Pre-Event Setup
Nebius entered Q1 2026 earnings on May 13 at $179-193, 33-43% above our $135 base case and within striking distance of the $195 bull case. The pre-event playbook flagged asymmetric downside risk and explicitly invoked the CoreWeave Q1 template: revenue beat + Q2 guidance miss = -11.68% selloff. Six scenarios were defined, with Scenarios 2-3 (moderate beat / in-line) carrying 55% combined probability and the strong-beat Scenario 1 only 12%.
What Actually Happened
| Metric | Pre-Event Expectation | Actual Q1 2026 | Verdict |
| Group revenue | $388M consensus, $400M bull trigger | $399M (+684% YoY, +75% QoQ) | Beat consensus, $1M short of strict bull trigger |
| Nebius AI revenue | n/a | $390M (+841% YoY, 98% of group) | — |
| Adj EBITDA margin (group) | 24% Q4 → 30%+ to confirm bull | 32% ($130M, vs $15M Q4) | Strong beat |
| Adj EBITDA margin (Nebius AI) | n/a | 45% (vs 24% Q4 core cloud) | Sharp operating leverage |
| Annualized run-rate (Nebius AI) | $1.25B Q4 → acceleration | $1.9B (+50% QoQ) | On track |
| FY2026 revenue guidance | $3.0-3.4B → raise to confirm bull | Reiterated $3.0-3.4B | Held, not raised |
| FY2026 ARR guidance | $7-9B | Reaffirmed $7-9B | Held |
| Year-end contracted power | 3GW target | Raised to >4GW (vs 3GW prior) | Material capacity raise |
| 2026 CapEx | $16-20B | Raised to $20-25B (citing 2027 demand) | Up |
| New site disclosure | None modeled | 1.2GW Pennsylvania AI factory (250-300MW live by end-2027, full by 2030) | Net new |
| Capital raised | n/a | >$6B in debt + equity | Liquidity reinforced |
Stock reaction: +7% on print day (May 13 close $207.27), then gapped on May 14 to intraday high $223.66, closed $222.82, and printed an all-time high of $233.73 intraday — a 15-21% move from the pre-earnings $193 close. Sell-side responded with broad PT raises: BofA $205→$240, Northland $215→$248, Citizens $175→$270, Citi to $287.
The CoreWeave Template Broke
Our pre-event playbook anchored on the CRWV Q1 precedent: a revenue beat plus a forward-guidance miss equals a sharp selloff. NBIS arguably delivered a strictly similar setup — revenue beat ($399M vs $388M consensus), revenue guidance reiterated (NOT raised), $1M short of the $400M strict bull trigger. By the CRWV template the right call was -8 to -15% to the $165-175 range.
That isn't what happened. The +15% move to ATH was a Scenario-1 reaction on a Scenario-2 print. Two things broke the template:
- The margin print did the talking. Group adj. EBITDA margin jumped from 7% (Q4) to 32% in a single quarter — and the Nebius AI segment hit 45%, well above the 40% year-end target. This was a one-quarter pull-forward of the operating leverage narrative that bears had argued NBIS could not credibly deliver. The market read this as the bottleneck risk being resolved, irrespective of the headline revenue line.
- The capacity raise reframed the multiple. Year-end contracted power guidance raised from 3GW to >4GW, 2026 capex raised to $20-25B, and a new 1.2GW Pennsylvania site disclosed — all citing pre-committed 2027 customer demand. This shifted the implied 2027 setup before management even issued explicit 2027 revenue guidance.
The lesson for the playbook: when a neocloud delivers a margin breakthrough alongside a capacity raise, the "beat-then-sell" template does not apply. CRWV's selloff was driven by margin pressure under volume growth; NBIS was the inverse setup.
Prediction Scoring (Brier Method)
| # | Prediction | Confidence | Outcome | Brier |
| 1 | Q1 revenue meets/exceeds $388M consensus | 0.50 | ✓ TRUE ($399M) | 0.250 |
| 2 | Q1 revenue exceeds $400M (strict bull trigger) | 0.35 | ✗ FALSE ($399M, $1M short) | 0.123 |
| 3 | Connected power >250MW at Q1 end | 0.50 | — UNRESOLVED (disclosure metric changed to contracted) | n/a |
| 4 | Initial 2027 revenue guidance or quantitative color | 0.65 | ✗ FALSE (no specific 2027 revenue guide) | 0.423 |
| 5 | FY2026 revenue guidance raised above $3.4B | 0.15 | ✗ FALSE (reiterated $3.0-3.4B) | 0.023 |
| 6 | Q1 adj. EBITDA margin (group) >30% | 0.30 | ✓ TRUE (32%) | 0.490 |
| 7 | Eigen AI close confirmed or imminent on call | 0.75 | ✗ FALSE (not specifically confirmed in transcript) | 0.563 |
| 8 | Token Factory inference ARR/customer metrics disclosed | 0.30 | ✗ FALSE | 0.090 |
| 9 | NBIS stock moves >12% within 2 trading days | 0.55 | ✓ TRUE (+15%) | 0.203 |
| 10 | Third hyperscaler $3B+ TCV contract announced | 0.08 | ✗ FALSE | 0.006 |
Average Brier (9 resolved): 0.241 (lower is better; 0 = perfect, 0.25 = pure coin flip)
Calibration: - Above-50% confidence predictions: 2/4 correct (#1, #9 hit; #4, #7 missed) - Below-50% confidence predictions: 4/5 correctly low (#6 was the surprise — the model underweighted the margin breakthrough)
Worst miss (Brier 0.563): Prediction #7 — overconfident at 0.75 that Eigen AI would be specifically called out as closed/imminent during earnings. The deal was already announced (May 1, BusinessWire) and management spoke broadly about Token Factory integration, but the transcript did not contain explicit close-status confirmation. Lesson: separate "deal announced" from "deal closed" thresholds more crisply in future pre-event cards.
Second-worst miss (Brier 0.490): Prediction #6 — only 0.30 confidence on >30% group adj. EBITDA margin. The actual 32% was a clean beat and the AI segment 45% was a blowout. Bears (including our pre-event note) overweighted the Q4 7% group margin as a baseline. Lesson: when a company has explicit margin target convergence ahead, weight the trajectory more than the trailing-quarter level.
Thesis Updates Triggered
| Ticker | Update Type | Action |
| **NBIS** | Major: scenarioAnalysis, watchTriggers, priceTargets, monitoring | Retire $135 base case; revise upward to $190-220 range; bull case toward $260+. Drop the "Q1 rev <$400M" bear trigger (technically tripped at $399M, market reaction proved it irrelevant). Update monitoring to track contracted capacity (target: 4GW year-end) rather than connected MW, which the company de-emphasized in disclosure. |
| **META** | Minor: notes | $27B partnership structure clarified ($12B dedicated + $15B optional). NBIS execution de-risks Vera Rubin 2027 deployment timing — log to notes; no scenario change. |
| **NVDA** | Minor: notes | NBIS hyper-acceleration validates neocloud GPU demand beyond hyperscalers; the $2B equity stake looks well-timed. Note only. |
| **CRWV** | Moderate: scenarioAnalysis, watchTriggers | NBIS demonstrates the neocloud category supports multiple high-margin winners (NBIS AI 45% vs CRWV stress on guidance). The 5x scale lead CRWV had in operational MW is narrowing as NBIS contracted base hits 3.5GW. Add NBIS execution milestones to CRWV watch triggers. |
Trade Recommendations
- ✓ Bull confirmation triggers fired (rev beat, margin breakthrough, +12% move), but stock already at ATH $233 — pre-event positioning suggestion explicitly said "do not chase" on large up moves. Hold any existing position; do NOT add at $222-233.
- Original $160 accumulation zone is invalidated by the post-earnings revaluation. New evidence-based accumulation window: $185-200 on any normal pullback (subject to re-validation at Q2 print).
- ✗ Prediction #2 (rev >$400M strict bull trigger) marginally failed at $399M — bull confirmation logic should be applied in spirit; the EBITDA and capacity beat compensate for the $1M shortfall.
- CRWV relative trade: If the execution gap persists into Q2, consider CRWV underweight vs NBIS overweight rotation. NBIS now has the better operating-leverage story.
- Remaining risk: 2027 revenue guidance still undisclosed. The Meta $27B / NBIS execution credibility test is in Q2-Q3 earnings.
Playbook Lessons
- Margin breakthroughs override revenue trigger arithmetic. A strict $400M revenue floor was wrong framing when the company delivered 32% group EBITDA — markets correctly priced the margin signal as more important than a $1M revenue shortfall.
- Disclosure-metric changes are themselves a signal. NBIS de-emphasized connected MW and elevated contracted GW — this is bullish (the bigger metric is healthier). Future predictions need to anticipate disclosure pivots, not be invalidated by them.
- "Beat-then-sell" template requires the right setup. CRWV's template needs margin compression under volume growth; NBIS was margin expansion under volume growth, the inverse setup. Distinguish the two next time.
- Pre-event positioning won. "Do not add pre-earnings" was correct — the asymmetric downside risk was real, but the asymmetric upside also wasn't capturable without taking the risk. The positioning was correct given the information available.