NBIS | Earnings Review — Q1 2026

Nebius printed the first dual beat of the pure-play Nebius era. Group revenue of $399.0M beat FMP consensus of $375.1M by +6.4% and grew +622% YoY vs $55.3M (management stated +684% on a slightly different perimeter). Street EPS (FMP non-GAAP proxy) was −$0.23 vs −$0.77 estimate — a +$0.54 beat. GAAP diluted EPS of $2.11 is not the operating story: it embeds a non-cash ClickHouse equity revaluation gain of $780.6M. Net income of $621.2M is similarly distorted; loss from operations was still −$128.0M.

Operating inflection is the real print. Group adjusted EBITDA swung to $129.5M (32.5% margin) from $15.0M in Q4'25 and −$62.6M in Q1'25. Core AI cloud revenue of $390.0M was ~98% of group; Core AI adj. EBITDA of $174.0M implies ~45% margin (from 24% in Q4'25 — ~+2,100 bps). Nebius AI ARR exited March at $1,920M, +54% vs YE25 $1,250M.

Guidance: hold on P&L, raise on build. FY2026 package reiterated: ARR $7.0B–$9.0B, group revenue $3.0B–$3.4B, group adj. EBITDA margin ~40%. CapEx raised to $20–25B (from $16–20B) and contracted-power target to ≥4 GW (4,000 MW) for 2027 demand already under commitment — not component inflation (low-single-digit % of 2026 program per management). Street FY2026 rev $3.37B already sits at the top of the guide band.

Tone: confident, execution-led, multi-vector (capacity, product, customers, capital). Not promotional; no P&L raise after a large beat. Explicit Q2 adj. EBITDA margin dip, then Q3 back to Q1 levels and Q4 higher. Demand language remains sold-out; pipeline +3.5× QoQ excluding hyperscaler mega-deals.

Contradictions (3): medium flag on Meta $27B "formal contract" vs $12B firm dedicated + $15B Nebius-option capacity; two low-severity ARR hygiene slips ($700 vs $750 prior guide; YE25 ARR $1.20B vs $1.25B base). No chronic contradiction pattern.

Near-term catalysts: Aug 12 Q2 print (~$578M street rev; pre-wired margin dip); H2 capacity / MSFT tranche; ABF scale after first ~$775M secured facility; Vera Rubin first rack validated Jul 23; Meta $15B option economics.

Trajectory over absolutes: revenue YoY re-accelerated two quarters off a Q3'25 trough (+237% → +501% → +622%). Capacity is the binding constraint, not demand. GAAP EPS is unusable; underwrite ARR, revenue growth, and adj. EBITDA margin.

Verdict — accelerating. Revenue YoY re-accelerated from a +237% trough in Q3'25 to +501% (Q4'25) and +622% (Q1'26) on an absolute base of $399.0M. Adj. EBITDA flipped from multi-quarter losses to 32.5% margin; Core AI ~45%. ARR path $249M → $1.92B in four quarters under a $7–9B YE26 guide.

Revenue, ARR, margins

| Metric | Q2'24 | Q3'24 | Q4'24 | Q1'25 | Q2'25 | Q3'25 | Q4'25 | Q1'26 | |---|---|---|---|---|---|---|---|---| | Group revenue ($M) | $24.9 | $43.3 | $37.9 | $55.3 | $105.1 | $146.1 | $227.7 | $399.0 | | Revenue YoY % | +430% | +766% | +466% | +385% | +322% | +237% | +501% | +622% | | Core AI cloud rev ($M) | — | — | — | — | — | — | $214.0 | $390.0 | | ARR e/o ($M) | $80 | $120 | — | $249 | $430 | $551 | $1,250 | $1,920 | | Gross margin % | 27.3 | 56.4 | 27.4 | 46.7 | 71.4 | 70.6 | 69.9 | 74.0 | | Adj. EBITDA ($M) | −68.1 | −51.9 | −75.5 | −62.6 | −21.0 | −5.2 | $15.0 | $129.5 | | Adj. EBITDA margin % | nm | nm | nm | (113) | (20) | (3.6) | 6.6 | 32.5 | | Core AI adj. EBITDA ($M) | — | — | — | — | — | — | $51.8 | $174.0 | | Core AI adj. EBITDA margin % | — | — | — | — | — | 19 | 24 | ~45 | | GAAP diluted EPS ($) | — | — | — | −0.48 | — | — | — | $2.11 |

Read: absolute revenue roughly doubled every 1–2 quarters from mid-2024. YoY decelerated five quarters into Q3'25 as the base rose and capacity stayed sold-out, then re-accelerated two quarters as MSFT/Meta capacity converted. Gross margin stabilized ~70%+ (Q1'26 74.0%, +2,733 bps YoY). Adj. EBITDA path (113)% → (20)% → (4)% → +7% → +32.5% over five quarters is the operating leverage story.

Annual (Nebius-comparable USD)

| Metric | FY2023 | FY2024 | FY2025 | |---|---|---|---| | Revenue ($M) | $20.9 | $117.5 | $529.8 | | Revenue YoY % | — | +462% | +351% | | Gross margin % | (52.6) | 37.5 | 68.6 | | Adj. EBITDA ($M) | −282.8 | −266.4 | −64.9 | | ARR e/o year ($M) | — | — | $1,250 |

Q1'26 alone is ~75% of full-year FY2025 revenue and positive adj. EBITDA exceeding the entire FY2025 loss. FY2026 guide mid $3.20B implies ~+504% YoY vs FY2025.

Balance-sheet / cash (context for CapEx)

| Metric | Q4'25 | Q1'26 | |---|---|---| | Cash & equiv. ($M) | $3,678 | $9,298 | | Operating cash flow ($M) | — | $2,258 (prepayments) | | CapEx ($M) | — | −$2,473 | | Deferred rev current + non-current ($B) | ~$1.58 | ~$4.78 ($686M + $4,093M) |

Headline: first clean dual beat of the Nebius pure-play window. Revenue had missed street five consecutive quarters before this +6.4% beat. Street EPS is a non-GAAP/operating proxy (FMP); do not score beat/miss on GAAP EPS when investment marks dominate.

| Quarter | Rev Cons. | Rev Act. | Rev Surp. | Rev | EPS Cons. | EPS Act. (street) | EPS Surp. | EPS | |---|---|---|---|---|---|---|---|---| | 2024Q4 | $58.1M | $37.9M | −34.8% | Miss | −$0.39 | −$0.43 | −9% | Miss | | 2025Q1 | $57.7M | $55.3M | −4.2% | Miss | −$0.45 | −$0.44 | +2% | Beat | | 2025Q2 | $105.4M | $105.1M | −0.3% | Miss | −$0.50 | −$0.38 | +24% | Beat | | 2025Q3 | $157.9M | $146.1M | −7.5% | Miss | −$0.56 | −$0.39 | +30% | Beat | | 2025Q4 | $246.0M | $227.7M | −7.5% | Miss | −$0.58 | −$0.99 | −71% | Miss | | ▶ 2026Q1 (THIS) | $375.1M | $399.0M | +6.4% | Beat | −$0.77 | −$0.23 | +70% | Beat |

| Window | Revenue beat rate | EPS beat rate (street) | |---|---|---| | Available history (L6Q) | 1 / 6 = 16.7% | 4 / 6 = 66.7% | | Last 4 quarters (L4Q) | 1 / 4 = 25.0% | 3 / 4 = 75.0% |

Pattern: Mixed — chronic revenue misser until this quarter; EPS mostly beats (losses less deep than street) with one large dual miss in Q4'25. Surprise magnitude improving into this print: revenue surprises moved from large misses → small misses → first clear beat; Q1'26 is the largest EPS $ surprise in the sample (+$0.54).

Management variance explanation (not consensus-referenced on call): sold-out capacity, pricing raises still clearing, ARR +50%+ QoQ, utilization of capacity put in service. ClickHouse mark explains GAAP-only profit.

EPS definition bridge (required):

| Definition | Q1'26 | Use for beat/miss? | |---|---|---| | Street estimate | −$0.77 (FMP) | Denominator | | Street actual | −$0.23 (FMP) | Yes | | GAAP diluted EPS | $2.11 | No — investment mark | | Equity reval gain | $780.6M | Explains gap |

FY2026 formal package — reiterated P&L, raised CapEx / power

| Metric | Prior (Q4'25 call) | New (Q1'26 call) | Change | Daloopa | |---|---|---|---|---| | ARR YE2026 | $7–9B | $7–9B | Unchanged | $7.0B / $9.0B | | Group revenue FY2026 | $3.0–3.4B | $3.0–3.4B | Unchanged | $3.0B / $3.4B | | Group adj. EBITDA margin | ~40% | ~40% | Unchanged | ~40% | | CapEx FY2026 | $16–20B | $20–25B | +$4–5B range | Transcript | | Contracted power YE2026 | >3 GW | ≥4 GW | Raised ~+1 GW | 4,000 MW | | Connected power YE2026 | 800 MW–1 GW | 800 MW–1 GW on track | Unchanged | 800–1,000 MW |

Waterfall — revenue envelope vs CapEx

Guide vs street

| Metric | New mid | Consensus | vs Consensus | |---|---|---|---| | Group revenue FY2026 | $3.20B | $3.37B (11 analysts, FMP) | Street near high end (−$0.03 vs high / +$0.17 vs mid) | | ARR YE | $8.0B mid | Street models unevenly | North-star KPI for management | | Adj. EBITDA margin | ~40% | Models track if capacity lands | Aligned if H2 ramp delivers | | FY2026E EPS (street) | Not guided | −$2.64 (FMP, 6) | Definition noise (revals) | | Q2'26 rev (no FQ guide) | Path: sequential growth; step is Q3+ | $577.7M | Street implies ~+45% QoQ — aggressive vs back-end capacity |

ARR path to guide

| Point | ARR | Source | |---|---|---| | 2025Q1 | $249M | Daloopa | | 2025Q2 | $430M | Daloopa | | 2025Q3 | $551M | Daloopa | | 2025Q4 | $1,250M | Beat YE guide $0.9–1.1B | | 2026Q1 | $1,920M | +54% QoQ | | YE2026 guide mid | $8.0B | Need ~4.2× from Q1 exit |

Guidance evolution (raises, not flips)

| Call | ARR | Group rev | Adj. EBITDA | CapEx | Power | |---|---|---|---|---|---| | Q2'25 | YE25 raised $0.9–1.1B | Core $400–600M | Core positive ahead | ~$2B 2025 | >1 GW process | | Q3'25 | Introduced YE26 $7–9B | FY25 tightened | Slightly + YE25 | Capex for 800 MW–1 GW | Contracted 2.5 GW | | Q4'25 | $7–9B formal | $3.0–3.4B | ~40% | $16–20B | >3 GW target | | Q1'26 | Reiterated $7–9B | Reiterated $3.0–3.4B | Reiterated ~40% | Raised $20–25B | ≥4 GW |

Pattern: raise capacity/ARR ambition; keep in-year revenue deliberately conservative vs ARR math (capacity back-end weighted). Opposite of chronic over-promising on the P&L.

Tone — Q4'25 vs Q1'26

| Dimension | Q4'25 | Q1'26 | Delta | |---|---|---|---| | Posture | Confident entry; first full FY26 guide | "Great start"; execution across 4 dimensions | More operational / multi-vector | | ARR confidence | High; not dependent on new mega-deals | Reinforced with $1.9B exit + pipeline +3.5× | Data-backed | | Margin path | Group ~40% from 7% Q4 exit | Q1 already 32%/45% AI; Q2 dip → Q3/Q4 rebuild | More granular transparency | | CapEx / capital | $16–20B; exploring ABF | $20–25B; $9.3B cash; converts done; NVIDIA $2B; >90% of prior range secured | Aggressive build + tools in place | | Demand | Sold out | Sold out again; price raises still clear; Meta $27B structure | Hotter demand tone |

Synthesis: reiteration after a strong Q1 improves credibility of the annual package; street has already "maxed" revenue guide. Real guide change = CapEx/power for 2027. FQ+1 risk: street $578M Q2 looks optimistic vs management's H2-weighted capacity and near-term margin downshift.

| Metric | Q2'24 | Q3'24 | Q4'24 | Q1'25 | Q2'25 | Q3'25 | Q4'25 | Q1'26 | |---|---|---|---|---|---|---|---|---| | Revenue YoY % | +429.8% | +766.0% | +465.7% | +385.1% | +322.1% | +237.4% | +500.8% | +621.5% | | Rev accel (bps QoQ) | +27,645 | +33,621 | −30,033 | −8,058 | −6,300 | −8,467 | +26,338 | +12,073 |

Inflection markers

1. Q3'24 — peak YoY in window (+766%); last pure-acceleration quarter off a tiny 2023 base.
2. Q4'24 — first hard deceleration (−30,033 bps); absolute revenue slipped QoQ.
3. Q3'25 — YoY trough (+237%); capacity sold-out / mega-deal ramp timing.
4. Q4'25 — re-acceleration inflection (+26,338 bps to +501%); ARR beat YE guide; Meta capacity live early Feb'26.
5. Q1'26 (just-reported) — continued acceleration (+12,073 bps to +622%) on $399M; ARR exit $1.9B.

EPS YoY is not an operating signal. GAAP basic EPS swings (Q2'25 +$2.45; Q4'25 −$0.99; Q1'26 +$2.40) track investment revals, not unit economics. Underwrite revenue, ARR, and adj. EBITDA.

Drivers (transcripts): capacity is the bottleneck ("everything we deploy, we sell"); hyperscaler mega-deals (MSFT, Meta) convert; AI mix ~98% of group; ARR outruns quarterly revenue toward $7–9B; pricing/utilization tailwind (AI margin 24% → 45%); pipeline breadth +3.5× ex-mega-deals; CapEx raise signals persistence if power/GPU supply lands.

| Priority | Catalyst | Timing | Scoreboard | |---|---|---|---| | P0 | Q2'26 earnings print | 2026-08-12 | Rev vs ~$578M street; ARR step; Q2 margin dip magnitude; FY guide language (reiterate base case; street already at high end of $3.0–3.4B) | | P0 | H2 capacity step-up / MSFT tranches | Q3–Q4 2026 | Bridge from $1.92B ARR → $7–9B; Korolenko: "Q3 very significant… Q4 also very significant"; MSFT majority H2 | | P1 | ABF follow-on after $775M pilot | 2H26–27 | $20–25B CapEx needs scale beyond cash + converts; first secured facility ~$775M (Jul 17) validates path | | P1 | Vera Rubin customer-ready | H2 2026 | First NVL72 rack validated Jul 23 (Finland); Meta dedicated design / premium ASP optics | | P1 | Meta $12B dedicated from early 2027 | Structure live; rev mostly 2027 | FY2027 street rev ~$11.6B embeds Meta/MSFT; $15B option is financing + optional multi-tenant upside | | P2 | Pricing / sold-out / deferred rev | Continuous | Deferred rev stock ~$4.8B; OCF $2.26B; defends 40% YE margin if GPU scarcity eases | | P2 | Token Factory / inference attach | Ongoing | Differentiation vs bare-metal neocloud peers | | P3 | US site politics / power ≥4 GW | Through YE26 | Contracted already >3.5 GW; PA 1.2 GW long-cycle |

Already partially de-risked into Aug 12: ABF pilot closed; first Vera Rubin rack live; formal Q2 date locked. Live overhang: Meta Compute / neocloud sector sentiment (mid-summer drawdowns).

Management pre-wires for Q2: adj. EBITDA margin "a little bit lower" than Q1, return to Q1 levels in Q3, higher in Q4; full-year group margin still ~40%; capacity heavily H2-weighted.

Portal-driven Q&A (IR-selected), not live open queue — multi-part asks sometimes only partially answered. 8 of 15 Well Answered; 7 Deflected/Avoided.

Well answered (high signal)

| # | Analyst | Topic | Takeaway | |---|---|---|---| | 2 | Portal | CapEx raise: capacity vs inflation | Capacity / 2027 demand — not cost pressure; inflation low-single-digit % of 2026 program | | 3 | Kisner (Water Tower) | 45% AI margin vs ~40% FY guide | Explicit path: Q2 lower → Q3 back to Q1 → Q4 higher; FY group ~40% held | | 4 | Beale (Arete) | Capacity timing / Pennsylvania | PA lights-up end-2027 ~250–300 MW → 1.2 GW by ~2030; 2026 H2-weighted | | 6 | Platt | Meta $15B option | $12B dedicated from early'27 + $15B at Nebius discretion (Meta backstop); can exceed $27B economics | | 7 | Duval (GS) | Eigen / Clarifai M&A | Software as enabler of full-stack cloud; selective M&A for LTV / road map | | 10 | Chokshi (Northland) | Pipeline +3.5× | AI cloud only, excludes hyperscalers; conversion qualitative | | 11 | Portal | Software momentum | Inference fastest-growing; Token Factory PMF | | 15 | Portal | US DC politics | Efficiency / transparency / community partnership playbook |

Deflected / avoided (modeling gaps)

| # | Analyst | What was missing | |---|---|---| | 1 | Duval (GS) | Share of older shorter-term contracts that reprice — zero % | | 5 | Baer (MS) | Vineland NJ never named; portfolio "on schedule" only | | 8 | Portal | Token Factory / agentic monetization as distinct P&L — declined | | 9 | Liani (BofA) | Noncore disposals (ClickHouse / TripleTen / Avride) ignored | | 12 | Portal | Customer concentration % Meta/MSFT — pure narrative | | 13 | Kisner (Water Tower) | Concrete NVIDIA deliverables / Rubin H2 MW — strategy only | | 14 | Portal | Sold-out % of 2026/2027 capacity — "vast majority" only |

Common pattern: highly specific on strategy, structure, sequencing; systematically soft on book composition numbers (concentration, sold-out %, duration mix, software take-rate). Consistent with a hyper-growth neocloud protecting commercial terms — leaves residual on how durable/diversified revenue is once mega-deals are stripped out.

Q&A trajectory implications: capacity is the binding constraint; CapEx raise is pull-forward of 2027 demand; Meta $27B structure is the financing keystone; largest open risks are concentration, site-level delivery (Vineland), noncore capital recycling, software economics.

3 found — 1 medium, 2 low. No chronic multi-quarter narrative flip.

Not contradictions (explicit raises): contracted power path 1 → 2.5 → >3 → ≥4 GW; CapEx $2B → $5B (2025) → $16–20B → $20–25B (2026); ARR/revenue guides revised up with acknowledgment; mega-deal layering after signing; converts/NVIDIA equity after Q4 "minimal debt" language.

Management-quality lens: cross-quarter guidance honesty good (upward revisions labeled); numeric hygiene minor slips only; single decision-useful flag is Meta contract firmness.

Macro / industrial (AI supercycle, not classic consumer macro)

Management does not discuss Fed/CPI/retail PMI. Useful "macro" is AI-industrial: sold-out multi-quarter capacity, rising ASPs and contract duration, prepayments, mild component inflation (low-single-digit % of CapEx), power as strategic input, local DC politics, and government AI industrial policy (U.K., Israel).

| Theme | Read-through | |---|---| | Sold-out sector-wide | Bullish scaled capacity owners with power; constrains underfunded AI-native buyers without multi-year prepay | | CapEx raise = volume, not cost blow-up | Better for GPU / server / construction OEMs; financing markets open for contracted capacity | | Memory / storage shortages (Q4 color) | Early bookers locked 2025 pricing; late buyers face price and lead-time risk | | Power contracted ≥4 GW path | Constructive utilities, IPPs, transformers, switchgear, EPC; power is binding industrial constraint | | US DC political opposition | Rising NIMBY / grid-politics headwind; operators with community playbooks better positioned | | Capital markets | $4.3B converts at low coupons; NVIDIA $2B equity; ABF expected vs MSFT/Meta credit — not a "rates kill CapEx" narrative |

Companies mentioned (second-order)

| Company | Relationship | Implication | |---|---|---| | Meta (META) | $12B dedicated + $15B option structure | Meta still underbuilt vs internal AI demand; NBIS gets financing collateral + multi-tenant optionality | | Microsoft (MSFT) | Multi-year offtake; bulk H2'26 | External capacity for AI; execution risk headline-sensitive (Vineland) | | NVIDIA (NVDA) | $2B equity; Exemplar; Rubin/Vera supply; 5 GW LOS by 2030 | Preferred builder alignment; GPU scarcity still strategic | | Shopify, Cloudflare, monday.com, Revolut, Cursor, 1X, Sword Health | Customers / Token Factory | Broadening real offtake (enterprise, fintech, robotics, healthcare) beyond pure AI labs | | Eigen / Clarifai / Tavily | Acquisitions | Full-stack inference / agentic attach; pressure on pure-play optimizers without capacity | | ClickHouse | Equity stake (reval drove GAAP NI) | Data-infra valuations elevated; NBIS noncore optionality | | Unnamed neocloud peers | Competitors sold out into 2027 | Sector tightness; late entrants disadvantaged |

Sharpest third-party read-throughs: Meta $27B architecture and NVIDIA $2B equity are the strongest external validators of Nebius capacity quality and financing access.

NBIS Q1'26 is a capacity-sold-out, dual-beat, margin-inflection print with P&L guide held and CapEx/power raised for 2027. Trajectory is re-accelerating on revenue YoY for two quarters; ARR at $1.92B sits under a $7–9B YE guide that requires H2 capacity delivery, not demand discovery. Street already sits at the high end of FY rev guide into an Aug 12 print that should show sequential growth with a pre-telegraphed margin dip. Model Meta as $12B firm + $15B option, strip ClickHouse marks from earnings quality, and underwrite adj. EBITDA / ARR — not GAAP EPS.

Data sourced from Daloopa.