China's two biggest memory makers are moving toward each other's core businesses. Reuters reported on September 18, 2026 that CXMT (ChangXin Memory Technologies), a DRAM maker, is planning an R&D production line for NAND flash memory in Beijing. In April, Reuters also reported that YMTC (Yangtze Memory Technologies), which focuses on NAND, had sent prototype low-power DRAM (LPDDR) to customers.
Taken together, the two moves suggest that the business areas of China's DRAM and NAND specialists are beginning to overlap. But one is a plan for a research line, and the other is a report about prototypes handed to customers. Neither confirms mass production or continued shipments in the new field. What would each company's new products have to clear before they turn into real supply competition?
How far along is each company's crossover?
According to Reuters' September 18 report, CXMT is planning an NAND R&D production line at a new plant in Beijing. It is also holding talks with customers, including companies that want to use the chips in storage products for AI systems and supercomputers. However, it is not known when the line will start operating or whether it will move on to large-scale commercial production. Interest in buying is not confirmation of a supply contract or of deliveries.
Reuters' April 14 report on YMTC, meanwhile, says the company sent LPDDR prototypes to customers. According to Reuters, YMTC will collect customer feedback by the end of the year and then decide on DRAM production. Whether customers will decide to adopt the product after receiving the prototypes, and whether it will go into mass production, has not been disclosed.
Lining up what each company already mass-produces clarifies where the new moves stand. In an official announcement on September 20, CXMT said it is mass-producing its fifth-generation DRAM manufacturing platform, "G5," and two 24Gb LPDDR5X products that use it. YMTC states on its official corporate profile that it began mass production of fifth-generation TLC 3D NAND in 2024. Both are track records in each company's existing core field.
| Company | Core products announced | Move into the other's field | Not yet confirmed |
|---|---|---|---|
| CXMT | DRAM. Announced mass production of two G5-based LPDDR5X products on September 20, 2026 | Reuters reported on September 18 that it is planning an NAND R&D production line in Beijing | Line start-up, chip specifications, commercial mass production, continued shipments |
| YMTC | 3D NAND. Its website says it mass-produced fifth-generation TLC products in 2024 | Reuters reported on April 14 that it sent LPDDR prototypes to customers | Evaluation results, a decision on DRAM mass production, continued shipments |
YMTC has reached the stage of sending LPDDR prototypes to customers, while CXMT is at the stage of planning an NAND R&D production line. Mass production and continued shipments have not been confirmed for either company's crossover products. The table shows the progress that can be verified for each company's new business. Handing LPDDR prototypes to customers indicates that a channel now exists for checking operation and quality for specific applications. But customer evaluation results, the devices that will adopt the chips and order volumes have not been disclosed. CXMT's customer talks likewise do not mean NAND chip supply has begun. Describing both companies together as "entering the other's market" glosses over the fact that the types of confirmed facts differ. CXMT's G5 mass production cannot be read as NAND mass production, and YMTC's NAND mass production cannot be read as DRAM mass production.
Factory capacity is not the same as supply of new products
Reuters reported that YMTC's third Wuhan fab is expected to start operating at the end of 2026 and to be able to process 50,000 wafers a month in 2027. It added that two additional planned fabs would each have a design capacity of 100,000 wafers a month when completed. According to people familiar with the matter, there is also a concept of shifting part of the three fabs to DRAM, but the allocation will depend on how development progresses.
Monthly capacity here refers to the wafer volume the equipment could process once completed, not current DRAM shipments. Building a factory, installing equipment, trial production and customer quality verification are separate stages. Moreover, the mere idea of assigning part of the capacity to DRAM does not make it possible to calculate a specific split between NAND and DRAM. Adding up 50,000 or 100,000 wafers a month as confirmed DRAM supply would lead to a misjudgment. Nor is the number of wafers that can be processed the same as the number of good chips obtained. Even when a new type of memory is made on the same site, the volume that can actually be supplied changes with the equipment configuration and the maturity of the process. Nothing in the fab plans Reuters described provides a basis for working backward to a DRAM market share.
CXMT, for its part, cannot necessarily transfer its DRAM progress directly to NAND. The scaling and 24Gb capacity figures the company has published for G5 describe a DRAM manufacturing platform and LPDDR5X products. For NAND, it would have to separately establish the cell-stacking structure, read/write characteristics, reliability and manufacturing yield. The NAND plan has no published figures for layer count, capacity, performance or production scale.
The same caution applies in the other direction. YMTC's Xtacking technology is a scheme in which NAND memory cells and control peripheral circuits are built on separate wafers and then bonded. Owning stacking technology is a fact, but that description alone does not show that YMTC has achieved LPDDR or HBM manufacturing processes, yields or customer qualification. DRAM holds data at high speed while a device is running, whereas NAND keeps data after the power is turned off. The requirements the products must meet also differ.
Supply and demand outlooks also diverge for DRAM and NAND
If the crossovers materialize, Chinese device makers may gain more sourcing options. But at the research-line or prototype stage, it is impossible to predict that prices for PC memory or SSDs will fall. To affect prices, a company must keep delivering chips with competitive performance and quality, in the volumes customers need and when they need them.
TrendForce's outlook, released on July 30, 2026, also shows that the two markets are not heading in the same direction. It forecasts that DRAM supply will remain tight in 2027 because of demand from AI servers and the allocation of manufacturing capacity to HBM. NAND, by contrast, is expected to loosen in the second half of 2027 as new capacity comes online and consumer demand is weak. This is a forecast of industry-wide supply and demand, not a forecast that CXMT's or YMTC's new products will enter mass production on schedule. Even if both companies press ahead with development, the time they can bring products to market may not coincide with supply and demand at that point. For that reason, the design capacity of the factories Reuters described cannot be converted into current shipment volumes or future confirmed supply.
Even within the same "memory shortage," DRAM and NAND differ in investment timing and in the customers expected to generate revenue. YMTC's move toward DRAM and CXMT's move toward NAND could add to China's sources of supply, but they also highlight the challenge of building manufacturing technology and a customer base outside the markets where each company has excelled so far.
What to check next is customers' evaluation of YMTC's LPDDR prototypes and its mass-production decision, and the start-up date and product specifications of CXMT's NAND line. Only after yields, delivery destinations and continued supply become visible can the scale at which the two companies actually compete in each other's markets be measured. What can be seen now is two different stages of preparation for that competition.