Changxin Technology, the domestic DRAM leader better known as CXMT, became one of the hottest stories in China’s tech and capital markets after listing on the STAR Market on July 27. The stock opened at 49.50 yuan per share, up more than 470% from its 8.66 yuan offer price. With a market capitalization above 3.6 trillion yuan, it overtook Industrial and Commercial Bank of China to rank first by market value in China’s equity market and stood at more than twice the size of Kweichow Moutai.
CXMT’s debut also pulled a Nomura research note into the spotlight. Ahead of trading on July 27, Nomura initiated coverage with a Buy rating and a target price of 116 yuan. The report described the industrial value of CXMT’s DRAM chips as “the jewel in China’s crown.” Based on the opening price, that target implied more than 12x upside.
That call immediately triggered questions over how far valuation can stretch. One investor who has long followed China’s semiconductor industry told Tencent Technology that a target above 100 yuan would imply a market value of more than 7 trillion yuan and looked aggressive, though not impossible if market sentiment stays extremely strong and pulls forward at least three years of growth.
The central issue behind the debate is straightforward: what is the logic behind Nomura’s bullish view, and how does the bank frame the risks around CXMT’s outlook?
Nomura projects revenue of 560.8 billion yuan in 2027
According to data disclosed in CXMT’s listing press release, the company expects revenue for January to June 2026 to come in at 110 billion yuan to 120 billion yuan, up 612.53% to 677.31% year on year. It expects net profit attributable to shareholders to reach 50 billion yuan to 57 billion yuan, up 2244.03% to 2544.19%.
CXMT said the surge in earnings was driven by a recovery in the memory industry cycle, an improved product mix, and scale effects.
Nomura’s model is more aggressive than the company’s own near-term guidance. The bank projects revenue rising from 61.8 billion yuan in 2025 to 290.7 billion yuan in 2026, 560.8 billion yuan in 2027, and 773.3 billion yuan in 2028. It expects net profit attributable to shareholders to climb from less than 1.9 billion yuan in 2025 to 130.3 billion yuan in 2026, 277.2 billion yuan in 2027, and 393.1 billion yuan in 2028.
That implies compound annual growth rates of 63% for revenue and 74% for net profit.
Nomura’s framework rests on three main drivers: capacity expansion, technology upgrades, and higher prices. The first two line up closely with the use of proceeds disclosed in CXMT’s prospectus.
The company previously said the funds raised would mainly go toward technology upgrades for memory wafer mass-production lines, DRAM technology upgrades, and forward-looking DRAM research and development, with the stated goal of improving advanced manufacturing capability and innovation.
On pricing, the report points to two factors. One is a higher average selling price per wafer as products move up the technology curve. The other is a broad increase in memory chip prices during the current supercycle. In the overview section of its prospectus, CXMT wrote that since the second half of 2025, continuing price increases had lifted gross margin and profit levels and helped the company turn profitable in 2025.
Product mix shifts as Apple testing draws attention
CXMT’s shipment mix is also changing. In an earlier response to regulatory inquiries, the company said mobile devices remain the foundation of revenue. Xiaomi, Transsion, Honor, vivo, and OPPO are tied to its LPDDR product line, while cloud customers such as Alibaba and ByteDance mainly correspond to DDR products.
At that point, the two customer groups contributed revenue in roughly a 9:1 ratio.
By May 17, 2026, AI server-related products represented by the DDR line accounted for more than 30% of revenue, while LPDDR remained the main business, contributing more than 66%.
One possible customer is drawing particular attention during this ramp-up phase. According to the Financial Times, Apple has started testing CXMT’s DRAM chips for use in entry-level iPhones and other devices sold in the Chinese market. The report also said Apple has been lobbying the U.S. government since May and June this year in an effort to obtain the required approvals.
If that relationship is finalized, CXMT could take part in Apple’s global DRAM procurement. The significance, as framed in the source article, goes beyond the revenue contribution. It would also signal a shift from being seen largely as a domestic substitution supplier toward becoming one recognized by mainstream global customers.
Not every institution shares the same level of optimism. Some believe Apple may only purchase small volumes from CXMT and may also be using the introduction of a new supply variable as leverage in talks with Samsung and SK Hynix.
Nomura’s 116 yuan target implies about 20x forward earnings
Using Nomura’s 116 yuan target price and 2028 earnings per share estimate of 5.8 yuan, CXMT would trade at about 20x forward price-to-earnings.
Chen Qi, an investor who has long followed the semiconductor sector, previously told Tencent Technology that more than 100 billion yuan in profit could reasonably support a market value of 2 trillion yuan. In his view, that corresponds to 20x PE.
For comparison, Wall Street expects Micron to trade at about 10x 2026 calendar-year earnings, while SK Hynix is closer to 5x. Chen said “anything can happen” when the semiconductor cycle turns up. He added that while 10x PE is reasonable, 20x PE is also within the range of possibility because CXMT stands opposite Samsung, SK Hynix, and Micron and represents China’s memory industry taking a seat at the table.
Nomura’s valuation logic has another layer. The bank treats Micron as the global valuation anchor for DRAM, with a historical center for forward PE around 10x. Since China’s A-share semiconductor sector has long traded at a 1x to 3x premium to comparable U.S. names, applying a midpoint premium of roughly 2x leads to a 20x multiple for CXMT.
AI demand and supply limits sit at the center of the bull case
A key assumption behind the valuation is that AI is pushing global memory demand into a steep expansion path while supply remains constrained by physical bottlenecks.
Nomura estimates that even if the compression effect from storage-efficiency technologies is discounted to 40% of its theoretical impact, global memory usage demand would still rise more than sevenfold between 2026 and 2030, implying a compound annual growth rate above 60%.
The bank also outlined a more extreme scenario. If AI robots were to operate autonomously without being limited by the pace of human interaction, the upper bound on demand would be determined mainly by authorization limits, infrastructure capacity, and capital expenditure budgets.
Supply, in Nomura’s view, cannot keep up. The bank expects industry capacity expansion to grow at only 30% to 40% annually on a compound basis. For CXMT, it estimates bit output expansion of around 40% to 45% a year from 2026 to 2030. That is above the industry average, but still below the pace of demand growth.
At the same time Nomura released its report, major Korean and U.S. chip players were moving into a new round of cooperation. On July 24 U.S. time, Samsung Electronics and Broadcom signed a memorandum of cooperation worth more than $200 billion, covering HBM4 high-end memory supply, 2-nanometer foundry manufacturing, and advanced packaging. During the same period, SK Group and SK Hynix reached a long-term cooperation intention with Nvidia worth more than $500 billion, focused on joint development and stable supply of next-generation HBM.
In this framework, the more top players focus on HBM, the larger the supply gap left in traditional DRAM. Nomura therefore expects CXMT’s share of the global DRAM market to rise from about 10% currently to 18% by the end of 2028.
The domestic market is another part of the story. WSTS data cited in the article shows China accounted for about 25% of global DRAM consumption in 2025, but domestic manufacturers represented only about 10% of global share by production revenue, implying self-sufficiency of roughly 30%. On that basis alone, the room for substitution inside China remains large.
Export controls and memory cyclicality remain the main risks
Nomura spent substantial space on downside scenarios as well. The first is the risk that the U.S. MATCH Act and related restrictions could block access to key equipment and materials, including lithography tools, etching tools, and advanced photoresists. Under the bank’s worst-case scenario, CXMT’s 2027 and 2028 revenue would fall by about 13% to 14%, while net profit would decline by about 30% to 33%, in addition to constraints on capacity expansion.
Those external risks overlap heavily with issues already listed in the prospectus. CXMT referred to “geopolitical risks,” including the 1260H list, and to competition risk in a highly concentrated global market. The most important point, though, is the industry cycle itself: DRAM is highly cyclical, and earnings can swing sharply.
The sector is still in an upswing, but consensus around memory is beginning to crack. Micron, for example, hit a record high of $1,255 on June 25 and then fell to as low as $804 within three weeks, a maximum drawdown of 36%. SK Hynix saw an even steeper move. Its Korea-listed shares dropped from a June 25 high of 2.987 million won to 1.678 million won, a drawdown of 43.8%. Its ADR jumped 27% in a single day on July 14, then fell 9% the next day.
The investor who follows China’s semiconductor sector told Tencent Technology that the cycle always comes back to basic supply and demand. “As long as the market is euphoric, expansion can become unlimited. Demand, on the other hand, can stop abruptly at a certain point of adoption. No market is exempt. Memory is just a more extreme case.”
That investor also cited Huali in the 1990s, when it invested in memory near the top of the internet bubble, endured a violent boom-and-bust cycle, and ultimately shifted to logic chip foundry work. The point was to underline the force of memory cyclicality. According to the same investor, few other markets see prices rise 5x to 10x in one to two years during a boom and then lose 90% in one to two years after the break. Since 1990, that pattern has happened at least five times, roughly once every six years.
For CXMT, the long-term test may be whether it can keep financial discipline during the upcycle and avoid overly aggressive expansion.