3 Artificial Intelligence (AI) Stocks to Buy at a Discount

Micron reported revenue of $23.9B this quarter (vs $13.6B last quarter and $8.0B year-ago) and expects $33.5B next quarter; its shares are up ~300% over the past year but trade at ~6.8x forward EPS, with management saying it can only fulfill ~50–66% of orders. Nvidia posted Q4 revenue growth of 73% YoY and is guiding ~77% growth for Q1 while trading at ~21.1x forward earnings (S&P ~20.6x). Microsoft revenue rose 17% YoY and EPS increased ~60% (partly from OpenAI-related gains), leaving the stock at historically low valuations by operating P/E metrics.

Analysis

AI-driven GPU demand is re-rating capital allocation across the data-center supply chain rather than just boosting single names. With lead times and node constraints at foundries and HBM suppliers, marginal dollars from hyperscalers will disproportionately flow to firms able to secure advanced memory and packaging capacity — that amplifies outsized cash conversion for a handful of suppliers for 12–36 months.

Micron’s current earnings strength is a function of structural mismatch between near-term demand and supply-side inertia; that same inertia is the double-edged sword. If customers accelerate restocking and Micron front-loads capex, pricing could normalize within 12–24 months; conversely, continued export frictions or slower fab builds could extend pricing power and force OEMs into higher-BOM designs (raising industry gross margins).

Nvidia sits at the largest nonlinear payoff: its architecture concentration makes it both the prime beneficiary of compute pull-through and the single-point risk if a large buyer substitutes or delays orders. Microsoft is a lower-volatility way to capture AI platform upside but carries an OpenAI/carrying-cost binary that will show up in lumpy mark-to-market gains and quarter-to-quarter valuation moves.

Net-net: this market is pricing a mixture of cyclical restock risk and secular AI adoption; prefer convex, funded long exposure to leaders (option structures or LEAPs) and use short-dated hedges to protect against 1–3 quarter inventory normalizations. Size risk so single-event reversals (earnings/controls) are portfolio-manageable.

Key Decisions for Investors

• NVDA — Buy a funded bullish LEAP spread (long Jan-2027 calls, finance with 1:1 shorter-dated calls) sized 2–3% portfolio. Timeframe 6–18 months. Rationale: captures secular GPU upcycle with defined max loss (premium); target 30–80% upside if demand stays strong; hedge tail-short with 5–10% portfolio cash or short-dated puts to limit gamma on earnings weeks.
• MU — Initiate a core position (3–4% portfolio) expressed via long 18–24 month LEAP calls or buy shares paired with 1y ATM protective puts (50–60% hedge). Timeframe 12–36 months. R/R: asymmetric upside (2x–3x if shortage persists) vs downside capped to 40–60% if memory pricing collapses; reduce size after a 25–35% rally and re-buy protective puts.
• MSFT — Accumulate on any further 8–12% pullback, sized 2% portfolio, and sell 3–6 month covered calls to harvest premium while retaining upside. Timeframe 6–24 months. R/R: modest absolute upside but low volatility hedge for portfolio AI exposure; risk is concentrated mark-to-market from investor sentiment on OpenAI valuation events.
• Pair/Relative trade — Long NVDA / Short SMH (equal notional) for 6–12 months sized 2% net exposure. Rationale: express NVDA-specific outperformance vs broad semiconductor cyclicality; target 30% relative alpha; stop and reassess if broad-capacity signals (ASML/LRCX earnings) indicate industry-wide re-stocking (>2 consecutive quarters of positive guide).