CXMT Reportedly Plans a Beijing NAND R&D Line—Not a Commercial Flash Fab Yet

CXMT is reportedly preparing a research-and-development and trial-production line for NAND flash memory at its new Beijing site. If it proceeds, the initiative would move the Chinese chipmaker beyond its core DRAM business and toward persistent storage for AI systems, servers, and other computing products. But the key point is timing: this is a reported R&D plan, not an announced high-volume NAND manufacturing program. 1

What CXMT is reportedly building in Beijing

People familiar with the matter told Reuters that ChangXin Memory Technologies (CXMT) plans to establish a NAND flash R&D production line at its new Beijing plant. The company has also set up a Beijing research institute with NAND development among its projects. 1

CXMT has reportedly discussed the NAND initiative with prospective customers, including a startup seeking storage chips for AI systems and supercomputers. That indicates an effort to validate demand alongside technology development rather than a confirmed volume-supply agreement. 1

NAND and DRAM serve different roles in computing systems. DRAM provides fast working memory, while NAND is persistent flash storage used in products such as SSDs and data-center storage systems. A NAND program could therefore let CXMT offer a broader set of memory components to customers already buying—or evaluating—its DRAM products.

Why CXMT would expand into NAND now

The reported move comes amid a global memory shortage tied to AI infrastructure demand. Reuters reported that industry executives expected supply pressure to persist through at least 2027, while NAND supply additions have been constrained as major manufacturers prioritized DRAM and high-bandwidth memory investments. 1

For CXMT, NAND could create access to a wider group of domestic buyers, including AI-system builders, server makers, data-center operators, storage vendors, PC manufacturers, and device companies. The reported customer discussions around AI and supercomputer storage offer the clearest evidence of that commercial logic. 1

It would also give CXMT a more complete memory portfolio. That does not guarantee that the company will become a meaningful NAND supplier quickly, but it could make CXMT more relevant to customers looking to source both working memory and storage memory from Chinese vendors.

A direct challenge to YMTC—and a longer-term challenge to global suppliers

CXMT has been identified mainly with DRAM, while Yangtze Memory Technologies Co. (YMTC) has been China’s established NAND specialist. A CXMT NAND initiative would blur that division and bring the two companies closer together in the same market. Reuters also reported that YMTC had sent low-power DRAM samples to customers, suggesting that the historical DRAM-versus-NAND split between the two companies was already becoming less distinct. 1

In the broader NAND market, CXMT would become a potential additional competitor to established suppliers including Samsung Electronics, SK hynix, and Micron. However, the reported Beijing project should not be read as evidence that CXMT is already comparable to those companies in NAND technology, scale, output, or customer qualification. The known plan is for R&D and trial production, while the incumbents operate established commercial NAND businesses. 1

What is still unknown

The Reuters report left the most important execution questions unanswered:

• Start date: There is no reported date for the R&D line to begin operating. 1
• Commercial transition: CXMT has not publicly confirmed whether the trial line will become a high-volume NAND fab. 1
• Technology: No process node, layer count, NAND architecture, or named stacking or bonding technology was disclosed for the proposed line. 1
• Capacity and output: There are no reported wafer-start targets, yield figures, shipment dates, sales targets, or commercial volumes. 1

Those omissions matter. An R&D line can help develop process expertise and customer samples, but it does not by itself add meaningful near-term NAND supply. There is not enough public evidence to conclude that CXMT-made NAND will materially ease the current storage shortage or soon become a major source of SSD and server-storage flash. 1

Keep CXMT’s separate projects distinct

Several CXMT reports concern different technologies and facilities, and they should not be combined into one confirmed NAND roadmap.

The proposed second Beijing memory fab

In August, Reuters reported that CXMT was considering a second 12-inch memory-chip plant in Beijing’s Yizhuang area and was in funding discussions with a local-government-backed technology manufacturing hub. That report described a possible capacity expansion; it did not establish that the project would be a commercial NAND fab or disclose a committed NAND production schedule. 3

HBM is a separate DRAM-based effort

CXMT has also reportedly begun producing HBM3E in small quantities, with Chinese chip designers including Alibaba’s T-Head and Cambricon testing the memory with their processors. HBM is a stacked DRAM product designed for AI computing, not NAND flash. 2

Separate reporting put CXMT’s early HBM3 trial-production yield at about 25%, versus an approximately 80% benchmark cited for mass production. That reported figure concerns HBM manufacturing and should not be treated as a NAND yield or as evidence about the proposed Beijing NAND line. 17

No confirmed XStacking roadmap for the reported CXMT NAND line

The September 18 Reuters report did not identify XStacking, or any other named NAND bonding or stacking process, for CXMT’s proposed line. On the available evidence, it is not possible to attribute a specific NAND technology, node, or production roadmap to the project. 1

The practical takeaway

CXMT’s reported Beijing NAND plan is strategically significant because it could expand the company from DRAM into a complementary storage market at a time of tight global memory supply. It could also intensify competition with YMTC inside China and, over time, add another challenger to the global NAND market. 1

For now, though, it is best understood as an early research and pilot-production initiative. The indicators that would turn it into a major market event are still missing: an operating date, technical specifications, customer qualifications, commercial capacity, yields, and a decision to ramp mass production.