MU vs. SKHY: Which Memory Stock Should You Buy in the AI Selloff?

Micron is presented as the stronger risk-adjusted buy versus SK Hynix after five-session declines of about 10% and 13%, respectively. Micron forecasts fiscal Q1 2027 revenue of $61.5 billion (±$1.5 billion) and adjusted EPS of $38.15 (±$1), with most of its 2027 HBM supply contracted at higher prices and $32 billion in customer financial commitments. SK Hynix reported Q2 2026 revenue up 257% year over year to 79.32 trillion won and operating profit up 557% to 60.54 trillion won, but the article flags its heavy planned investment and exposure to pricing normalization; it also cites CXMT’s claimed 30–50% DRAM density improvement, noting that mass production at scale remains unproven.

Analysis

The key distinction is not simply HBM leadership versus diversification; it is how much of each company’s future output is insulated from a spot-price reset. Customer commitments can dampen near-term earnings volatility for Micron (MU), but they do not eliminate utilization risk if AI customers defer deployments, and the article does not establish how binding or price-protective those agreements are. Verify cancellation terms, deposits, and contracted volumes before assigning them full downside value.

A relative-value long MU / short SK hynix (SKHY) is more defensible than treating either as a clean directional buy: both remain exposed to the same hyperscaler capex cycle, while heavy capacity additions make the downside asymmetric if supply arrives ahead of demand. The short leg is not a judgment that SK hynix is technologically impaired; a successful HBM ramp or stronger pricing could squeeze it sharply. NVIDIA (NVDA) is a demand-side watchpoint: changes in accelerator shipments, memory content, or deployment economics can transmit quickly to both suppliers.

Near term, positioning and inflation-driven discount-rate sensitivity may dominate fundamentals. Over 1–3 months, monitor supplier pricing, customer capex commentary, and realized HBM mix; over 6–18 months, new capacity utilization and CXMT’s demonstrable volume yields matter more than architecture claims. The article’s growth outlook is unusually aggressive; independently verify reported figures and guidance before sizing. The bullish thesis fails if MU’s contracted HBM economics weaken, customer commitments do not translate into shipments, or pricing/utilization deteriorates.

Key Decisions for Investors

• Prefer a staged, beta-adjusted MU-over-SKHY relative position rather than an outright sector chase. Enter only after confirming current prices, liquidity, and the spread’s recent volatility; avoid an unhedged short in SKHY given HBM-ramp and squeeze risk.
• Treat MU’s customer agreements as a potential volatility buffer, not guaranteed earnings protection. Before increasing exposure, verify contract duration, cancellation provisions, deposit terms, and the share of expected revenue that is genuinely committed.
• Use the next 1–3 months of company updates to track HBM pricing and shipment conversion, NAND profitability, and hyperscaler capital-spending plans. Reduce the relative position if MU’s guidance or realized contracted volumes disappoint, or if SKHY demonstrates stronger margins and utilization than expected.
• Keep CXMT on a structural watchlist, not as a near-term short thesis: require evidence of repeatable yields and commercial-scale shipments before treating its density claims as a material supply shock.