CXMT claims 4F² memory breakthrough as South Korea’s KOSPI falls 6%

CXMT claimed over the weekend that it had achieved a 4F² memory-chip design, a disclosure followed on Tuesday by a 6% fall in South Korea’s KOSPI. The Chinese memory-chip maker said the proposed design would place greater computing capacity in a smaller area, but disclosed no further technical details about the architecture.

CXMT claims a 4F² server chip before year-end

CXMT President Cao Kanyu said the company had developed a chip using a 4F² architecture. He said the design would provide greater computing capacity within a smaller area.

CXMT’s spokesperson said the proposed product was for data-centre servers and gave its launch window as before the end of the year. Cao provided no further technical details about how the architecture would work.

South Korea’s KOSPI falls 6% as memory shares decline

KOSPI data showed that South Korea’s benchmark fell 6% on Tuesday, its steepest single-day decline since July. The movement followed CXMT’s disclosure of the claimed memory technology.

South Korean market data showed Samsung Electronics and SK Hynix shares each fell more than 7% on Tuesday. United States market data showed Micron Technology shares had fallen 4% overnight.

CXMT’s 4F² layout differs from the 6F² standard

The existing memory-chip manufacturing standard is 6F², while the technology efforts disclosed for Samsung Electronics and SK Hynix include work on moving to 4F².

Cao said CXMT was pursuing a different transistor layout through a stacking technique called Vertical Surrounding Gate. The 4F² claim had not been independently verified.

Morningstar analyst Jing Jie Yu said memory-chip and equipment manufacturers had highlighted challenges in the transition to the 4F² architecture. Yu said the time between CXMT’s unveiling and commercial production at viable yields could be considerable.

CXMT trails Samsung and SK Hynix in high-bandwidth memory

Industry data cited with CXMT’s disclosure placed the company behind Samsung Electronics and SK Hynix in high-bandwidth memory. Nvidia uses that type of memory in its artificial-intelligence accelerators.

Counterpoint Research data showed CXMT’s broader DRAM market share at 10% in the second quarter, compared with 4% a year earlier. The company said it had been expanding advanced-memory production for artificial-intelligence applications.

Disclaimer: This article has been written exclusively for informational and educational purposes and does not constitute investment advice. Investments in securities are subject to market risks; please read all scheme/offer documents carefully and consult a SEBI-registered advisor before investing.