AI受益者与被高估者：内存和韩国科技股的评级/目标价变动

AI beneficiaries vs overvalued: Rating/POchanges among memory and Korea tech Rating Change 24 October 2025 Thesis: time todistinguish AI beneficiaries vs overvaluedAlmost all memory and Korea tech stocksunder our coverage have given double- or EquityAsia PacificTechnology triple-digit returns YTD, primarily driven by the AI growth theme, in our view. That said,the degree of AI impact will vary significantly among various tech names. As such, webelieve it is time to distinguish real AI beneficiaries vs low-growth / overvalued stocks. Simon Woo, CFA>>Research AnalystMerrill Lynch (Seoul)+82 2 3707 0554simon.woo@bofa.com Top picks:Hynix, Nanya, SIMO, LGE and SDSOur top picks for global memory and Korea tech are: SK Hynix (catalyst: No.1 position in HBM; new PO: W700,000); Nanya Tech (legacy DRAM upturn; NT$150); Silicon Motion(SIMO; solid demand for NAND controller IC; US$130); LG Electronics (well-managed UStariffs, resilient earnings, growth opportunities with data center cooling system and autoparts, India subsidiary IPO completion; W130k); and Samsung SDS (AI-based IT services;W240k). These companies’share prices have reached or surpassed prior high level. But,new POs imply 40%+ upside potential on average, despite our conservative target P/E(10-20x) or P/B (2-3x). New upgrade to Buy from Neutral on LG Innotek (50%+ marketshare likely well-retained for Apple’s iPhone 17) is also interesting. Dai Shen>>Research AnalystMerrill Lynch (Hong Kong)dai.shen@bofa.com Matt Shin>>Research AnalystMerrill Lynch (Seoul)matt.shin2@bofa.com Downgrades: Hanmi, Soulbrain, SEMCO, LGD, SDI, DuksanWe downgrade Hanmi Semi (from Buy to Neutral; TCB equipment supplier; new PO W170k); Soulbrain (to U/P; hydrogen fluoride etchant; W270k); SEMCO (to Neutral;MLCC/substrate manufacturer; W250k); LG Display (to Underperform, OLED panels forIT/TV; W13,000; Samsung SDI (to Underperform; EV/ESS batteries; W260k); and Duksan(to Neutral; OLED materials; unchanged W50k) Rich valuations (Hanmi Semi, etc.) or lowAI exposure (Samsung SDI, LGD, Soulbrain) are the key reasons behind our Neutral orU/P ratings. See Exhibit 3-5 for more details. Catalysts: AI growth, US exposure, value-up programHBM demand has been more diversified to include OpenAI, AMD, and others beside NVIDIA. If OpenAI’s suggested long-term advanced memory demand reaches the 900kwpm level (Hynix’s current HBM capacity only 160k), memory cycle will remain strongwith high margins through 2026-30. US domestic demand (AI chip and even Applesupply chain and home appliances) can be stronger than other regions (China, Europe,Middle East) due to Big Tech’s capex hike or replacement cycle (iPhone 17/18, etc.).Value-up program can also be a catalyst for names with net cash (e.g., Samsung SDS). Risks: US, China rare-earth export control, double-booking We don’t see any significant impact of US tariffs on memory and Korean tech names.For instance, LGE and Samsung Electronics have already reported upbeat 3Q preliminaryOP. China’s control over rare-earth exports should also be a risk, but non-Chinacountries’rare earth resources should be sufficient, at least for memory and Korea tech.We also checked for double-booking status (HBM, DRAM, NAND, etc), but we learnedthat it is mostly actual demand as alluded by major OEMs or module makers. >> Employed by a non-US affiliate of BofAS and is not registered/qualified as a research analystunder the FINRA rules.Refer to "Other Important Disclosures" for information on certain BofA Securities entities that takeresponsibility for the information herein in particular jurisdictions.BofA Securities does and seeks to do business with issuers covered in its researchreports. As a result, investors should be aware that the firm may have a conflict ofinterest that could affect the objectivity of this report. Investors should consider thisreport as only a single factor in making their investment decision.Refer to important disclosures on page 38 to 41. Analyst Certification on page 35. PriceObjective Basis/Risk on page 29. Glossary can be found at the end of this report Identifying the real beneficiaries of AI We believe it is time to distinguish real AI beneficiaries vs low-growth / overvaluedstocks. Our bullish view on HBM and memory cycle is based on strongly growing AI chipdemand vs tight capacity. Legacy DRAM price rally (up 100%+ YTD), however, has beendriven by capacity/capex shift (to HBM) instead of demand (not yet meaningfully strongdue to PC/smartphone/TV saturation). Memory chipmakers’2026-27 capex spend willalso grow but the incremental spend (vs 2025 peak) should be for HBM4, node migration(1c DRAM, 300+/- layers NAND) and infrastructure (shell fab, power/utilities, in-housecloud/data centers). Thus, we still see limited capex spend or production volume increasefor conventional/commodity/legacy memory (negative implications for low-endequipment and material suppliers). Separately, we still observe very low impact of AI among d