Back to News
Market Impact: 0.35

Better Artificial Intelligence Stock: Nebius vs. Super Micro Computer

NBISNFLXNDAQ
Artificial IntelligenceTechnology & InnovationCorporate EarningsCorporate Guidance & OutlookAnalyst EstimatesCompany FundamentalsRegulation & LegislationSanctions & Export Controls
Better Artificial Intelligence Stock: Nebius vs. Super Micro Computer

Nebius (formerly Yandex) is executing a rapid cloud-based AI infrastructure expansion: 2024 revenue jumped 462% to $118M with adjusted EBITDA of -$266M, analysts model 2025 revenue of $556M (up 373%) and expect adjusted EBITDA to improve to -$75M; management targets a $7–$9B annualized revenue run rate by end-2026 and analysts forecast $7.8B revenue in 2027 (CAGR ~305% from 2024). Super Micro Computer grew fiscal 2025 revenue 47% to $22.0B with adjusted EBITDA of $1.6B and is projected to reach $36.3B revenue in fiscal 2026 with $1.8B adj. EBITDA; analysts see 2025–2028 CAGRs of ~33% (revenue) and 26% (adj. EBITDA). Valuation contrasts are stark: Nebius EV $24.2B (~7x next-year sales) vs. Supermicro EV $16.6B (<1x next-year sales and ~6x adj. EBITDA), and although Supermicro faced past accounting/regulatory headwinds it now appears the more conservative buy while Nebius remains a high-growth but unproven, speculative bet.

Analysis

Market structure: The AI stack bifurcates into two clear winners — hyperscale/cloud GPU renters (Nebius/NBIS) and specialized server OEMs (Supermicro/SMCI), with Nvidia (NVDA) as the choke-point supplier. Nebius’ sold-out capacity and aggressive $7–9B run-rate target by end-2026 signal immediate pricing power for capacity, but SMCI’s <1x FY26 sales and 6x EBITDA valuation indicate the market still prices it as cyclical hardware rather than high-growth AI exposure. Persistent GPU and liquid-cooling constraints imply demand > supply for 12–24 months, supporting both server pricing and colo rents. Risk assessment: Tail risks include renewed export controls on datacenter GPUs or sanctions (high impact, ~10–30% downside instant shock), execution risk for NBIS scaling capex (must hit EBITDA+ by 2026) and repeat governance/regulatory issues at SMCI (audit/regulatory headlines can re-rate multiples quickly). Timeline: expect headline-driven volatility in days; supply/GPU cadence and colo builds matter over 3–12 months; structural profitability informative in 12–36 months when NBIS must convert run-rate to recurring ARR. Hidden dependency: both firms are hostage to NVDA GPU allocations and energy pricing for liquid cooling. Trade implications: Primary actionable bias is long SMCI (value + growth) and tactical, small speculative exposure to NBIS. Implement a pair trade: long SMCI (2–4% portfolio) financed by a short/equivalent notional NBIS position (1–2% or short calls) over a 6–12 month horizon to capture execution/valuation divergence. Use options: buy SMCI 12–18 month LEAP calls to cap downside while selling nearer-term covered calls post-earnings if IV falls; for NBIS, buy deep-OTM 2027 calls (<1% allocation) rather than a cash long to limit dilution and execution risk. Contrarian angles: Consensus underestimates NVDA allocation risk — if NVDA tightens supply, NBIS growth could stall while SMCI (with channel ties) captures share, meaning current consensus growth for NBIS is overoptimistic. Historical parallel: 2017–19 GPU cycles show rapid boom then price deflation once capacity exceeds demand; a cautioned play is to avoid paying >7x forward sales for NBIS today without EBITDA proof. Unintended consequence: aggressive NBIS capex could force deep dilution or discounted leases if demand normalizes; set hard triggers (see decisions) before adding material exposure.