Samsung and SK Hynix Are Upgrading NAND in China. Why Does Samsung Say It Is Not Expanding?

What the companies confirmed—and what Reuters sources reported

An October 21 Reuters investigation says Samsung Electronics is adapting its Xi'an NAND plant in China for 286-layer V9 products, while SK hynix is looking at 240-layer-plus chips at its Dalian facility. Sources described the existing processes as roughly 200 layers at Xi'an and 192 at Dalian.

Both companies confirmed upgrade work and consultation with US authorities. Neither confirmed the reported detailed layer counts or timelines as a public production schedule. Xi'an's reported internal codename and Dalian's process roadmap derive from Reuters interviews with people familiar with the plans.

Reuters was unable to establish exactly which required US-origin equipment had received export clearance. A proposed manufacturing transition is not equivalent to blanket equipment authorization or a completed commercial ramp.

YMTC has passed Samsung in China—not globally

According to Omdia data cited by Reuters, YMTC's China NAND share reached 38.5% in Q2 2026, compared with 28% for the full year 2025. Samsung was at 33.5% versus 36%, while SK hynix was at 17% versus 20% on the same respective period definitions.

The distinction is geographic and temporal. These are market-share estimates for NAND sold in China, not a reversal of the global NAND rankings. The 2025 annual average and a single 2026 quarter also do not provide a same-period growth rate.

Reuters sources said YMTC's leading volumes were still in 160- and 232-layer devices, with limited output of 294-layer chips and plans for greater than 300 layers. Higher layer counts matter for bit density, but they are not a complete measure of yield, device costs or end-customer performance.

Technology already shipped elsewhere may lag in a China fab

Samsung formally began ninth-generation V-NAND mass production in April 2024, highlighting higher bit density and advanced channel-hole etching. SK hynix announced mass production of 321-layer NAND in November of the same year.

Yet Reuters characterizes their existing Chinese plant processes as operating at lower layers. A company's technology capability is not identical to the process generation currently deployed at a specific overseas fab.

Moving to a new layer generation involves etching and deposition tools, process tuning, defect analysis, yield improvement and customer qualification. If US-made equipment is subject to export permits, readiness at a domestic fab does not remove that additional condition.

US controls distinguish continuity of operation from technology upgrades

On September 2, 2025, the US Bureau of Industry and Security published a rule removing certain Samsung, SK hynix and Intel-linked facilities in China from the Validated End-User program. The rule became effective on December 31, 2025, replacing the former broad authorization with more specific licensing requirements for covered exports.

BIS said it would consider approvals necessary to operate existing factories but did not intend to license capacity expansion or technology upgrades. That makes routine upkeep and the introduction of newer processes potentially distinct licensing questions.

Samsung told Reuters its work complied with export regulations and served global customers, including US customers. SK hynix also confirmed US consultations. Public reporting does not establish that all upgrade equipment has been approved; it also does not support alleging unlawful installation.

How can output rise from 5,000 to 40,000 wafers without an expansion?

One Reuters source put Samsung's planned monthly production of 286-layer wafers at 5,000 in December 2026 and 40,000 by December 2027. Samsung nonetheless said it was not adding overall production capacity or constructing a new factory.

These measures are not inherently contradictory. The number of wafers assigned to one new product can rise while wafers assigned to older products decline. Fab-wide wafer throughput could stay constant even if production of the new generation grows rapidly.

For an illustration only, a factory making 100,000 wafers of older product might shift part of the same total output to a new process. This is not an estimate of Xi'an's actual capacity. Furthermore, usable memory bits and qualified shipments depend on density and yields as well as wafer starts.

Neither abandoning existing fabs nor funding unlimited upgrades is straightforward

Reuters sources put the Xi'an plants at roughly 40% of Samsung's worldwide NAND output. That is a significant installed network of machinery, production teams, qualified processes and customer-supply commitments that could not be replicated elsewhere overnight.

SK hynix acquired the Dalian production site with Intel's NAND business in 2021. Its Solidigm business depends on the site, and an October 15 Reuters report described talks over the yet-to-be-fully-equipped Mod B area alongside a potential new US NAND factory. Those talks are not a finalized investment or equipment license.

The strategic tension is clear: YMTC is advancing in China while export-control uncertainty changes the economics and timing of process migration. Holding steady carries an aging-product risk; attempting upgrades entails licensing, capital and manufacturing-transition risks.

The execution capability between R&D and the fab becomes strategic

A design team can establish a newer chip architecture without knowing exactly when a particular overseas plant can convert. Equipment permissions, process compatibility, yield and customer qualifications create different practical timelines.

This does not prove Samsung or SK hynix is adding headcount in specific roles. No resulting hiring announcement or organizational reallocation was confirmed by the cited report. It is an operational inference about engineering capabilities needed for constrained fab transitions.

Useful indicators to monitor include licenses, actual first production of upgraded products, yields and subsequent role-specific hiring disclosures. Until then, layer-count targets should remain distinct from a claimed workforce trend.

Three questions determine whether the modernization works

Will Samsung hit the reported 286-layer wafer output targets? Which pieces of the Dalian transition and possible 300-plus-layer R&D fall within actual US approvals? And will upgrades translate into commercially competitive NAND sales against YMTC?

Layer count, wafer starts, qualified output and China-market share each answer different questions. None alone establishes the return on investment of a manufacturing transition.