J.P. Morgan: SK Hynix's Headwinds Have Passed, Maintains Overweight

Overview

In response to SK Hynix's recent sharp stock decline, J.P. Morgan released a research report addressing three core market concerns one by one: HBM pricing and content levels, uncertainty regarding shareholder returns, and massive capital expenditures. The institution believes current pessimism has been fully priced in. Normalization of HBM content is a necessary step to address chip shortages, and reports of a "50% discount" are inaccurate. Management has clarified that the shareholder return policy will be updated by the end of Q3 2026 (late September). Based on confidence in the memory industry's upward cycle and the company's execution in AI solutions, the report maintains an "Overweight" rating and a 2.75 million KRW target price.

Core views

On HBM Pricing and Content Level Adjustments: The report explicitly states that media reports claiming "HBM4 will be discounted by 50% compared to competitors in 2027" are inaccurate. While investor expectations for HBM ASP doubling year-over-year may be too high, the institution's relatively conservative assumption still includes year-over-year growth below 40% starting from 2026. There are three logical reasons behind this pricing strategy: First, memory suppliers prioritizing supply of products like DDR5/LPDDR5/NAND, which have significantly higher margins than HBM, aligns with their own interests. Second, as NVIDIA's largest partner, SK Hynix can negotiate from a multi-year procurement perspective. Third, HBM is re-priced annually; if the company prioritizes Long-Term Agreement (LTA) shipment volumes for the next 3-5 years, short-term contract prices are not the primary priority. Furthermore, normalization of HBM content volume is viewed as a necessary step to handle excessive chip shortages rather than a signal of deteriorating demand. On the 54 Trillion KRW Infrastructure Capital Expenditure: The large investment causing market concern is actually a necessary preparation for the company to achieve its goal of one million wafers of capacity by 2030. Specifically, 35.2 trillion KRW is allocated for the Yongin Y2 cluster DRAM factory (planned to break ground in July 2027, mainly serving capacity expansion post-2031), and 19.1 trillion KRW is for the Cheongju M17 NAND facility (planned to break ground in February 2026, with the first cleanroom targeted for completion by the end of 2028). This investment pace is consistent with the company's previously disclosed plan to "have new NAND factories ready before 2029," aiming to meet unprecedented demand driven by AI, rather than blind expansion. On Shareholder Returns and Solidigm IPO: Management has advanced the disclosure timeline for the new shareholder return plan from the originally scheduled "by year-end" to "by the end of Q3 2026," eliminating market concerns that the company might delay returns while seeking a subsidiary IPO. Given the expectation of generating over 80 trillion KRW in cumulative free cash flow over the next three years, a gradual shareholder return policy is likely to be well-received by the market. Regarding the Solidigm US IPO, the report takes a cautious stance, arguing that given the NAND supercycle, gross margins exceeding 70%, and ample parent company cash, the strategic financing value brought by the IPO is limited. It may also conflict with South Korean Commercial Law rules on double listing. Its core purpose is more about expanding the investor base than solving funding gaps. Valuation and Investment Logic: The core basis for maintaining the Overweight rating is that SK Hynix's execution in AI solutions has significantly enhanced profitability, and it is currently in the early stages of a memory upcycle. The target price of 2.75 million KRW is based on a 7x P/E ratio applied to the average expected EPS for 2026-2027. This multiple reflects the institution's judgment of a stronger and longer memory upcycle, as well as the sustainability of the company's profitability over the next three years. The high Beta nature of pure storage business means the company will gain greater elasticity during traditional fundamental improvements.

Analysis framework

The report adopts an "event-driven + expectation gap correction" analytical framework. It first identifies three specific market rumors leading to the short-term stock plunge (HBM discount, delayed returns, excessive CAPEX), then falsifies or clarifies each using industry logic and company guidance. In terms of valuation, rather than relying solely on current performance, it uses "memory cycle length" and "structural profit improvement driven by AI" as core variables for granting valuation premiums, anchoring mid-to-long-term value through the P/E multiple method. Simultaneously, it assesses the feasibility of shareholder returns by analyzing free cash flow generation capabilities, directly linking financial indicators with market sentiment recovery.

Methodology notes

• Valuation MethodPE/PEG valuation

  P/E valuation based on forward average EPS

  The report derives the target price by multiplying the average expected EPS for 2026-2027 by a 7x P/E. This method smooths out the large annual profit fluctuations characteristic of the semiconductor industry, better reflecting average profitability across cycles, making it suitable for valuing anchor points in strongly cyclical industries.

• Industry/Industrial Analysis FrameworkSupply and Demand Framework

  Analysis matching supply-side capacity planning with demand

  When analyzing massive capital expenditures, the investment amount is not viewed in isolation but matched against the 2030 one-million-wafer capacity goal and the timing of the AI demand explosion. This reflects the characteristic of the memory industry where "supply determines the cycle," meaning current CAPEX is intended to fill future supply-demand gaps, rather than signaling current oversupply.

• Company Fundamentals and Financial FrameworkFree cash flow analysis

  Linking FCF generation capability with shareholder return potential

  The report emphasizes that cumulative FCF exceeding 80 trillion KRW over the next three years is the foundation supporting shareholder returns. For heavy-asset semiconductor companies, net profit is often distorted by depreciation and amortization, whereas FCF is the core indicator for measuring true dividend-paying capacity and flexibility in capital allocation.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• SK Hynix (000660.KS)

  Core Beneficiary: Strong execution in AI memory solutions, early stage of memory upcycle, high profit elasticity

  Strengths

  Leading HBM technology and tied to NVIDIA; NAND business entering a supercycle; abundant cash flow supports shareholder returns

  Weaknesses

  High Beta nature of pure storage business leads to volatile stock prices; low transparency in HBM pricing easily triggers market misreading

  Comparison

  Compared to Japanese and Korean peers, profit generation capability over the past 12 months (including sale of Kioxia stake) has been stronger, and market expects greater intensity in returns

  Risks

  DRAM prices and margins fall short of expectations; terminal demand or inventory destocking slower than expected

Key data

• Target Price2,750,000 KRWBased on 7x P/E of average 2026-2027E EPS, corresponding to June 2027
• Infrastructure Capex Plan54 Trillion KRWIncludes 35.2 Trillion KRW for Yongin Y2 DRAM plant and 19.1 Trillion KRW for Cheongju M17 NAND plant
• Projected Cumulative Free Cash Flow (Next 3 Years)>80 Trillion KRWFinancial basis supporting the gradual shareholder return policy
• Shareholder Return Plan Update TimeEnd of Q3 2026Advanced from previous guidance of "by year-end," eliminating market uncertainty
• HBM ASP Growth Assumption<40% YoY (2026E)Conservative institutional assumption, lower than the market-expected 100% growth, but refuting the 50% discount theory

Impact & implications

The report believes the worst is past for SK Hynix. As several key catalysts around late September materialize (shareholder return update, determination of HBM contract prices, progress on Solidigm listing), stock sentiment is expected to gradually recover. If HBM ASP ultimately exceeds the institution's conservative expectations, it will become an upside risk to earnings forecasts. For investors, the current pullback offers an allocation opportunity based on long-cycle logic, but close attention must be paid to downstream demand validation and inventory destocking conditions.

Risks

• DRAM prices and margins lower than expected
• Weak terminal demand or inventory destocking slower than expected
• Results of HBM pricing negotiations significantly lower than conservative assumptions
• Solidigm IPO process triggers uncertainty regarding equity structure or regulation

What to watch

• Specific updates to the shareholder return plan before the end of September 2026
• Latest confirmation of HBM contract prices before the end of September 2026
• Progress on the US subsidiary Solidigm's listing plan within the next month
• Actual groundbreaking and construction progress of Yongin Y2 and Cheongju M17 plants