Author: ICVIEWS Editorial Department
A historic moment.
Just now, Changxin Memory Technologies (CXMT) successfully went public on the STAR Market. With an issue price of 8.66 RMB per share, the stock surged 460% at the opening, reaching a peak price of approximately 49.88 RMB per share, and pushing its market capitalization to over 3.2 trillion RMB.
Notably, CXMT raised a total of 57.919 billion RMB in this IPO (before the exercise of the over-allotment option), surpassing the 53.2 billion RMB raised by SMIC (Semiconductor Manufacturing International Corporation) in 2020, setting a new record for IPO fundraising on the STAR Market. Its A-share fundraising amount even ranks in the top five in history, second only to the Agricultural Bank of China, PetroChina, China Shenhua, and China Construction Bank, making it the largest technology stock IPO in the history of the A-share market.
CXMT's listing marks the transition of China's DRAM industry from the stage of "whether we have it" to "whether we can make it stronger."
01. Crossing the Death Valley: From Billions in Losses to Nearly 270 Million in Daily Profits
Opening CXMT's prospectus, the most intuitive impression is a set of nearly steep growth curves. From 2023 to 2025, CXMT's revenue surged from 9.087 billion RMB to 61.799 billion RMB, with the revenue scale growing nearly 7 times and a three-year compound annual growth rate of 160.02%. A fundamental reversal has been achieved on the profitability front. In 2023, the net loss attributable to the parent company was 16.34 billion RMB; in 2024, the net loss attributable to the parent company was 7.145 billion RMB. In 2025, the company achieved an annual net profit attributable to the parent company of 1.875 billion RMB for the first time, officially ending years of consecutive large-scale losses.
The real earnings explosion occurred in 2026. In the first quarter alone, revenue reached 50.8 billion RMB, a year-on-year increase of 719%; net profit attributable to the parent company was 24.762 billion RMB, translating to an average daily profit of approximately 275 million RMB based on 90 days. For the first half of this year, the expected operating revenue is between 110 billion RMB and 120 billion RMB, a year-on-year increase of 612.53% to 677.31%; net profit is expected to be between 66 billion RMB and 75 billion RMB, and net profit attributable to the parent company is between 50 billion RMB and 57 billion RMB. The consistently realized profitability data indicates that CXMT has crossed the most perilous "death valley" for DRAM manufacturers and entered a stable operational cycle.
The improvement in profitability is the result of multiple factors, including overseas memory giants scaling back conventional DRAM capacity, an upward industry cycle, full capacity utilization, and product structure upgrades, with product upgrades being a crucial driver. CXMT has completed a comprehensive transition from DDR4 to DDR5. In 2025, it launched the LPDDR5X product, with a maximum speed reaching 10667Mbps, benchmarking against the mainstream products of top-tier international manufacturers. The customer list can be described as an "all-star" lineup of China's tech sector: Alibaba Cloud, ByteDance, Tencent, Lenovo, Xiaomi, Transsion, Honor, OPPO, and vivo.
CXMT's revenue comes almost entirely from its two major memory product lines, DDR and LPDDR, which together account for approximately 99% of total revenue, serving as the company's primary source of income. Among them, relying on the massive market demand for smartphones and tablets, the LPDDR series has long maintained a high sales proportion, with continuously growing recognition among domestic terminal customers. At the end of 2024, DDR5 achieved mass production and rapidly ramped up volume, driving an increase in the revenue proportion of the DDR series products.
Reviewing CXMT's complete product iteration roadmap is also equivalent to retracing the entire journey of domestic DRAM from 0 to 1 and then to large-scale commercialization: product engineering tape-out began in 2018; in 2019, trial production of 8Gb DDR4 engineering samples was realized, marking the breakthrough of domestic DRAM from 0 to 1; in 2020, the LPDDR4X product was launched; in 2023, the LPDDR5 product was officially launched, making it the first domestic brand to launch independently researched, developed, and produced LPDDR5 products, which were applied to models from mainstream domestic smartphone manufacturers; in 2024, DDR5 products were launched; in 2025, a new generation of LPDDR5X product series was released, with the highest specification speed in the product line reaching 10667Mbps.
02. AI Overturns the Memory Table: CXMT Seizes the Positioning
From the latest DRAM market competitive landscape chart released by Counterpoint, it is evident that CXMT is rising as the "fourth pole" in the global memory industry.
Unlike in the past, CXMT is standing in a memory cycle that is completely different from traditional logic—AI is reshaping the DRAM industry. In the traditional DRAM cycle, the core driver is the inventory fluctuation of PCs, smartphones, and servers. When demand is strong, prices rise; when demand is weak, prices face downward pressure. Manufacturers' capacity expansion decisions often lag behind market changes, ultimately falling into a cycle of overcapacity and falling prices.
This script has been repeatedly played out over the past decades. But AI has introduced structural variables. First, the demand side has been redefined. AI servers pose unprecedented requirements for memory capacity, bandwidth, and energy efficiency. HBM (High Bandwidth Memory) has become the highest-value segment, while high-capacity RDIMMs are also an unavoidable configuration for data center expansion. Memory is no longer just "good enough," but rather the performance bottleneck of computing power systems.
Second, the supply side is being reshuffled. As Samsung, SK Hynix, and Micron allocate more capacity resources to high-profit products such as HBM and high-capacity server DRAM, the supply of conventional DRAM, mobile DRAM, and some DDR4/DDR5 naturally tends to tighten. This is not a cyclical supply-demand mismatch, but an active migration of capacity structure.
Third, the procurement logic of customers has undergone a qualitative change. Downstream manufacturers no longer solely pursue the lowest prices; supply security has been placed on a par with cost. Smartphone manufacturers, PC manufacturers, server manufacturers, and automotive electronics customers all know that memory shortages can directly lead to a halt in complete machine shipments—the lesson of the chip shortage has spread from logic chips to memory chips.
With the superposition of these three variables, DRAM is transforming from a "standard cyclical commodity" back into a "strategic resource." This is precisely CXMT's window of opportunity. Without needing to confront the big three head-on across the entire track, as long as CXMT continues to ramp up volume in the LPDDR, DDR5, and certain server and terminal markets, it can become a non-negligible new force in the global supply chain. For customers, having one more stable supplier means one more bargaining chip; for Chinese terminal manufacturers, having one more domestic original manufacturer means one more layer of supply security.
This is the true industrial value of CXMT: not to completely replace existing giants, but to push the DRAM market from a "tripod-like" three-way standoff to a balanced four-way game, enabling the global memory supply chain to break free from the highly oligopolistic concentration pattern and move towards diversification and controllability.
03. The Fourth Pole in DRAM: Where is the "Trump Card"?
For a long time, the global DRAM market has been controlled by Samsung, SK Hynix, and Micron, with a combined market share of over 90%. But this landscape is fracturing. While the big three are still engaged in close-quarters combat on the HBM track, CXMT has already torn open a gap in the mature process DRAM field, growing into a de facto fourth pole.
CXMT's trump card lies in a carefully designed "misaligned competition" strategy, namely the precise positioning of a differentiated product matrix and process strategy. CXMT has formed a full-spectrum matrix covering DDR4, DDR5, LPDDR4X, LPDDR5, and LPDDR5X. Among them, DDR5 targets the high-end PC and server markets, LPDDR5X aims at flagship smartphones, while DDR4 and LPDDR4X harvest the scale dividends in the mid-to-low-end markets. This combined strategy of "building brands at the high end, grabbing market share at the mid end, and securing cash flow at the low end" has allowed CXMT to find space for survival and growth in the gaps between the giants.
At the process level, CXMT is currently mass-producing its fourth-generation process platform (equivalent to the industry's 1Xnm to 1Ynm level). Compared to the 1α and 1β nodes already mass-produced by Samsung, SK Hynix, and Micron, although there is a generational gap, the particularity of DRAM lies in the fact that mature processes are not without a market. DDR4, LPDDR4X, and some mid-to-low-end DDR5 have relatively relaxed requirements for the process but are extremely sensitive to cost control. CXMT has seized exactly this point: using mature processes to produce standard DRAM with high yield and low cost, to meet the massive consumer electronics and server demand in China and emerging markets.
In this special memory cycle, all current capacity and production line layouts are a race to seize market share. CXMT operates three 12-inch wafer fabs in Hefei and Beijing, with all capacity coming from 12-inch wafer manufacturing production lines. Over the past three years, the capacity utilization rates were 87.06%, 92.46%, and 95.73% respectively. The production and sales of major DRAM products have achieved rapid growth, with the production-to-sales ratio calculated by capacity at 99.45%, 97.94%, and 90.67% respectively. Counterpoint points out that in the short term, CXMT will expand its monthly wafer capacity from the current 320,000 pieces to 420,000 pieces by 2027. In the long term, CXMT plans to double its capacity by 2030 and expand it to three times the current level by 2035.
There is another deeper trump card: CXMT's long-term accumulation in intellectual property. As of the end of last year, CXMT owned a total of 3,929 domestic patents (including 3,165 invention patents) and 3,043 overseas patents. According to statistics from the World Intellectual Property Organization, CXMT ranked 22nd globally in the number of international patent application publications in 2023. According to data published by the authoritative US patent service agency IFI, CXMT ranked 42nd globally in US patent grants in 2024, ranking fourth among all listed Chinese companies.
In the prospectus, we also saw a striking piece of data: CXMT's comprehensive gross profit margin in 2025 was 40.99%, catching up with or even surpassing Samsung Electronics and Micron. But risks are hidden behind these numbers—the leap in CXMT's gross profit margin is mainly due to reaping the dual dividends of "low cost of mature processes + cyclical price increases." The depreciation of the mass-produced 1X-1Ynm process has been basically amortized, while standard DRAMs such as DDR5 and LPDDR5X face tight supply and rising prices because the big three are shifting capacity towards HBM, forming a profit scissors difference. Once the cycle turns, or if the big three shift HBM capacity back to the standard market, CXMT's process disadvantage will quickly translate into a cost disadvantage, and the decline in gross profit margin may be faster than that of its competitors.
The deeper gap lies in the structure. SK Hynix's 60% gross profit margin is supported by the HBM premium, which is highly sustainable; CXMT's 40% gross profit margin is built on a single product portfolio of "no HBM, no NAND, no logic chips," making it extremely weak against cycles. Affected by export controls on advanced lithography equipment, CXMT cannot cut into high-end processes supported by EUV (Extreme Ultraviolet) Lithography in the short term. High-end products with high bandwidth such as HBM are still in the R&D and validation stage. There are still obvious shortcomings in the high-end computing power memory track, and industrial breakthroughs still face multiple challenges.
04. The Chain Master Effect: What is Truly Ignited is the Domestic Supply Chain
CXMT's listing is not only a financing behavior for a single enterprise but also a "stress test" for the industrial chain. The spillover value of its supply chain localization process is reshaping the domestic semiconductor supporting ecosystem. The true bottleneck of the domestic semiconductor supply chain has never been the inability to manufacture from technical drawings, but the lack of a domestic "chain master" that dares to introduce in batches, continuously validate, and steadily expand production on real production lines. CXMT is exactly such a chain master, and it is explicitly written in the prospectus that it aims to "give full play to the leading role of the industrial chain and promote the localization and diversification of the supply chain."
The operation of a DRAM production line is inseparable from a large number of core equipment such as etching, thin film deposition, cleaning, CMP (Chemical Mechanical Polishing), coater/developer, inspection and metrology, and testing. Of the 29.5 billion RMB raised by CXMT in this IPO, about 22.066 billion RMB will be spent on equipment procurement and installation. 2026-2027 will be the golden window for the introduction of domestic equipment. In CXMT's equipment supply chain, domestic manufacturers have already occupied key positions. NAURA Technology has supplied various equipment covering etching, thin film deposition, heat treatment, and cleaning. AMEC (Advanced Micro-Fabrication Equipment) has supplied plasma etching equipment, thin film deposition equipment, and other supporting equipment. In addition, Piotech's PECVD/ALD/SACVD equipment, ACM Research's cleaning equipment, Zhichun Technology's wet cleaning equipment, and HWATSING's CMP equipment are all applied to CXMT's production lines. These equipment no longer stop at laboratory validation but run through yield and stabilize capacity in real mass production environments—this is precisely the most critical leap for domestic equipment from being usable to being dared to be used.
Having climbed over the mountain of equipment, materials are the next checkpoint that must be conquered. DRAM manufacturing has extremely high requirements for materials such as silicon wafers, electronic specialty gases, wet electronic chemicals, precursors, polishing slurry, target materials, and photoresist. From the prospectus, we can see that the cost advantage brought by localization is no longer an expectation, but a realized fact: Based on 2023, by 2025, the procurement unit price of silicon wafers for CXMT dropped by about 30%, target materials dropped by about 22%, spare parts dropped by nearly 47%, and chemical procurement costs decreased by about 26%. Behind the numbers is the commercial closed loop of domestic materials completing validation on real production lines and securing batch orders. Manufacturers such as Yake Technology, Guanggang Gases, Jinhong Gas, Tongcheng New, and Jingrui Electronic Materials have all entered the CXMT supply chain.
From wafers to the hands of end users, memory chips also need to go through the relay of packaging and testing, module manufacturing, and brand channels. Huatian Technology is a clear packaging cooperative supplier for CXMT; DingLi has reached a procurement cooperation with CXMT; Microgate is still in the product introduction stage; currently, multiple distributors have obtained authorization from CXMT. With the continuous release of capacity, CXMT is opening up the complete internal cycle of wafer—packaging and testing—module—terminal.
And this fundraising of 29.5 billion RMB is precisely the specific landing point for the chain master to drive the in-depth development of the industrial chain. The funds will be invested in three major projects: technical upgrading and transformation of mass production lines for memory wafer manufacturing, DRAM memory technology upgrading, and forward-looking technology R&D for dynamic random access memory. A large amount of investment falls on equipment renewal, material validation, process upgrading, and R&D. For enterprises in all links of the industrial chain, this is a certain incremental opportunity.
05. Only Heroes Can Drive Away Tigers and Leopards, No Brave Men Fear Bears
CXMT's listing is just a beginning. As Zhu Yiming, Chairman of CXMT, stated during the STAR Market IPO roadshow: "Listing on the capital market is a new starting point, and more importantly, it means new responsibilities. The company hopes to further enhance its technological R&D capabilities and industrialization level through the raised funds investment projects, give full play to the driving role of a leading enterprise, promote the coordinated development of the industrial chain, and contribute greater strength to the construction of China's IC (Integrated Circuit) industry ecosystem."
From the "new starting point" to the "new responsibility," this refers not only to CXMT but also to the collective mission of China's semiconductor industry. The bell-ringing of CXMT is a milestone moment, but looking at the overall situation, it is just one piece of the puzzle in the rising map of China's semiconductor industry.
Over the past decade, this puzzle has been put into place piece by piece. In the memory field, CXMT has grown from zero to the fourth globally, and YMTC's (Yangtze Memory Technologies Co.) 3D NAND has stepped onto the global tier; in the manufacturing field, SMIC's foundry ranking is third globally; in the equipment field, NAURA Technology is the only domestic company among the global top five, and AMEC's etching equipment has entered the production lines of TSMC (Taiwan Semiconductor Manufacturing Company) and Samsung; in the design field, this year Huawei proposed "Tao's Law," breaking through process bottlenecks with architectural innovation and opening up new technical paths for China's chip design.
Behind China's semiconductor industry, the figures of a group of Tsinghua alumni have always been shining. Zhu Yiming locked his holdings for ten years, and with long-term persistence, he led CXMT into the first tier of global DRAM; Wang Hui steers ACM Research, with self-developed cleaning equipment running through all links of memory manufacturing; Yu Renrong founded Will Semiconductor, and Zhao Lixin founded GalaxyCore, achieving localized breakthroughs in image sensors; in the domestic GPU track, Tsinghua alumni such as Chen Weiliang, founder of Muxi Semiconductor, and Zhao Lidong, founder of Enflame Technology, have emerged. This group of Tsinghua alumni shares the same spiritual background: self-improvement and actions speaking louder than words.
Following CXMT, another name that has been continuously watched is YMTC. On the NAND Flash track, YMTC is also advancing silently. On the day it takes the stage, for the two most difficult memory battle lines of DRAM and NAND Flash, China will stand firm simultaneously for the first time.
Only heroes can drive away tigers and leopards; no brave men fear bears. CXMT's listing is a step on this Long March of China's semiconductor industry. The road is still long, but there are more and more Chinese people on the road.