Nvidia Just Dumped This AI Stock, but Here's Why That Might Be a Buying Opportunity

Key event: Nvidia sold its entire 7.7 million-share stake in Applied Digital for about $177 million in Q4 2025, signaling a cautious repositioning. Applied Digital grew revenue from $8.5M (fiscal 2022) to $215.5M (fiscal 2025) and has ~$16B in contracted lease payments over 15 years (~$1B/year), but 93% of 2025 revenue came from a single customer and the company is expected to remain GAAP-unprofitable until ~fiscal 2028. Analysts project revenue rising to roughly $1.05B by fiscal 2028, but the planned ChronoScale spin-off and customer concentration create execution and growth risks despite long-term REIT-like upside potential.

Analysis

Macro/competitive: The mid-tier data-center landlord segment is exposed to a classic scale-versus-flexibility bifurcation — hyperscalers and large REITs will increasingly extract premium pricing and longer tenors as they internalize more GPU-heavy workloads, leaving smaller landlords to compete on bespoke power/cooling and geographic niche advantages. This raises margin compression risk for exposed landlords while boosting OEMs of specialized infrastructure (power distribution, liquid cooling) that can standardize deployment and shorten build cycles. Expect this dynamic to play out unevenly over the next 12–36 months as capacity additions and contract renewals reveal who can hit 80%+ utilization on multi-year contracts.

Risk & catalysts: The two binary drivers that will reprice marginal names are (1) conversion of contracted lease backlog into cashflow (timing and commencement certainty), and (2) the outcome of any corporate restructurings or asset-liability swaps that crystallize leverage ratios. Near-term indicators to monitor: monthly capacity commencements, single-customer revenue cadence, and secured financing terms on new builds — each can move multiples 30–50% within a single quarter if surprising. Tail risks include a counterparty retrenchment by a large tenant or a spike in capital costs that forces equity dilution; these can unfold in weeks but usually become irreversible over 6–18 months.

Contrarian/second-order view: The market may be overstating absolute demand risk and understating the value of long-duration contracted cashflows; properly underwritten 15-year leases behave like high-grade bonds and can support REIT-like valuations once leverage is rationalized. That implies asymmetric option-like payoffs: downside is limited to execution failure, upside material if the company achieves scale and converts to a REIT or secures covenant-light funding. For investors, skew toward trades that monetize high convexity (calendar spreads, pair trades vs scale incumbents) rather than outright equity punts on execution-risky names.

Key Decisions for Investors

• Pair trade (6–18 months): Long EQIX (or DLR) 3–5% portfolio weight vs short APLD 1–2% notional. Rationale: capture yield/stability premium of incumbent REITs while shorting execution and concentration risk at a smaller landlord; target 25–40% relative outperformance. Risk management: equal-dollar pair, stop-loss if APLD reports >20% sequential customer diversification or EQIX misses revenue guidance.
• Directional hedge (3–9 months): Buy CRWV (CoreWeave) downside via 3–6 month put spread (buy 25–30 delta puts, sell 10–15 delta puts) sized to offset 50% of GPU-dependent landlord exposure. Rationale: protects fund if GPU-cloud demand rebalances or unprofitable cloud operators cut capacity; reward: limited premium, high convexity if spending contracts. Risk: loss limited to premium; close if CRWV posts sustainable gross margin expansion.
• Asymmetric long (9–15 months): Buy NVDA 9–12 month call spread (buy ~50-delta LEAP, sell ~30-delta higher strike) sized to 1–2% portfolio. Rationale: maintain exposure to GPU demand tailwinds without full vega; expected payoff >2x if enterprise AI capex continues. Risk: premium decay and single-quarter softness; hedge by selling near-term calls against position if implied vol spikes.
• Event/arbitrage (3–9 months): Small, conditional long EKSO position (or merger-arb if applicable) sized 0.5–1% with strict deal-fail stop. Rationale: execution of corporate actions or asset consolidation can re-rate small-cap infra/service names. Risk/reward: asymmetric — limited capital at risk vs potential rerating; exit if definitive agreement terms change or financing conditionality appears.
• Convexity trade (12+ months): Buy long-dated OTM put protection on names with >60% customer concentration (size 0.5–1% of portfolio) rather than holding larger cash shorts. Rationale: caps downside from counterparty retrenchment while preserving upside if execution succeeds. Risk: premium cost over time; reevaluate quarterly and roll if concentration risk remains.