Samsung, SK Hynix upgrading China plants to make more advanced chips, people say: Reuters

Samsung and SK Hynix are upgrading their China factories to produce NAND chips with more than 200 layers, targeting AI-related demand while navigating US export-license restrictions. Samsung’s Xi’an plant, responsible for about 40% of its global NAND output, is targeting 5,000 wafers monthly in December this year and 40,000 by December 2027 for 286-layer chips; SK Hynix plans to shift Dalian production from 192-layer chips to 240 layers or more. The upgrades face uncertainty over US approvals as YMTC’s China NAND share rose to 38.5% in Q2, while Samsung’s fell to 33.5% and SK Hynix’s to 17%.

Analysis

The key economic distinction is technology refresh versus wafer-capacity expansion: higher layer counts can raise bits per wafer and lower unit costs even if formal wafer capacity is unchanged. If the reported ramps proceed, they could both defend Samsung and SK hynix’s cost position and add NAND supply—so the same upgrade that protects share may pressure industry pricing. That is a potential benefit to storage buyers, but not automatically to AI-related memory economics: NAND demand is not interchangeable with high-bandwidth memory demand.

The asymmetry is regulatory. The companies’ statements that work is being done in consultation with Washington do not establish that the required licenses have been granted. If approvals are narrow, delayed, or not renewed, Chinese sites risk a widening technology gap and stranded conversion spending; YMTC gains share and learning scale. If upgrades are permitted, YMTC faces stronger competition, but the reported production ramp could still weigh on NAND pricing. Watch NAND contract pricing, bits shipped, yields, and license disclosures rather than treating announced layer counts as realized output.

Near term, this is an uncertainty event, not a clean earnings catalyst. Over 1–3 months, license scope and customer qualification are the key checks. Over 6–18 months, the outcome could reshape China-market share and the global cost curve. The Solidigm IPO is a separate potential valuation catalyst for SK hynix; its terms and any proceeds or retained economics are not established here. Intel sold the NAND operations, so do not infer a direct Intel earnings benefit without evidence of continuing financial exposure.

Key Decisions for Investors

• No immediate directional NAND trade: reported upgrades have uncertain licensing and execution, while successful bit-output growth could be negative for sector pricing even as it helps the operators’ cost competitiveness.
• Keep SK hynix on a catalyst watch, not an automatic buy: revisit only after verifiable license scope, customer qualification, and production/yield evidence. A license denial, material delay, or weak NAND pricing would falsify the constructive upgrade case.
• Monitor YMTC’s share and product progression as the competitive downside indicator for incumbent pricing and mix; sustained share gains alongside falling NAND contract prices would argue against paying for the incumbents’ technology-refresh narrative.
• Treat the Solidigm IPO as a separate SK hynix catalyst. Before positioning, verify valuation, offering size, proceeds allocation, and SK hynix’s retained ownership or other economic interest.
• No Intel trade from this report alone: the former NAND operation is now Solidigm under SK hynix, and the article establishes no current Intel exposure or cash-flow linkage.