Exclusive-Samsung, SK Hynix upgrading China plants to make more advanced chips, people say

Samsung and SK Hynix are upgrading their China factories to produce NAND chips with more than 200 layers, aiming to meet AI demand and counter YMTC, while navigating US export controls. Samsung targets 5,000 wafers per month at its Xi’an plant in December this year and 40,000 by December 2027; the site accounts for about 40% of its global NAND output. YMTC’s China NAND share reached 38.5% in the second quarter, versus Samsung’s 33.5% and SK Hynix’s 17%, as US licensing constraints complicate upgrades and capacity decisions.

Analysis

The key economic distinction is wafer capacity versus bits per wafer: moving to denser NAND can improve unit costs and defend customer qualification without necessarily adding wafer starts. That makes a blanket “more China capacity = imminent glut” read too simplistic. But successful upgrades could still increase available bits and pressure NAND pricing if AI/server and consumer demand fail to absorb them.

For Samsung, China execution matters unusually much because Xi’an is a material share of its global NAND output. For SK hynix, Dalian is a concentrated operational and technology risk for the former Intel NAND business now called Solidigm; do not treat that as an Intel operating asset or infer Intel’s economic exposure without ownership and IPO details. YMTC’s local-market gains raise the cost of under-investing: restrictions may protect foreign suppliers elsewhere while ceding Chinese customers and scale learning to a domestic rival.

Near term (days): license ambiguity is the headline risk, not the announced process transition. Over 1–3 months, watch for explicit BIS license terms, capex disclosures and customer qualification; management claims of compliance do not establish approval scope. Over 6–18 months, successful transitions could narrow YMTC’s cost advantage, while delays risk share loss and stranded older-node tools. Contrarian point: the upgrade may be a defensive node migration rather than a capacity build, so supply fears could be overstated. Thesis weakens if approvals are denied, transitions slip, or NAND pricing weakens despite AI demand.

Key Decisions for Investors

• Keep SKHY neutral/watch rather than shorting on this report alone. Reassess on confirmed Dalian license scope and transition timing; the risk/reward is asymmetric if authorization fails, but a successful upgrade could reduce competitive and cost disadvantage.
• Track NAND contract pricing, supplier bit output and customer qualification over the next 1–3 months. Treat rising bit supply alongside flat pricing as a bearish signal for NAND-exposed producers; resilient pricing through ramp activity would argue demand is absorbing the density gains.
• Do not infer a direct Intel catalyst from Solidigm’s prospective IPO. Verify Intel’s current ownership, any retained economics and transaction timing before positioning INTC; absent that information, this is not a standalone Intel trade.
• Use denial or materially narrower-than-expected BIS approvals, or a disclosed transition delay, as the principal downside alert for SKHY; evidence of timely authorization and stable NAND pricing would falsify the near-term bearish case.