Nebius Stock Jumped 15% on Its Meta Deal. Is This the Next CoreWeave -- or Something Better?

Nebius secured a five-year Meta infrastructure deal worth up to $27 billion (including $12B of dedicated capacity/$15B contingent), and its stock is up nearly 400% over 12 months (202% in 2025 alone), while CoreWeave is larger ($40.7B vs Nebius $25.2B) but Nebius is growing faster. Q4 figures: CoreWeave revenue $1.57B (+110% YoY) with adjusted net loss $(284)M and Q4 capex $8.2B; Nebius revenue $227.7M (+547% YoY) with net loss $(173)M and Q4 capex ≈$2.0B (FY capex $4.06B). CoreWeave expects $12–13B revenue in 2026 (from $5.1B in 2025) backed by >$46B of deals (OpenAI and Meta), so investors should weigh Nebius's accelerating growth and landmark Meta/Microsoft contracts against both companies' heavy capex and sizable operating losses.

Analysis

Scale is the single biggest durable moat in GPU-infrastructure: the firm with the largest, diversified fleet secures earlier access to next-gen accelerators, lower per-unit build costs and more favorable power/PPA economics. That creates a nonlinear margin benefit — a 20–30% advantage in effective cost-per-GPU can translate into 400–800bps of gross-margin delta once utilization normalizes, because fixed site and power costs are amortized across more high-value cycles.

Counterpart concentration and build-timing create asymmetric outcomes. Large, multi-year offtakes de-risk top-line visibility but front-load balance-sheet and recognition risk: revenue only materializes as racks spin up and customers consume; meanwhile stranded fixed commitments (power, leases, GPUs) amplify downside if demand lags. A 12–24 month lag between capital deployment and stable cash conversion is realistic, so short-term earnings will remain volatile even if backlog is strong.

GPU supply and export-control regimes are the key external hooks that can flip this trade. A tightening allocation or country-specific sales curbs would disproportionately punish faster growers that bought forward hardware but lack diversified sourcing. Conversely, a loosening of supply (or a faster-than-expected secondary market for used accelerators) would compress replacement costs and accelerate margin inflection within 6–12 months.

Net: valuations are pricing a path to scale and margin convergence; the outcome is binary within a 12–24 month window. For investors, the cleanest way to capture upside is structured optionality around the monetization timeline while explicitly hedging the execution and regulatory tails that would cause heavy re-rating in either direction.

Key Decisions for Investors

• Long NBIS via structured options (ticker: NBIS) — buy a 9–15 month call spread (~30% position-sizing of a directional allocation, pay premium only) sized to limit max loss to the premium; target 2x–3x return if 12–18 month capacity monetization accelerates. Hard stop: exit if shares fall 30% from entry or if two consecutive quarterly build metrics miss internal targets.
• Relative pair: long NBIS / short CRWV (tickers: NBIS/CRWV) — equal-dollar pair over 6–12 months to isolate execution vs scale. Size to keep net market exposure neutral (e.g., 60% long NBIS : 40% short CRWV by notional if one is larger); target 20–30% relative outperformance, risk is a sector-wide GPU re-rating that lifts both.
• Tail hedges on CRWV (ticker: CRWV) — buy 6–9 month puts sized at 1–2% of portfolio to protect against a rapid re-rating from capex/margin disappointments or GPU supply squeeze. Expect to pay a premium; these acts as insurance if market re-prices growth without margin proof points.
• Event hedge via hyperscaler optionality (tickers: MSFT, META) — monitor earnings/commentary windows for these customers and use short-dated OTM put spreads on MSFT/META as cheap insurance around disclosures that could signal pull-through weakness. Keep these hedges small (0.5–1% portfolio) but live into the next two hyperscaler reporting cycles.