Recently, domestic semiconductor companies have struggled with declining stock prices. The volatility has been exacerbated by rising oil prices due to escalating conflicts in the Middle East, inflation concerns, soaring government bond yields in major countries, and the spread of cost-effective strategies in the AI industry.
The aggressive pursuit by Chinese memory semiconductor companies is also cited as a major reason for the stock decline. Changxin Memory Technologies (CXMT) reported a global DRAM market share of 9.5% in the second quarter of this year, more than double the 4% share from the same period last year. Considering that its market share was in the 1% range in 2023, this is a remarkable achievement. CXMT has capitalized on a gap in the general DRAM market while Samsung Electronics, SK Hynix, and Micron have focused their efforts on producing high-bandwidth memory (HBM).
Adding to the turmoil, CXMT announced on October 10 that it would launch next-generation server DRAM using a 4F² structure within the year, shocking the market. Concerns have grown that Chinese semiconductor companies are closing the technology gap. The 4F² structure is a design technology that reduces the memory cell area by 33% compared to the existing 6F². Completing this technology could significantly enhance data processing capabilities, energy efficiency, and productivity.
Initially, Chinese semiconductor companies were assessed to be 1-2 years behind in DRAM technology and over three years behind in HBM technology compared to South Korea. However, this announcement has sparked fears that China is rapidly narrowing the gap. On October 13, SK Hynix's stock plummeted nearly 10%, reflecting significant adjustments in the stock prices of domestic semiconductor companies and the prevailing anxiety in the market.
Industry experts and analysts have offered a more measured assessment, suggesting that CXMT's announcement is more of an exaggerated marketing claim. They point out the lack of specific product specifications or clear mass production timelines, emphasizing the need for objective verification of product performance, yield, and cost competitiveness.
The 4F² structure is not a new concept. Leading domestic companies like Samsung Electronics and SK Hynix have been researching related technologies and securing patents for a long time. The reason they have not yet applied it to mass production is not due to a lack of technology but rather the extremely challenging yield issues and the substantial production cost burdens that make commercialization less viable.
In particular, there is a prevailing analysis that the technology gap in the HBM sector, which is crucial for AI semiconductors, remains robust. With U.S. sanctions blocking the introduction of ASML's extreme ultraviolet (EUV) lithography equipment essential for advanced DRAM and HBM manufacturing, it is nearly impossible for China to mass-produce high-performance, high-yield products in the short term.
On Wall Street, analysts suggest that the recent CXMT shock is merely a pretext for profit-taking in the semiconductor stocks that had recently overheated, and it is difficult to view it as a substantial threat. Jordan Klein, a TMT analyst at Mizuho Securities in Japan, also presented seven reasons to refute CXMT's claims, highlighting the market's excessive concerns.
The profitability of memory semiconductors, which had approached 80%, cannot be sustained indefinitely. Investments in AI infrastructure are also facing practical constraints such as power grid shortages and capital input limits. In this context, it is natural for semiconductor stocks to take a breather. The significant market reaction to CXMT's announcement reflects accumulated caution. However, it is difficult to assert that CXMT's announcement will immediately change the industry landscape. The actual proof of CXMT's next-generation DRAM mass production declaration will require rigorous verification in the future.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.