This Under-the-Radar AI Stock Could Be a Better Buy Than Nvidia

Nvidia holds warrants/owns ~22 million Class A shares of Nebius valued at $2.1B (roughly a 7.7–8.3% stake). Nebius reported Q4 revenue up 547% y/y to $228M and core AI revenue up 802% y/y to $214M, and management expects ARR of $7–9B by year-end (vs $1.25B at end-2025); Wall Street analysts forecast revenue to nearly triple next year. The company is investing heavily in data-center buildout and is operating at major losses to fund growth, leaving long-term margin outcomes uncertain despite dramatic top-line expansion.

Analysis

Nebius’ commercial relationships with platform incumbents create asymmetric advantages in procurement and early access, but the real moat will be unit economics — not headlines. Data-center scale gives operating leverage on fixed costs (real estate, power, networking) so a 5–10 percentage-point increase in overall GPU utilization can plausibly move incremental gross margin by multiples; conversely, any prolonged GPU oversupply or spot-price erosion will compress that leverage quickly and expose a capital-intense buildout.

Second-order winners include high-density power and cooling vendors, optical networking suppliers, and third-party managed-service operators who can piggyback Nebius’ stack; losers are legacy colocators and smaller GPU cloud providers that lack preferential hardware flow and long-term contracts. Hyperscaler customers act as both demand anchors and governance risks — contract concentration accelerates scale but concentrates counterparty and pricing renegotiation risk if those partners internalize capacity.

Time horizons matter: expect supply-cycle and margin signals within 3–12 months (GPU allocation, utilization, spot-hour pricing) and a clearer margin inflection only after 12–36 months as software monetization and higher-margin offerings scale. Key reversal triggers are a sustained decline in GPU pricing, a shift of preferential allocations away from third parties, or slower enterprise model adoption; upside acceleration would come from SaaS-like revenue conversion of tooling and orchestration, which could convert a cash-burning build into 30–40%+ EBITDA margins at scale.

Key Decisions for Investors

• Establish a tactical long NBIS position sized 1–2% of portfolio (12–24 month horizon). Treat as asymmetric bet: upside target 2.5–3x if utilization/mix ramps and software monetization begins; hard stop at -50% to limit capex and execution risk.
• Buy long-dated NBIS call LEAPS (18–30 months) representing 0.5–1% notional and hedge with a near-term (6–9 month) put to cap downside to ~30%. Rationale: preserves upside convexity while protecting against short-cycle GPU/capex shocks.
• Pair trade: go long NBIS / short NVDA at a 1:0.6 notional ratio for 6–12 months to express idiosyncratic scale-up upside while hedging platform/market risk. If NBIS outperforms NVDA by >50% take profits; if NBIS underperforms by >30% trim both legs.
• Overweight MSFT or META by 1–2% (buy-and-hold 9–12 months) as a defensive hedge to Nebius concentration risk — these names capture AI spend upside regardless of which cloud vendor ultimately owns bulk capacity; expect moderate upside with lower drawdown vs pure NBIS exposure.