Stock Market Today, Oct. 28: SK Hynix Surges on Record Q3 Results and AI-Memory Demand

SK Hynix shares surged 13.30% to $164.70 after record third-quarter results, including an operating-profit increase of more than sevenfold and an 80% operating margin, amid rising AI-memory pricing. Management said DRAM and HBM production is effectively sold out through 2029 and that memory demand is expected to outpace supply through the end of the decade. Trading volume was 48.4 million shares, about 113% above its three-month average; memory-chip peers Micron and Sandisk rose 5.86% and 7.66%, respectively.

Analysis

The key read-through is not simply stronger AI demand: constrained high-bandwidth memory can shift bargaining power toward memory suppliers, while raising the cost of AI systems for accelerator makers, server vendors, and cloud providers. If those costs outrun monetization, customers may optimize memory per workload or defer deployments—an eventual check on the demand narrative.

Near term, the sharp repricing and unusually heavy trading make chasing SK hynix unattractive without evidence that estimates still lag the new pricing regime. Its stated forward visibility is a company claim, not proof that realized prices, yields, or margins remain elevated across the cycle. Verify HBM versus conventional DRAM mix, customer commitments, capex, and shipment yields. NAND exposure is not interchangeable with HBM: Sandisk’s sympathy move may be less directly supported by the tightest AI-memory bottleneck.

Over 1–3 months, contract-price updates, quarterly guidance, and capacity/yield disclosures should test whether scarcity is translating into durable earnings. Over 6–18 months, industry capex and new capacity create the reverse risk: supply additions, product transitions, or customer qualification of alternatives can compress pricing before headline demand weakens. The contrarian risk is treating sold-out capacity as guaranteed pricing power; volume visibility does not lock in price or margin.

Key Decisions for Investors

• Avoid chasing SKHY immediately after the outsized gap. Reassess on a pullback or after the next disclosure confirms HBM pricing, mix, and yields; the thesis weakens if guidance or realized memory prices stop improving.
• Watch a conditional relative-value trade: long MU versus SNDK only if disclosures show MU gaining meaningful HBM/DRAM exposure while NAND pricing or demand lags. Size against the risk that broad AI sentiment, rather than product-specific fundamentals, drives both shares.
• Track memory contract prices, supplier capex and yield progress, customer qualification timelines, and inventory. Rising capacity plans without corresponding pricing strength would be an early warning to reduce memory-sector exposure.
• Monitor AI infrastructure buyers for evidence that memory costs are slowing orders or pushing workload-level memory efficiency. Such evidence would challenge the demand-duration thesis even if current supplier backlogs remain strong.