10 AI Stocks I'd Buy Right Now

A 30-day sell-off in AI names is presented as a buying opportunity, with the author spotlighting 10 companies across chips, data‑center infrastructure and cloud platforms. Key mentions include Alphabet (near-$4 trillion market cap), Nvidia (Blackwell GPUs, ~22x projected fiscal 2028 earnings), Nebius (≈$20 billion in combined contracts), CoreWeave (tens of billions in committed revenue) and IREN (a reported $9.7 billion contract with Microsoft); the thesis rests on accelerating hyperscaler capex, expanding contracted backlogs and multiyear demand for AI infrastructure. The piece recommends long-term accumulation of quality AI infrastructure and platform exposures amid fear-driven volatility.

Analysis

Market structure: Winners are hardware and infra operators (NVDA, ASML, AMD, APLD, CRWV, GOOGL) as hyperscalers push 20–30%+ capex growth y/y into GPUs, data centers and leading-edge nodes; losers are single-product app plays and legacy CPU-only vendors without accelerator roadmaps. Pricing power consolidates at chipmakers and lithography (ASML) where delivery lead times create multi-quarter bottlenecks and allow >10% pricing premia on constrained SKUs.

Risk assessment: Tail risks include export controls/technology bans, a demand pullback if model costs fall or LLM utility disappoints, and grid/power permitting constraints for data centers; probability medium but impact high—plan for a 20–40% drawdown scenario over 6–12 months. Short-term (days–weeks) volatility will be event-driven (earnings, Blackwell launch, ASML shipments); medium-term (3–12 months) depends on contracted backlog recognition and hyperscaler contract renewals.

Trade implications: Favor concentrated exposure to infra winners via 6–18 month vehicles: NVDA and ASML for core longs, AMD as a value alternative, and APLD/CRWV for landlord cashflows. Consider relative-value trades: long ASML vs short high-beta AI app names (SOUN) and use option structures (buy LEAPS, sell nearer-term OTM call spreads) to capture elevated IV while limiting downside; scale into positions over 3 months and trim on 40–60% rallies.

Contrarian angles: The market underprices durable cash-flow landlords (APLD/CRWV) and overprices speculative application names; risk of GPU oversupply exists if foundry capacity expands aggressively—if inventory builds >25% industry-wide within 12–24 months, expect margin compression. Historical analogue: 2017–19 semiconductor cycles show infrastructure winners recover faster than app winners; monitor order backlogs and hyperscaler disclosures for early signals.

Key Decisions for Investors

• Establish a 2–3% portfolio long in NVDA via staggered buys over 0–3 months or buy 12–18 month LEAPS (e.g., Jan 2026 calls) sized to risk 2% of portfolio; set a stop-loss to reduce position by 50% if NVDA declines >30% from entry or if company guidance cuts GPU backlog >20%.
• Allocate 1.5–2% to ASML (stock or 9–12 month call spread) to own structural lithography scarcity; increase to 3% if ASML confirms EUV shipment acceleration in next 90 days or backlog growth >15% q/q.
• Take a 1–1.5% position in data-center landlords APLD or CRWV (preferred for lease-backed revenue) to capture contracted cash flows; exit/trim if weighted-average lease durations shorten below 3 years or occupancy falls >10%.
• Implement pair trade: go long ASML (1%) or NVDA (1.5%) and short SOUN (1%) or other small-cap AI application names with no profitable revenue; use options to limit downside and monitor IV—sell OTM call spreads against core longs if IV > 40% to harvest premium.