Nvidia Sold All of Its Applied Digital Stock. Should You Follow Suit?

Nvidia sold its entire Applied Digital stake (about 7.7 million shares, roughly $177M) during Q4 2025 per its Dec. 31, 2025 13F, after previously closing similar-sized positions in ARM and SoundHound AI. Applied Digital reported fiscal 2026 Q2 revenue up 250% year-over-year (period ended Nov. 30) while net losses widened due to heavy buildout investments to meet AI data-center demand. The piece frames Nvidia's sale as portfolio rebalancing/noise and argues Applied Digital’s ramping capacity and strong customer demand could generate significant cash flows over time, though investors should be patient. Motley Fool’s Stock Advisor did not include Applied Digital in its top 10 list and discloses positions in Nvidia and SoundHound AI.

Analysis

The structural demand curve for AI compute remains asymmetric: front-loaded capital intensity and long lead times for commissioned capacity create a multi-quarter mismatch between order intake and revenue recognition. That favors owners/operators that can convert leased space into high-margin recurring cashflows once fully commissioned, and penalizes those that carry buildout risk without customer-backed contracts or flexible financing.

Second-order winners are suppliers that shorten deployment cycles — modular data center integrators, immersion-cooling vendors, high-voltage switchgear manufacturers and local utilities — because each 2–6 week reduction in deployment time materially shortens cash conversion cycles for operators. Conversely, hyperscalers moving to captive build models (or to alternative accelerators) would compress utilization and pricing for third-party wholesale racks, so the competitive moat is contingent on multi-year customer stickiness, not just footprint.

Key risks: (1) a durable drop in training intensity driven by model-efficiency breakthroughs could leave newly commissioned capacity distressed within 6–18 months; (2) a refinancing/contract cliff at smaller operators would force dilutive equity raises; (3) regulatory or grid-permitting delays in constrained markets can push ROI timelines from quarters to years. Watch for large counterparty take-or-pay announcements, utility tariff changes, and quarterly utilization disclosures — any of which can re-rate the group quickly.

Key Decisions for Investors

• Long APLD (APLD) — tactical 12–24 month position sized 3–5% of tech book. Use a staggered entry (25% now, 25% at 10% lower, remainder on confirmed utilization guidance) and collar with a sold call ~20% OTM and bought put ~15% OTM to cap downside; target 2.5x downside-protected return if builds convert to contracted cashflow, stop-loss if contract cadence misses by two consecutive quarters.
• Long Intel (INTC) — 6–18 month thematic trade (2–4% of portfolio) to capture non-GPU AI acceleration and edge compute wins. Risk/reward: asymmetric upside if customers diversify away from single-accelerator dependency; limit risk with 8–12% trailing stop and take-profits in 40–60% range on material share gains in AI platforms.
• Short SoundHound (SOUNW) — concentrated, event-driven short (1–2% notional) over 3–9 months: high beta to sentiment and small-cap funding cycles. Set tight risk control: initial stop at 20% adverse move, target 40–70% downside if institutional support wanes or funding windows close.
• Buy Nasdaq calls (NDAQ) — 9–12 month play (equity/option mix) to capture elevated issuance and trading volumes from continued AI-capex and M&A; size small (1–3%). Expected payoff: moderate upside from fee expansion and listings; hedge with 30% position-level stop if volumes normalize and volatility collapses.