10 AI Stocks I'm Buying Right Now

The author lists a top-10 AI stock watchlist, highlighting Nvidia (Q4 revenue +73% YoY; Q1 revenue guide +77%) and Broadcom (guides AI-related sales to >$100B by end-2027 from current < $8.4B/quarter) as primary picks. Microsoft cloud/AI revenue rose 39% YoY; Meta posted 24% revenue growth, while Microsoft, Amazon and Meta shares are cited as trading ~35%, >22%, and ~34% below all-time highs respectively, signaling perceived buying opportunities. The note also flags higher-risk/longer-duration AI plays (IonQ quantum, Nebius partnership with Nvidia, SoundHound AI contract wins) as speculative upside exposures.

Analysis

AI’s multi-year capex wave is bifurcating the hardware stack: hyperscalers are moving part of demand toward bespoke ASICs (higher upfront design spend, lower per-unit inference cost), while general-purpose GPU demand remains for training and flexible inference. That creates a clear second-order beneficiary set: foundries and firms that sit on the contract-design/customization path (allocation wins, long-term multi-year revenue slots), and channel/cloud partners who get prioritized hardware access.

Concentration risk is underappreciated — a handful of foundries, a few hyperscalers, and a couple of IP suppliers now determine capacity and pricing. Short-term demand signals (quarterly cloud utilization) can flip price/margin dynamics quickly; regulatory moves (export controls) or a discreet hyperscaler design-win announcement can reallocate billions of dollars of orders within months.

Quantum and niche software players remain convex lottery tickets: protocol-level or algorithmic breakthroughs would re-rate those names long after market attention has moved on, but timelines are multi-year and binary. For portfolio construction this means a two-bucket approach — convex, capped-risk optionality in small positions for long-horizon upside, and concentrated, hedged exposure to the core compute and cloud platforms where execution and allocation risk dominate over the next 6–24 months.

Key Decisions for Investors

• Core asymmetric exposure to compute: buy NVDA via a long-dated call spread (buy NVDA Jan 2027 1x sell Jan 2027 2x at higher strike) sized 2–4% NAV. Rationale: captures multi-year GPU demand with capped premium; set max loss = premium (~100% of position) and target 3–6x payoff if leadership persists. Monitor quarterly capacity comments and TSM allocation notes as triggers to trim.
• Play custom ASIC share-shift: long AVGO stock sized 1–2% NAV and hedge with a small NVDA put position (3–6 month, ~15% OTM) sized to cover 30–50% of notional. Rationale: AVGO benefits if hyperscalers accelerate bespoke chips; hedge protects portfolio if NVDA retains/expands share unexpectedly. Target 12–18 month hold; stop-loss at -20% absolute on AVGO tranche.
• Satellite asymmetric picks (small, capped): allocate 0.5–1% NAV to NBIS (Nebius) and 0.5% to SOUN — buy shares and protective 12-month calls to limit downside (or buy 12-month calls outright if liquidity). Rationale: optionality on privileged hardware/software distribution; cap each position’s downside to ~100% premium and target 5x+ upside if strategic partnerships scale.
• Long-shot convex ticket: allocate <=0.5% NAV to IONQ using a multi-year option (buy 2–3 year calls) or stock+long-dated calls. Rationale: quantum remains a multi-year binary; position size should be treated as lottery-ticket capital with expected time-to-catalyst 3–7 years. Use this as a tail hedge for upside to the AI compute stack; accept total loss as possible outcome.