Charles Kau, the former president of Nanya Technology who is known as the "godfather" of the DRAM industry, has warned that the large-scale fab expansion by China's memory maker CXMT (ChangXin Memory Technologies) could push the global DRAM market into oversupply within two years. Kau also expressed skepticism that TSMC will enter DRAM manufacturing. But any assessment of his warning should also consider how much price competition CXMT can withstand. The company's gross margin for the July–September 2026 quarter reached 89.53%, and under certain conditions it could maintain a gross margin of about 79% even if its selling prices were cut in half. If oversupply does materialize, CXMT appears to have room to cut prices and expand its market share.
What Kau's remarks do and do not establish
According to a TechNews report dated November 10, Kau said CXMT is building 10 fabs in Beijing, Shanghai and elsewhere, and that their production capacity will be fully ramped within the next two years. He argued that even if CXMT cannot access EUV (extreme ultraviolet) lithography equipment and remains limited to mature process technology, a large volume of DRAM reaching the market could still cause oversupply. He also rebutted the view held by some Taiwanese industry insiders that China's added output will be absorbed by domestic demand and will not affect the global market, calling it a temporary perception created by the current boom.
That said, the figure of 10 fabs rests on Kau's statement, and the other sources reviewed for this article do not corroborate the number. According to Commercial Times reporting cited by TrendForce, CXMT's monthly wafer capacity is expected to grow from about 320,000 wafers today to 420,000 in 2027. The added output is expected to go mainly to demand within China, and some institutional investors believe China's capacity expansion will not clearly affect global supply and demand until 2027 or later. Meanwhile, TrendForce's tally shows CXMT's share of global DRAM revenue reached 9.5% in April–June 2026, making it the world's fourth-largest supplier. Kau's warning has a credible backdrop, but his remarks should not be treated as a confirmed production plan.
Kau also took a cautious view of next-generation DRAM technology. Regarding the "4F²" structure, which reduces the memory cell area to four times the square of the minimum feature size F, he noted that 6F² is currently the mainstream and that no company has achieved mass production of 4F². Even if CXMT spares no expense on development, he said, it is not guaranteed to clear the hurdles to volume production. On the other hand, according to The Paper's reporting relayed by TrendForce, CXMT has announced that it completed verification of engineering samples on its sixth-generation platform, which uses 4F² vertical-channel DRAM. However, validating prototypes and commercial mass production are separate stages, and manufacturing yields and production costs still need to be verified.
The oversupply scenario does not necessarily hinge on whether 4F² technology succeeds. Kau himself pointed out that increased DRAM output using mature processes could be enough to cause a glut. As for TSMC, Kau ruled out a move into DRAM, but a TSMC spokesperson said only that the company "has consistently focused on its core business," without explicitly addressing DRAM manufacturing. As XenoSpectrum noted in its November 9 article, that statement cannot be read as a clear denial of DRAM entry. Kau's comments are his own view and need to be distinguished from TSMC's official position.
From an 89.53% gross margin to about 79% after a price halving
CXMT's high gross margin is an important factor in judging its competitiveness if oversupply occurs. According to CXMT's July–September 2026 earnings report, its gross margin rose 1.94 percentage points from the previous quarter to 89.53%, and reached 86.85% for the January–September period. Management cited a shift in sales mix toward higher-value products and higher selling prices as the drivers of the improvement. In TrendForce's comparison, Micron's gross margin for the quarter ended September 3, 2026 was about 87%, and Samsung Electronics' consolidated gross margin for July–September was 69.3%. However, each company's product mix and reporting period differ, so a simple comparison cannot be used to judge the cost competitiveness of their DRAM businesses.
Gross margin is the share of revenue left after subtracting cost of sales from sales. Based on CXMT's 89.53% gross margin, its cost of sales equals 10.47% of revenue. Suppose sales volume and product mix stay the same, all product prices are cut uniformly by half, and total cost of sales does not change. The gross margin could then be calculated as follows:
Gross margin after a price halving = 1 − (0.1047 ÷ 0.5) = 79.06%
In other words, under this simple model, a gross margin of about 79.1% remains even if selling prices are halved. Gross profit would reach zero only if prices fell to about 10.5% of the current level, a decline of about 89.5%. However, this is a hypothetical calculation based on the company-wide gross margin for July–September 2026, and it does not mean that every individual DRAM product has the same room to cut prices. Actual cost of sales includes items such as depreciation of manufacturing equipment, and unit costs fluctuate with utilization rates, production volumes and product mix.
Applying the same calculation to Micron's 87% gross margin gives a post-halving gross margin of about 74%, and Samsung's 69.3% gives about 39%. On paper CXMT has the most room, but Samsung's figure includes non-memory businesses and Micron's reporting period differs, so the result cannot be treated as a direct comparison of each company's ability to cut prices. CXMT's management has also referred to cost increases across the supply chain, so there is no guarantee that the current cost level will hold. The roughly 79% figure is therefore not a forecast of future gross margin, but a reference value showing how much of a price decline the current earnings structure could absorb under certain conditions.
What matters is that CXMT is not currently using this pricing headroom for aggressive price cuts. According to Star Market Daily reporting cited by TrendForce, CXMT says its product prices are broadly in line with those of overseas competitors. For now, the company appears to be pursuing a strategy of securing high margins while keeping prices comparable to rivals, rather than winning demand through price cuts. It will likely be able to maintain this approach while DRAM supply remains tight. If the oversupply Kau warns about materializes, however, cutting prices to expand sales volume and market share could become an option for CXMT.
Such price competition could also affect component procurement for smartphones and PCs. A August report by Korea's Aju News, relayed by TrendForce, said Apple is testing CXMT memory for the iPhone and MacBook. If CXMT moved to cut prices, device makers considering its memory could gain leverage in price negotiations with existing suppliers. Whether testing progresses to actual adoption, however, depends not only on price but also on quality and supply stability.
What will decide oversupply: Chinese domestic demand and what to watch next
The strongest counterargument to Kau's warning is the possibility that CXMT's added output will be absorbed by demand within China. TrendForce's September 24 report suggested that growth in China's AI and cloud market could absorb new capacity through 2027. CXMT's management also says DRAM supply will remain tight and expects its gross margin to be stable in October–December 2026. Indeed, the materials reviewed for this article contain no clear evidence that oversupply will occur before 2027.
The reason to focus on CXMT's high gross margin anyway is not to predict that oversupply will certainly happen. It is that CXMT may have enough earnings headroom to turn to price competition if oversupply does occur. While domestic demand in China can absorb the added output, there is little need for aggressive price cuts. But if demand growth fails to keep pace with capacity expansion, the incentive to cut prices to secure customers will strengthen. CXMT says it is negotiating long-term supply agreements with several major customers and that progress is going well, but the pricing terms have not been disclosed. If contracts fix prices or supply volumes for a set period, the room for short-term price cuts would change as well.
Testing this view requires continued tracking of the relationship between selling prices and gross margin. Taking July–September 2026 as the baseline and letting x be the rate of price decline (x = 0.1 for a 10% decline), the gross margin when sales volume, product mix and total cost of sales are unchanged is given by "1 − 0.1047 ÷ (1 − x)." If the actual gross margin falls well below this estimate, cost increases or shifts in product mix may have reduced the pricing headroom originally assumed. The gap alone cannot identify the cause, however, and sales volumes and manufacturing costs would also need to be checked.
Conversely, if CXMT's gross margin holds at around 89% even as monthly wafer capacity approaches 420,000 toward 2027, that would mean the company is maintaining high profitability. Even so, it would not prove that oversupply has not occurred, since a shift to higher-value products or lower manufacturing costs could be offsetting falling selling prices. The focus going forward is whether Chinese domestic demand can keep absorbing the added output, and how far CXMT would cut prices if the market loosens. In upcoming earnings reports, checking selling prices, production capacity, utilization and product mix alongside gross margin will be key to evaluating both Kau's warning and CXMT's price competitiveness.