Research
研报Overweight4月22日 · Morgan Stanley

SK hynix: 1Q26 – Super Memory Cycle Reaffirmed

SK hynix: The AI-Driven Super Memory Cycle Is Intact and Underpriced

Core Thesis

SK hynix is in a structural, AI-driven "Super Memory Cycle." Its leadership in HBM technology, enhanced visibility from long-term supply agreements (LTAs), and tightening industry-wide inventory are driving record profitability and pricing power. The current valuation, at ~2.7x 2026e P/E, prices in a cyclical downturn, materially undervaluing the durability of AI-driven demand and the company's improved earnings stability.

What the Market is Mispricing

The prevailing valuation reflects an expectation of a traditional memory industry downturn. This overlooks three structural shifts: 1) HBM demand, particularly for HBM3e/4, is driven by persistent AI server build-outs, not cyclical PC/phone upgrades; 2) LTAs with major AI customers lock in multi-year demand and profit streams, dampening cyclical volatility; 3) Industry inventories across key segments have normalized, making memory a critical datacenter bottleneck and granting suppliers sustained pricing power.

Evidence Chain

Profitability has reached a record zenith, with pricing strength extending beyond HBM. The 1Q26 operating profit of W37.6 trillion (a 72% margin), up 96% QoQ and 405% YoY, was driven by broad-based price increases. Notably, HBM's share of DRAM revenue fell to ~20%, indicating non-HBM DRAM pricing surged, a sign of industry-wide tightness. This demonstrates that the cycle's benefits are not confined to the premium HBM segment but are lifting the entire product portfolio.

The HBM technology roadmap is on track, aiming to cement long-term leadership. HBM4 development is proceeding as scheduled. Management's explicit target is to sustain market share leadership in HBM3e and HBM4 despite growing competition. This execution provides a clear path to maintaining its premium position in the highest-growth, highest-margin segment of the memory market for years to come.

Long-Term Agreements (LTAs) have fundamentally improved business visibility. AI customers, seeking to secure critical HBM supply for their own roadmaps, are entering into multi-year LTAs. For SK hynix, this translates into predictable, long-term demand and profit flows. This contractual shift reduces earnings volatility and provides a firmer foundation for capital allocation decisions, moving the business model toward greater stability.

Persistent inventory drawdowns across the ecosystem underpin continued pricing power. Inventories continue to decline in servers, PCs, mobile devices, and NAND. This broad-based normalization, after a prolonged digestion phase, positions memory as a key bottleneck in datacenter expansion. The resulting supply-demand tension supports ongoing firm pricing, extending the cycle's duration.

Key Risks and Divergences

  • End demand, particularly outside AI, weakens more than expected.
  • Intensifying competition in segments like DDR5 could trigger supply-side over-investment, undermining industry discipline.
  • Inventory levels at cloud service providers and Chinese smartphone customers remain elevated, posing a potential headwind.

Valuation and Investment Implication

At 2.7x 2026e P/E and 1.4x 2027e P/E, the valuation is near historical lows and embeds a severe cyclical downturn. This fails to account for the structural support from AI demand, earnings stability afforded by LTAs, and demonstrated pricing power. The stock offers compelling value as the earnings recovery narrative continues to unfold. We maintain an Overweight rating.

Related (同 ticker)