China’s CXMT defends bet on risky new DRAM architecture to circumvent US chip curbs

Chinese memory chip manufacturer ChangXin Memory Technologies (CXMT) is moving forward with an unconventional chip architecture, the 4F² design, in an attempt to circumvent US chip curbs. This alternative approach, which US rival Micron Technology has chosen to bypass, has sparked an industry debate over the best method to fit more data onto silicon while adhering to cost and geopolitical constraints.

According to CXMT, the company plans to send customers samples of its first server-grade dynamic random-access memory (DRAM) product utilizing the 4F² architecture by the end of this year. However, the timeline for mass production has not been disclosed. CXMT stated that commercial roll-out would depend on client validation and market demand.

For CXMT, the shift represents a crucial workaround to US export restrictions. With Washington prohibiting Chinese chipmakers from purchasing advanced extreme ultraviolet (EUV) lithography systems from Dutch supplier ASML, this has severed the standard route used by global peers to miniaturize memory circuitry. BofA Securities research note suggested that the 4F² architecture offers an alternative path to higher memory density without the need for EUV tools.

Successful volume manufacturing could bolster CXMT's competitiveness, although the bank cautioned that the process entails higher costs and execution risks. DRAM chips, providing high-speed working memory for various devices, from smartphones to AI data centers, are in high demand as computing workloads increase. Traditional DRAM chips follow a 6F² layout.

The 4F² design, by adopting a tighter layout, reduces each cell footprint by roughly a third at the same feature size, potentially enabling more bits within the same area. Nevertheless, the architecture necessitates additional manufacturing steps and carries a risk of lower production yields, which could diminish its cost advantages.

Micron Technology, the largest US memory producer, remains skeptical that the design makes economic sense at the most advanced level. During a recent webinar, Micron's president and chief technology and products officer, Scott DeBoer, highlighted the physical limits of shrinking memory chips, stating that it is more complicated, complex, and costly.

DeBoer also mentioned that Micron Technology has been researching 4F² for more than two decades and has built prototypes but found fundamental limitations to further shrinking the architecture. When questioned about overseas rivals opting out of the 4F² architecture, CXMT defended its strategy, emphasizing that manufacturing choices reflect each firm's technical experiences, production lines, patent portfolios, and investment plans.

Earlier this year, CXMT reported a 613% surge in third-quarter revenue to 118.71 billion yuan (US$17.8 billion), with profit reaching 54 billion yuan. The company allocated 11.07 billion yuan to research and development during the first nine months, marking a 73% increase from the previous year. In contrast, South Korea's industry leaders, Samsung Electronics and SK Hynix, are maintaining a wait-and-see approach.

Samsung Electronics showcased a working 16-gigabit 4F² test chip at the International Solid-State Circuits Conference in February. Meanwhile, SK Hynix reiterated its focus on extending 6F² using advanced lithography and preparing to transition to 4F² or 3D DRAM when market economics and technical readiness warrant a move. "The timing of technological development and the timing of commercialization are distinct from each other," SK Hynix stated.