These Growth Stocks Surged 300% and 1,400% While Everyone Was Watching Nvidia, and They Are Still Cheap. Here's Why.

Micron and Sandisk have surged roughly 300% and 1,400% respectively over the past year, yet trade at modest forward P/Es of ~3.7x (Micron) and ~8x (Sandisk) for fiscal 2027. AI-driven demand for HBM (which can consume up to 3x the wafer capacity of ordinary DRAM) and rising SSD/NAND needs have tightened supply and driven prices and margins higher; Micron has secured a five-year HBM contract and DRAM deals are shifting toward 3–5 year terms. Key risks remain the memory market's historical cyclicality and potential demand-reducing compression algorithms (e.g., Alphabet's TurboQuant), while high-bandwidth flash (HBF) for NAND is early-stage. Analyst view favors Micron as the better buy because it appears closer to shedding cyclical dynamics.

Analysis

The market is beginning to price a regime shift where HBM transitions from a niche premium to a contracted, multi-year revenue stream — that’s the structural difference that separates an oligopolist with pricing power (Micron) from a commoditized NAND pure‑play (Sandisk). If HBM contracts standardize at 3–5 years with minimums, DRAM suppliers can convert spot-driven boom profits into annuity‑like FCF; model a 20–30% uplift in sustainable gross margin for the HBM supplier cohort over 24–36 months versus a one‑time cyclical peak for NAND.

Second‑order beneficiaries and constraints matter: foundry/wafer capacity owners and back‑end test/assembly vendors will see utilization-led margin expansion before NAND makers do, potentially increasing capex lead times by 6–12 months. Conversely, cloud/AI hyperscalers face higher unit compute costs that create real demand elasticity — meaningful adoption of compression/sparsity or architectural shifts (systolic arrays, in-memory compute) within 12–24 months is a plausible dampener on memory intensity per model.

Tail risks are concentrated and asymmetric: a sustained advance in model compression or a faster-than-expected ramp of HBF would flatten current pricing power and could erase >50% of excess margin in 12–18 months. The prudent view is conditional: overweight exposures where contract duration and counterparty credit limit downside, and underweight pure NAND here unless HBF adoption shows clear multi‑year demand pull-through.

Key Decisions for Investors

• Long MU (equity) — 18–36 month horizon, initial sizing 2–4% of portfolio. Enter on ≤10% pullback or on announcement of a second multi‑year HBM contract. R/R: target 60–100% upside if HBM contracts become industry standard; downside 40–50% in a cycle reversion. Use a 25% trailing stop or hedge with 1–year puts at 20% OTM if downside risk materializes.
• Pair trade: Long MU / Short SNDK (equal dollar) — 6–12 month horizon to capture relative rerating of HBM vs NAND. Rationale: duration of cash flows and pricing power favors MU; target 30–50% relative spread compression. Risk: both rally on broad AI capex; cap losses by stop‑loss of 20% on the short leg or reduce size to 0.5x if market rallies.
• Options collar on MU: buy Jan‑2028 LEAPS (or nearest 12–18 month calls) sized 1.5x intended equity exposure and finance by selling 3–6 month calls. Objective is to retain upside convexity to multi‑year HBM adoption while reducing net premium. Max loss is premium paid; break‑even moves out materially only if HBM adoption accelerates.
• Tactical hedge for AI concentration: buy 3–6 month puts on NVDA sized at 0.5x MU exposure as tail protection against rapid GPU demand normalization. Cost is limited premium; payoff protects portfolio if compression/architecture shifts quickly reduce memory intensity.