China's CXMT Rises Amidst Generic Memory Shortage, Will Global DRAM Market Structure Change?

This situation has led to generic DRAM and NAND flash prices surging by at least 4 times, and even 5-6 times, compared to the same period last year.

Professor Kwon Seok-jun, Associate Dean of the College of Engineering at Sungkyunkwan University, explained, "The shortage in generic semiconductors was unexpectedly large, causing prices of generic semiconductors, for example, DRAM and NAND flash, to rise by at least four times, and even five to six times compared to the same period last year. That itself is actually the biggest driving force of the memory super cycle."

The boom in the memory market is providing unprecedented opportunities not only for established players but also for latecomers. Several latecomers, including Micron, are reporting record profits, with the rise of emerging players like China's CXMT and Taiwan's Nanya Technology being particularly notable.

Professor Kwon Seok-jun stated, "Companies that were previously 4th in the industry, like China's CXMT, and 5th, like Taiwan's Nanya Technology—companies that were somewhat considered second-tier, even including Micron Electronics—are also achieving record profits." He emphasized their growth.

This capital acquisition, combined with China's national policy to foster AI semiconductors, is accelerating CXMT's growth.

Professor Kwon Seok-jun explained, "The capital power, the dynamics of cash flow, and the policy background where China inevitably has to support AI semiconductors at a national level—these factors combined have suddenly caused CXMT's scale and status to surge." This indicates that policy support played a crucial role in CXMT's growth.

With China's CXMT growing rapidly, there's a possibility it could reconfigure into a Samsung, SK Hynix, and CXMT trio.

Professor Kwon Seok-jun stated, "Until now, everyone knows the three-company rule in DRAM: Samsung, Hynix, Micron. But now, it could change to Samsung, Hynix, CXMT." He predicts that within the next 5-6 years, CXMT will expand its global market share by catering to the Chinese domestic market, leading this transformation. This raises vigilance regarding the existing market structure.

In CXMT's case, out of the 100 trillion won secured, 70 trillion won is planned for mass production and 30 trillion won for research and development (R&D).

Professor Kwon Seok-jun explained, "If about 100 trillion won in capital is generated, about 70 trillion won of that is used for mass production. About 30 trillion goes into R&D. Naturally, a significant portion of that R&D is used to optimize processes internally, but it also goes into purchasing more domestically produced equipment in China and optimizing that equipment." This is analyzed as a strategy to rapidly reduce the technology gap through equipment localization and lead time reduction. Given that capital can be a key driver of technological advancement, the global semiconductor industry is closely watching China's moves.

Professor Kwon Seok-jun pointed out that even a reduction in the investment growth rate of AI data center operators from 10% to 7-5% could send a very negative signal to the market.

He commented, "This could send a very frightening signal to the market. 'Oh, investment is slowing down. Oh, perhaps they won't invest as much as before. Then perhaps they'll buy less memory?'" He predicted that the memory industry might prepare to adjust the pace of investment due to concerns about reduced demand.

The majority of AI operators believe that current computing capacity remains insufficient.

Professor Kwon Seok-jun reported, "The situation on the ground is still, for now, somewhat dominated by optimism. This investment will continue." They emphasize the need for continuous investment, stressing that scaling up beyond trillion-won frontier models to hundreds of trillions of won is necessary.

Professor Kwon Seok-jun pointed out a lack of clear return on investment (ROI) models beyond code automation. Additionally, a significant challenge lies in acquiring the large quantities of domain-specific data needed to implement physical AI, which is proving to be more difficult than expected.

From an economic perspective of token generation, analysis suggests that Chinese AI models are more efficient. Professor Kwon Seok-jun conveyed that this causes fear among American model developers.

He said, "Coincidentally, from the economic perspective of token generation, Chinese models are currently leading. From the perspective of American model developers, there is also a sense of fear." Currently, the market is divided between high-performance AI models from the US and cost-effective models from China, creating a dilemma for US big tech companies regarding the direction of their model development in terms of profitability.

Signs are emerging that even Google, a titan in the industry, is facing cash flow difficulties. Professor Kwon Seok-jun revealed indications that their retained earnings and accumulated cash reserves are running low.

He stated, "Even Google, a giant, is showing signs that its cash flow is becoming increasingly difficult, and its retained earnings and accumulated cash are quite depleted." He analyzed that this could signal the end of an era where big tech companies could sustain themselves purely on their own cash flow. Experts predict an increased reliance on corporate bond issuance or external funding in the future.