SK Hynix Leads Memory Stocks Higher as AI Demand Drives Record Sales, Profits

SK Hynix reported record third-quarter sales of 104.62 trillion won ($77.92 billion), above the 99.5 trillion won analyst consensus, and operating profit of 83.33 trillion won, topping estimates of 78.28 trillion won. The company said AI-driven memory shortages could extend through 2030 and that long-term deals with 14 customers cover more than half of production capacity over the next two years. Shares rose over 10%; the Roundhill Memory ETF gained around 5%. SK Hynix reiterated plans for 100 trillion won in shareholder returns this year and said its board approved a new 20 trillion won buyback plan.

Analysis

The surprise is most useful as evidence that tight memory supply is translating into pricing power—not as proof that today’s margins or shortage persist to 2030. Long-term customer commitments reduce demand visibility risk, but the unspecified pricing and reset terms matter: fixed-price volume could cap upside if spot prices rise, while market-linked contracts preserve it. Verify contract economics before treating booked capacity as earnings certainty.

Near term, SK hynix’s rally raises the bar for follow-through; the more interesting read-through is whether Micron can convert the sector signal into upward guidance revisions. Do not assume equal benefit: product mix, HBM qualification and customer allocation determine who captures incremental AI spend. Samsung Electronics is a relevant competitive swing factor if it improves HBM yields or adds usable capacity.

Over 1–3 months, watch memory pricing, HBM shipment/qualification commentary and capex plans. Over 6–18 months, aggressive investment can turn today’s scarcity into oversupply, especially if AI infrastructure spending slows or customers optimize memory per accelerator. Buybacks may support demand for shares but are not an earnings floor; large capacity commitments and capital returns also compete for cash. The bullish thesis weakens on falling memory prices, downward guidance revisions, delayed customer ramps, or evidence that new supply is arriving ahead of demand.

Key Decisions for Investors

• Avoid chasing SKHY after the sharp gap. Hold/add only on consolidation if subsequent pricing data and guidance confirm earnings momentum; cut the thesis if memory prices roll over or management lowers its outlook.
• Treat MU as a conditional read-through, not an automatic sympathy buy. Consider a staged long on a pullback or after a positive guidance/earnings revision; verify HBM shipments, customer qualification and realized pricing first. Falsifier: MU fails to raise outlook despite sustained sector pricing strength.
• For a relative-value expression, consider long MU versus a broad semiconductor basket only if MU’s next update confirms improving HBM mix and execution; size for correlation and AI-capex exposure. If differentiation is not confirmed, stay flat rather than extrapolating SK hynix’s results across memory makers.
• Track spot and contract memory prices, competitor capacity/yield updates (especially Samsung Electronics), and major AI-customer spending plans over the next 1–3 months. A supply ramp or AI capex slowdown would invalidate the 6–18 month scarcity thesis.