Nebius Just Signed $46 Billion in AI Cloud Deals With Microsoft and Meta. Can This Stock 10X?

Nebius has stacked multiyear GPU-infrastructure contracts that create an initial $22.4B backlog and could expand to >$46B (Microsoft: $17.4B + $2B optional; Meta: $3B initial then a $12B extension with up to $27B possible). Revenue surged 479% to $530M last year; management plans to scale connected capacity to 800–1,000 MW by end-2026 (from 170 MW at end-2025) with Nvidia support to target 5,000 MW by 2030, and analysts forecast ~$28B cumulative revenue over the next three years, signaling material upside versus the current $29B market cap.

Analysis

Nebius sits at an inflection where hyperscaler demand, vendor financing and GPU supply dynamics interact — the immediate second-order winners are GPU fabs and power providers that can scale capacity quickly, while legacy CPU-centric vendors and generalist colo operators face margin pressure as GPU-optimized clusters capture a premium. The real moat will come not from raw racks but from integrated procurement and site-readiness (land, PPA, substations, permitting) — customers pay for guaranteed ramp timing as much as raw FLOPS, which creates stickier, higher-visibility revenue if Nebius nails project delivery.

Key risks cluster around execution cadence and component lead times: a 12–36 month horizon is where revenue recognition and cash conversion will be decided, and slippage in sites, transformers or GPU shipments can flip attractive forward economics into heavy dilution. Counterparty concentration (a few large hyperscalers) amplifies this — contract duration mitigates but does not eliminate demand re-pricing or scope renegotiation risk, and software/stack commoditization over 2–5 years could cap large gross-margin expansion.

From a portfolio perspective, the upside is asymmetric if Nebius converts capacity on schedule because scarcity pricing for GPU-hours is real; downside is concentrated and binary if buildouts stall or architectural changes shift demand back on-prem or to vertically integrated hyperscalers. Watch four operational KPIs weekly/monthly — MW connected, GPU installation rate, realized $/GPU-hour, and customer take rates — as they map directly to cash burn inflection points and implied valuation re-rates.

Key Decisions for Investors

• Long NBIS (size 2–4%): enter on any 10–20% pullback; horizon 12–36 months. Reduce cost basis with a covered-call program (sell 9–12 month OTM calls ~30% above entry). Hedge execution risk by buying 18–30 month puts 25–30% OTM sized to limit downside to ~15% of position value. Risk/reward: asymmetric — limited financed downside vs multi-bagger upside if buildout and pricing hold.
• Long NVDA call spread (ticker NVDA): small tactical position (0.5–1% fund) using 9–15 month calls to capture GPU supply tightness; sell higher strike to fund premium. Timeframe 6–12 months. R/R: limited downside ( premium ) and 3–5x capped upside if GPU ASPs and win rates accelerate.
• Short INTC (size 1–2%): thematic hedge against CPU-centric incumbents losing share to GPU-first infrastructure over 12–24 months. Use a short stock or buy 12–18 month deep OTM puts to cap tail risk. R/R: asymmetric — moderate expected decline if displacement accelerates, but protect for idiosyncratic recovery.
• Event monitoring & stop triggers: if NBIS misses monthly MW connected targets by >15% for two consecutive months or realized $/GPU-hour falls >20% vs guidance, reduce NBIS exposure by 50% within 48 hours. Conversely, if NBIS posts sequential tightening in realized pricing and GPU installation >20% MoM, add to NBIS up to +50% of initial sizing.