Is Micron a Buy as Revenue Explodes Higher?

Micron reported fiscal Q2 revenue of $23.86B (up from $8.05B year-over-year) with DRAM revenue of $18.8B and gross margin of 74.4% (vs 36.8% a year ago); adjusted EPS was $12.20 vs $1.56 a year ago and ahead of the $9.31 analyst expectation. Management raised capex to $25B for the fiscal year and guided fiscal Q3 revenue of $32.75–$34.25B, gross margin ≈81%, and adjusted EPS $18.75–$19.55 (consensus was $12.05 on $24.3B), citing sustained AI/HBM-driven memory tightness and long-term capacity constraints. The results and aggressive guidance imply a durable secular tailwind from AI infrastructure, though cyclicality and a recent ~350% share run-up introduce execution and valuation risk despite the attractive forward P/E.

Analysis

Micron is operating with a multi-year structural story (HBM-heavy AI infrastructure) but the real edge is the interaction between wafer-intensity of HBM and 12–24 month capacity lead times. That means price power can persist even as unit growth slows because incremental supply cannot be brought online quickly — a classic oligopoly margin shock that favors incumbents with committed capex and supply-booking relationships with hyperscalers.

Second-order winners are not just DRAM peers but the equipment and packaging chain: high wafer-demand for HBM amplifies spend at AMAT/LRCX/KLA and accelerates bump/EMIB/silos at advancedOSATs, squeezing commodity NAND suppliers who lack HBM IP. On the demand side, higher memory line-item costs create architecture risk at hyperscalers — expect design moves (model sparsity, parameter sharding, larger on-chip caches) that can either prolong HBM intensity or blunt it; the direction will be decided in the next 6–18 months as production models and cost-per-inference metrics stabilize.

Key risks are timing and policy: a hyperscaler pause or inventory digestion in the next 3–9 months would punish multiple-of-cash valuations quickly, while capacity additions from Samsung/SK Hynix or aggressive Chinese expansion 12–24 months out could normalize pricing. Monitor three leading indicators daily/weekly: (1) cloud capex cadence commentary from AWS/GOOG/MSFT, (2) DRAM spot vs contract spreads, and (3) new tool orders at equipment suppliers — those will resolve whether this is a structural re-rating or a cyclical overshoot.

Key Decisions for Investors

• Core long (9–18 months): Buy MU via a calendar/vertical call spread to capture continued AI memory intensity while limiting premium decay. Example: buy MU 12–15 month calls and sell shorter-dated calls 3–6 months out to finance ~50–70% of cost. Risk/reward: max loss = premium paid (~100% of cost), target 2–3x if HBM pricing/pull-through holds.
• Pair trade (3–9 months): Long MU / Short INTC equal-dollar exposure. Rationale: levered upside to memory price-power vs Intel’s slower server recovery and capital intensity. Risk: broad AI capex reversal would hurt both; size so that pair delta ~0 to limit market beta.
• Hedge (6–12 months): Buy a tail hedge on NVDA (out-of-the-money puts) sized to cover 10–20% of MU position value. Rationale: an abrupt AI spending shock collapses GPU and memory demand simultaneously; puts cap portfolio drawdown at defined cost. Cost should be <1–2% annualized of portfolio allocation.
• Tactical income (0–3 months): Sell near-term covered calls against existing MU exposure into earnings/guidance windows to harvest premium and lower basis. Be prepared to re-establish if assigned; set buyback triggers at 10–15% below strike.