Global chip stocks tumble as China's CXMT unveils next-generation DRAM plans
CXMT President Cao Kanyu said that the company made "key progress" on server memory products that use 4F² technology, a new cell-based architecture that boosts memory density.
A logo of China's CXMT on the exterior of the company's production facility in Beijing on July 29REUTERS/YONHAP
Global semiconductor stocks tumbled on Tuesday, as Chinese memory chipmaker CXMT announced plans to launch server memory products that use next-generation 4F² technology by the end of this year.
CXMT President Cao Kanyu said that the company made “key progress” on a new cell-based architecture that boosts memory density — the 4F² technology — during a conference in Shanghai on Saturday, the South China Morning Post reported on Monday. The chipmaker is also expected to release a new generation of DDR5 registered dual in-line memory modules using the technology by the end of the year.
Markets reacted immediately. Shares of U.S. memory chipmaker Micron Technology fell 3.88 percent on the New York Stock Exchange on Monday, and the Philadelphia Semiconductor Index dropped 2.67 percent.
In the Korean stock market, Samsung Electronics and SK hynix shares fell more than 6 percent and 9 percent, respectively, at one point during intraday trading on Tuesday.
The declines reflected growing concerns that CXMT could move beyond conventional dynamic random-access memory (DRAM) chips to compete in next-generation technologies and high-value server memory products currently being developed by industry leaders.
The 4F² architecture is a next-generation design that reduces the memory cells’ area compared to the 6F² architecture widely used in commercial DRAM chips. In terms of cell area alone, it reduces each cell’s size by about 33 percent, theoretically allowing 50 percent more cells to fit overall.
Rather than relying solely on making chips smaller, the technology aims to increase memory density through structural changes, including arranging transistors vertically and separating memory cells from peripheral circuits.
Samsung Electronics is also developing next-generation DRAM using 4F² architecture and vertical channel transistors. SK hynix unveiled a technology road map last year, outlining plans to apply 4F² vertical gate technology to DRAM chips using sub-10-nanometer processes, alongside wafer bonding technology.
The logo of China's CXMT on the exterior of the company's production facility in Beijing on July 29REUTERS/YONHAP
While CXMT’s announcement shook the market, some say that it is too early to assess its ability to compete in mass production. CXMT not disclosed detailed product specifications, yield rates, die sizes, power consumption, wafer production capacity or customer qualification results.
“Even if CXMT manages to develop a 4F² product by the end of the year, it remains questionable whether the company can produce enough chips to have a meaningful impact on the market,” Chung In-sung, an author and former SK hynix researcher, said.
Considerable delays followed previous announcements.
In 2022, Taiwanese media and other outlets reported that CXMT was targeting mass production of DDR5 chips by the end of that year. However, DDR5 products incorporating CXMT chips did not begin appearing in the market in earnest until late 2024. Industry observers also reported last year that CXMT was struggling to expand production of its newer DDR5 chips due to yield and reliability issues.
Following growing competition, Samsung Electronics is accelerating its efforts to maintain its technological edge. It unveiled its eighth-generation high bandwidth memory (HBM), known as HBM5; custom HBM; Compute Express Link-based memory expansion technologies; and next-generation solid-state drives at the 2026 OCP Global Summit in Silicon Valley, California, on Tuesday. OCP stands for Open Compute Project.
Shim Ho-jun, the head of Samsung Electronics' memory application engineering team, speaks during the Open Compute Project Global Summit 2026 in Silicon Valley in California on Oct. 12.SAMSUNG ELECTRONICS
The move reflects Samsung’s strategy to extend its technological competitiveness beyond HBM to the broader AI infrastructure market, including system memory and storage.
Some experts argue that Korean memory chipmakers must also expand production capacity more quickly to widen their lead over Chinese competitors. SK Group Chairman Chey Tae-won’s remarks in Shanghai on Sunday reflected similar concerns.
“If China or Micron in the United States expands production capacity faster than we do, our country’s market share will decline,” he said. “We need to expand [production facilities] quickly.”
Experts also stressed the need for government support.
“To maintain a global market share of around 80 percent and establish an overwhelming lead in production capacity and technology, what matters is not only corporate investment but also how quickly the government implements support measures,” said Lee Jong-hwan, a professor in the Department of System Semiconductor Engineering at Sangmyung University.
This article was originally written in Korean and translated by a bilingual reporter with the help of generative AI tools. It was then edited by a native English-speaking editor. All AI-assisted translations are reviewed and refined by our newsroom.