Meta Just Signed a $27 Billion Artificial Intelligence (AI) Deal With This Under-the-Radar Stock. Is Nebius a Buy for 2026?

Meta agreed to a five-year, $27.0B AI infrastructure deal with Nebius (including $12.0B of dedicated capacity starting in early 2027 and up to $15.0B in additional purchases). Nebius shares jumped ~17%, lifting market cap from just over $28B to above $32B; Nvidia also committed $2.0B in a strategic investment last week. Nebius reported 2025 revenue of $530M (+479% YoY) but an operating loss of $596M (worsened 49%), and the stock trades at ~57x sales, signaling high growth but high risk.

Analysis

This Meta–neocloud dynamic is a validation event for the specialist GPUaaS model: large AI buyers are willing to outsource incremental scale rather than solo-build every increment. That creates a two-tier demand structure where hyperscalers keep base layers internal while outsourcing burst/scale layers to specialist operators — a structural tailwind for well-capitalized neoclouds but one that centralizes negotiating leverage upstream with the GPU suppliers.

Upstream capacity and non-GPU infrastructure (power, substations, optical backhaul, cryo/cooling equipment) become strategic chokepoints. Firms that secure long lead-time power contracts, brownfield sites, and preferential GPU allocations will see materially higher long-run utilization and margin; conversely, firms that are late to secure these inputs face multi-quarter delivery slippage and margin dilution.

Key risks are execution and supply concentration: delivery timelines for gigawatt-scale capacity, single-vendor GPU dependency, and rapid competitive response from hyperscalers that accelerate insourcing or vertical integration. Near-term sentiment can stay elevated, but disappointment on utilization, revenue recognition cadence, or GPU delivery windows would compress multiples sharply.

Market pricing now implies near-perfect scale and GPU allocation certainty; that appears vulnerable to delivery slippage or an industry-wide supply rebalancing. The clean arbitrage is between upstream suppliers who capture scarcity rents versus operators who must demonstrate sustainable unit economics once depreciation, power, and long-term contracts are accounted for.

Key Decisions for Investors

• Small tactical long NBIS (size 0.5–1.5% NAV) via long-dated call spread (12–24 month LEAPs) to cap downside: buy NBIS Jan-2028 call, sell a higher strike Jan-2028 call to fund ~50–70% of premium. Entry on a <15% pullback from current levels or after the company reports a step-up in utilization metrics; risk: large downside if delivery misses; reward: multi-bagger if scale and pricing hold.
• Buy exposure to GPU supplier leadership (NVDA) via 12–18 month LEAPS calls (target 2:1 upside potential vs capital at risk). Rationale: vendor with de facto allocation power will capture scarcity margin even if some operators fail; exit on signs of centralized GPU oversupply or material guidance cuts from the vendor.
• Pair trade: long hyperscaler cloud leaders (AMZN or GOOGL or MSFT, overweight one based on valuation/timing) vs short selective neoclouds with weak balance sheets (size neutral) over 6–24 months. Thesis: hyperscalers can internalize more capacity faster and squeeze specialist spreads if GPU supply normalizes; keep pairs small and hedge market beta.
• Income trade on NBIS if you prefer conservative exposure: sell near-term covered calls or sell a short-dated call spread after a rally to monetize elevated volatility. Target 6–8% premium capture over 30–90 days, but keep position size capped and monitor utilization updates as hard triggers to lift covered calls.