3 Undervalued AI Stocks to Buy Right Now

Microsoft: Azure revenue +39% and overall revenue +17% YoY; stock is down >30% from its all-time high and near its lowest P/E of the past decade, presented as a buying opportunity. Nvidia: trading at 20.6x forward earnings (in line with the S&P 500) while management now targets $1 trillion of Rubin/Blackwell chip sales by end-2027, implying underpriced multi-year growth. Micron: trades at ~6.1x forward earnings amid a cyclical memory market but forecasts HBM market growth from $35B in 2025 to ~$100B by 2028 (3x), with current supply unable to meet demand.

Analysis

The AI compute wave is bifurcating value capture: hyperscalers and platform hosts that sell bundled compute+services (MSFT-style) will capture recurring annuity-like cash flows and equity upside from model creators, while chip/IP owners (NVDA) extract outsized margin on the hardware stack. A less obvious beneficiary: firms upstream in advanced packaging and substrate supply chains will see multi-quarter lead-times translate into pricing power — meaning constrained HBM supply could persist even as fab output climbs. That widens gross margins for incumbents but also raises execution risk on capital intensity and working capital.

Micron’s demand spike is real but exposed to classic memory cyclicality plus policy and efficiency shocks. Model-efficiency gains (TurboQuant-style) are likely to compress HBM per-inference demand but will lower training costs, which tends to expand total training volumes (a Jevons-style effect) — net demand is ambiguous and timing-sensitive. Export controls or preferential supplier allocation to geopolitical allies could cap revenue growth versus a pure-demand narrative.

Key catalysts and tail risks are staggered across horizons: days–weeks around earnings/guidance updates and GPU bookings; months for Rubin/next-gen chip ramps and HBM capacity additions; 12–24 months for TAM realization or policy-induced market bifurcation. Reversals come from sudden capex announcements flooding supply, a sharp macro slowdown that stalls AI projects, or regulatory action that forces material changes to Microsoft’s partnership/equity model.

Tactically, prefer asymmetric, hedged exposure that isolates secular AI adoption from memory cyclicality and policy risk. Use pairs and options to express convex upside to NVDA-led compute growth while limiting outright exposure to Micron’s inventory cycle; treat MSFT as a defensive anchor to collect platform rents but hedge execution/regulatory kinks with tail protection.

Key Decisions for Investors

• Long NVDA (6–12 month): Buy LEAP call calendar/diagonal to capture next-gen GPU revenue convexity — buy 12–18 month LEAP calls and sell 2–3 month calls to fund. Position size ~3–5% NAV; target +50–100% on a successful ramp, max loss = net premium (~100% of premium).
• Pair trade — Long NVDA / Short INTC (6–12 month): 1:3 notional hedge to express accelerated AI share shift versus Intel; set stop if pair underperforms by 20% and size to cap downside at 4% NAV. Expect NVDA to outperform by 25–75% if hyperscaler spend continues.
• MSFT protected long (12 months): Accumulate on 2-day >5% selloffs; buy shares and simultaneously buy 9–12 month 7–10% OTM puts to limit downside. Target +25–35% in 12 months from multiple re-rating and platform monetization; hedge cost ~2–4% of position value.
• Micron tactical entry via cash-secured puts (6–9 month): Sell puts ~15–20% below current price to collect premium and build a position only at a lower cost basis; size small — equivalent to 1–2% NAV. Reward: capture multi-year HBM upside if cycle holds; risk: full assignment and potential 40–60% drawdown if memory crash occurs.