $12M | $19.96B | +49% | $10.09B | $12.84B | -$2.75B | +35 | -$1.58B |
| 2030 Q3E | $21.65B | $12M | $21.66B | +45% | $10.93B | $13.67B | -$2.74B | +32 | -$1.52B |
| 2030 Q4E | $23.38B | $13M | $23.39B | +41% | $11.78B | $14.56B | -$2.78B | +29 | -$1.48B |
| 2031 Q1E | $25.13B | $13M | $25.14B | +38% | $12.64B | $15.49B | -$2.85B | +26 | -$1.47B |
| 2031 Q2E | $26.92B | $14M | $26.93B | +35% | $13.52B | $16.47B | -$2.95B | +24 | -$1.47B |
Every row is projected. A year-over-year change is shown only where the quarter it compares against exists — an em dash means there is no comparable quarter, not a flat year.
Model revisions
Assumptions are marked to reality as each quarter prints. Every change is appended here, with the fair value the model produced at the time, so the model's own history stays visible.
| Date | Changed | Fair value then | Note |
|---|---|---|---|
| 2026-08-27 | all | $136.85 | First build, on the 2026 Q2 basis. Calibrated to the reaffirmed FY2026 guide, the $7-9B exit ARR, the $37.5B RPO runoff schedule and the $11.96B FY2027 consensus. |