Stock Market Today, March 19: Micron Falls Despite Record Revenue Amid Margin and Capex Concerns

Micron reported record Q2 results with EPS of $12.07 (vs $1.41 a year ago) and sales that nearly tripled, yet shares fell 3.78% to $444.27 on Thursday. Management guided Q3 sales and EPS to rise sequentially by 40% and 59% and raised the dividend by 30%, but warned capex will exceed $25 billion in 2026 to expand DRAM/NAND capacity, which pressured margins and investor sentiment. Trading volume was 73.7M shares, about 102% above the three-month average, and the company is valued near $500B with an analyst projecting potential upside to $1T if AI-driven demand materializes.

Analysis

The market is wrestling with a classic capital-intensity paradox: aggressive capacity buildouts can both secure long-run share in an AI-driven memory market and create a short-to-medium-term overhang as wafer starts and NAND/DRAM supply ramp over 12–24 months. That dynamic disproportionately helps upstream equipment and materials suppliers (tools, photoresists, specialty gases) in the near term while creating margin risk for memory OEMs if demand growth or pricing elasticity underperforms build schedules.

AI inference’s migration from monolithic training clusters to distributed, low-latency deployments changes the product mix: demand shifts toward higher-density, lower-power DRAM, HBM variants and fast, low-latency NAND used for model sharding and retrieval. This favors players with differentiated node roadmaps and HBM partnerships but accelerates commoditization of mainstream DDR if multiple suppliers come online simultaneously — a two- to three-year timing story for ASP normalization.

Key catalysts to watch are (1) early HBM/DDR shipping cadence into hyperscaler inference racks (near-term 3–9 months), (2) equipment order books and lead indicators from wafer-equipment suppliers (6–18 months), and (3) signs of model-architecture changes (sparsity, quantization, offloading to memory-centric accelerators) that could materially change $/inference intensity over years. Tail risks include export-control shocks, faster-than-expected competitor node transitions that drop $/GB, or hyperscalers consolidating supplier relationships; any of these can flip the thesis in one earnings cycle.

Key Decisions for Investors

• Tactical long via defined-cost options: Buy a 12–24 month MU call spread (LEAP buy / higher-strike sell) to express multi-year AI-inference upside while capping premium paid. Position size 3–5% of equity risk; target 2.5–4x payoff if memory market re-rates, max loss = premium.
• Hedge short-term execution risk: Buy a 3–6 month MU put spread to protect existing exposure against ASP compression from a supply ramp. Use as 30–40% hedge of directional long exposure; defined loss limits downside to premium while providing asymmetric protection.
• Equipment play: Long AMAT and LRCX (equal-weighted) on a 6–18 month horizon to capture capex flow-through to tools and consumables. Risk: if capex is deferred, expect 15–25% downside in 3–6 months; reward: 30–60% upside if orders convert and backlog prints.
• Pair trade to neutralize industry cyclicality: Go long MU and short Samsung (equal notional) for 6–12 months to express Micron-specific execution and product-mix wins vs a vertically integrated competitor. This reduces macro/ASP tail risk; skew position toward MU if evidence of inventory discipline and early HBM design wins appear.