Yangtze Memory Technologies Co. (YMTC), China’s top NAND flash manufacturer and one of the most heavily scrutinized names in the chip war, is pushing ahead with plans for a third semiconductor plant in Wuhan. The facility is expected to begin mass production in 2027, in a move that underscores Beijing’s intention to double down on domestic memory capacity even as YMTC remains barred from acquiring advanced U.S. chipmaking tools.
The Wuhan expansion marks the company’s most ambitious project since being placed on the U.S. Entity List in December 2022, a designation that has crippled access to EUV and advanced DUV lithography, precision etching gear, and key U.S.-origin design IP. Yet, instead of collapsing, YMTC is adapting slowly with the full backing of China’s industrial policy apparatus.
Surviving the blacklist
YMTC’s situation is unlike that of Huawei, which was forced into a radical pivot toward software and cloud services following its sanctions. YMTC, by contrast, continues to manufacture supply-constrained NAND flash chips at scale, albeit at reduced efficiency and with aging process nodes. Its most advanced product, the 232-layer Xtacking 3.0 NAND, was benchmarked in late 2022 as competitive with Micron’s and Samsung’s offerings. But sustaining that lead has proven impossible without access to ASML’s cutting-edge tooling.
Various reports suggest that YMTC’s Fab 2 has been operating with domestic workarounds and modified legacy tools, relying in part on procurement from third-party intermediaries to secure essential chemicals and maintenance components. However, these workarounds are slow and painful.
One potential candidate is SMEE’s immersion lithography equipment, but this has yet to deliver consistent yields beyond 28nm logic equivalents, far from the sub-20nm half-pitch necessary for leading-edge NAND stacking. Still, YMTC’s new fab suggests confidence that by 2027, domestic toolchains will be capable enough to support high-volume NAND production if not at the bleeding edge, then at commercial scale.
Memory nationalism
(Image credit: YMTC)
The report by Nikkei Asia suggests that the timing of this announcement also aligns with China’s broader recalibration toward AI infrastructure resilience. With U.S. export restrictions locking out Chinese firms from Nvidia H100 and A100-class GPUs, Beijing has redirected its subsidies and engineering talent toward local accelerator design, sovereign foundation models, and the underlying memory and packaging stacks that make them viable.
In that context, NAND flash is arguably foundational. While DRAM remains the performance bottleneck in AI inference, NAND underpins persistent storage for massive model checkpoints, training data lakes, and parameter weights for retrieval-augmented generation. YMTC’s success here is thus critical not because its chips are the fastest, but because they’re Chinese.
... continue reading