SK hynix Inc. Q3 FY2026 earnings call

• Strategic Investments: Discontinued Solidigm IPO review due to market volatility; will pursue alternative financing. Returned gains from Kioxia investment via special dividend.
• Shareholder Returns: Announced total shareholder return of ~KRW 100 trillion for FY2026 (KRW 30T+ dividends, KRW 60T+ share buybacks/cancellations). Paid KRW 20,000/share dividend (including KRW 10,000 special dividend).
• AI Market Outlook: Projected global CSP CapEx growth >60% in 2027. Demand outpaces supply through 2030. Investment base is diversifying beyond big tech to sovereign and enterprise AI.
• Supply Chain: Structural supply-demand imbalance persists. Expected 2027 growth: DRAM mid-20s%, NAND high-teens.
• Product Roadmap: HBM4 surpassing HBM3E in volume/revenue in Q4. M15X ramp-up accelerating. 321-layer NAND capacity expanding to >50% domestically.
• Long-Term Agreements (LTAs): Secured >50% of 2027/2028 capacity via LTAs with major CSPs, providing stable visibility and profitability foundation.

Total revenue reached KRW 104.6 trillion, a record high, up 32% sequentially and 328% year-on-year. Operating profit was KRW 83.3 trillion (up 38% sequentially, 632% year-on-year) with an operating margin of 80%. Net income was KRW 39.8 trillion with a net margin of 38%. For DRAM, bit shipments increased approximately 10% sequentially due to HBM4 mass production and strong server DDR5 growth; ASP rose by the high 20s%. For NAND, bit shipments decreased slightly as the company focused on high-value products, but ASP rose by the mid-20s% driven by enterprise SSD demand.

• Revenue & Profit: Expects significantly higher absolute revenue and profit levels next year compared to current year.
• Shipment Growth: Q4 DRAM shipment volume expected to grow mid-20% sequentially. Q4 NAND bit shipments expected to increase mid-single digits sequentially.
• Product Mix: HBM4 expected to become flagship product in Q4, exceeding HBM3E in share.
• Capital Expenditures: Significant increase in CapEx expected in 2027 compared to 2026 for leading-edge expansion and Yongin fab.
• Yield & Production: HBM4 yield exceeds targets; HBM3E yield mature. Focus on maximizing productivity to meet overwhelming demand.

• Exchange Rate: Strong Korean won caused KRW 11.6 trillion foreign exchange loss.
• Investment Valuation: KRW 22.9 trillion loss from valuation of investment assets.
• Supply Constraints: Leading-edge process complexity and facility lead times prevent full satisfaction of customer volume requirements.
• Overproduction Risk: Management acknowledges past cycles but argues current investments are for tech migration/high-value capacity, not just bit output. Capacity ramp-up is staged and flexible based on demand visibility.

Q: Citigroup asked about the sustainability of AI investment momentum despite slow monetization.

A: Management cited two reasons: 1) Hyperscalers report 30-80% cloud revenue growth driven by AI, with order backlogs growing 2-5x, proving ROI. 2) The investor base is structurally diversifying beyond big tech to sovereign and enterprise AI, reducing reliance on single entities' spending decisions. Supply constraints will persist through 2030.

Q: Meritz Securities asked about the basis for the dividend vs. share buyback ratio.

A: Dividends were set sustainably reflecting earnings and peer yields, plus a special dividend for Kioxia gains. Buybacks were chosen because management believes the stock is undervalued relative to cash flow and earnings. The goal is to raise per-share value while maintaining sustainable cash distributions.

Q: Goldman Sachs asked about the impact of LTAs covering >50% of 2027/2028 capacity.

A: This secures medium-to-long-term demand visibility, effectively selling DRAM/HBM volumes through 2029. It provides a stable earnings base. Non-LTA capacity remains available to flexibly respond to market improvements and next-generation product demand.

Q: Bank of America asked about overproduction risks given aggressive fab investments (e.g., Yongin).

A: Management stated that more fabs do not equal more supply immediately due to complex processes and tech migration needs. Investment is staged: shell structures are built early, but clean rooms and equipment move-in are phased based on demand visibility and financial health. They retain flexibility to adjust investment pace if conditions change.

Q: SK Securities asked about 'stack lowering' in HBM (e.g., 12-stack to 8-stack) and its impact.

A: Stack lowering is a customer strategy to maximize cube count under supply constraints, not a reduction in total demand. Increased die capacity per stack offsets lower stack counts. Profitability depends on mix, yield, and cost, not just stack height. HBM4 price increases and yield improvements will drive significant YoY profit growth.