The 2 AI Stocks Quietly Dominating While Everyone Watches Nvidia

Key figure: Motley Fool's Stock Advisor reports a 930% total average return as of Mar 16, 2026 (example outcomes: $1,000 into Netflix on Dec 17, 2004 → $514,000; $1,000 into Nvidia on Apr 15, 2005 → $1,105,029). The article/promotional video (morning prices used Mar 11, 2026; published Mar 12, 2026) highlights two under-the-radar AI stocks that the author says have posted 'tremendous results' and could be significant winners in the AI era. Disclosure: the author/Motley Fool hold positions in Digital Realty Trust, Equinix, and Nvidia and the author may receive affiliate compensation for promoting Motley Fool services.

Analysis

The market is front-running AI winners (highly convex semiconductor exposure) while leaving nuanced parts of the stack mispriced — notably how real estate and interconnection economics re-route value. Hyperscalers will increasingly demand contiguous, power-dense “GPU pod” footprints and direct fiber builds; that structural demand favors bulk-leased, build-to-suit landlords that can scale power and cooling quickly, and penalizes networks that monetize port-to-port connectivity if hyperscalers internalize cross-connects. Expect a multi-quarter lead time between GPU availability and visible leasing revenue — supply constraints can keep landlords lightly occupied even as demand data points look strong.

Tail risks are concentrated: model-level compute intensity can fall via better sparsity/quantization, regulatory limits on large-model deployments, or a rapid supply surge from a new competitor that drives spot GPU pricing down. Near-term price action will be governed by quarterly capex guidance from hyperscalers and NVDA supply commentary (days–weeks), while meaningful re-rating of REITs will take 6–24 months as leases turn. A regime flip (oversupply or a faster-than-expected vertical integration by hyperscalers) would quickly reverse the landlord winners and compress REIT cap rates.

The consensus misses asymmetry in contract types — longer-duration, fixed-dollar leases with built-to-suit clauses create option-like upside for landlords that can scale power, whereas interconnection-heavy players are more exposed to volume declines. That argues for capital-efficient, optionality-focused exposures rather than naked long semis. Position sizing should prioritize convexity (options, pairs) and event triggers around hyperscaler earnings and NVDA supply cadence rather than simple buy-and-hold equities.

Key Decisions for Investors

• Pair trade (6–12 months): Long Digital Realty (DLR) 1.5–2.0% net fund exposure / Short Equinix (EQIX) 1.5–2.0% net exposure. Rationale: play landlord bulk-footprint > interconnection. Target: DLR outperforms EQIX by 400–800 bps total return. Stop: tighten or flip if spread narrows by >200 bps within 3 months or if hyperscaler capex guides decline >10% yoy.
• Defined-risk NVDA directional (9 months): Buy a call spread — buy ATM 9-month calls, sell ~30% OTM 9-month calls sized to finance ~60–75% of premium. Position size 0.5–1.0% fund. Reward: asymmetric upside if NVDA continues to re-rate; max loss = upfront premium. Exit: take profits at 50–100% return or roll if NVDA/compute demand signals remain intact.
• Optionality/convexity on Intel (12 months): Small, opportunistic long-dated call position (0.5% fund) using 12-month LEAPs ~25–35% OTM. Rationale: cheap way to capture upside if Intel cracks sustainable AI accelerator share or secures hyperscaler design wins. Risk: total premium loss; allocate small size only.
• Event trigger rule: Monitor next two hyperscaler earnings for capex guidance. If combined guidance rises >15% yoy, add incremental 0.5–1.0% gross to DLR (or roll NVDA call spread wider). Conversely, if guidance falls >10% yoy, reduce gross AI exposures by 25% and shift to cash/hedges.