On July 16, 2026, ChangXin Memory Technologies, Inc. (CXMT) officially launched its online and offline subscription on the STAR Market. The offering price was RMB 8.66 per share (approx. USD 1.28), corresponding to a market capitalization of about RMB 579.2 billion (approx. USD 85.4 billion), with an initial fundraising scale of RMB 57.919 billion (approx. USD 8.54 billion), rising to RMB 66.607 billion (approx. USD 9.82 billion) after full exercise of the over-allotment option. This scale surpassed the RMB 53.2 billion (approx. USD 7.85 billion) raised by SMIC in its 2020 IPO, making it the largest IPO by fundraising amount since the STAR Market was established. The core question drawing market attention is this: how could a company that posted losses for three consecutive years from 2022 to 2024, with cumulative losses exceeding RMB 30 billion (approx. USD 4.42 billion), list on the capital market at a valuation of nearly RMB 580 billion (approx. USD 85.5 billion) after turning around to profit in 2025? The factors supporting this valuation need to be examined from three dimensions: the DRAM industry cycle, the company’s competitive position, and its equity structure.
Industry Cycle: AI-Driven Structural Supply-Demand Imbalance
In the first quarter of 2026, global DRAM market revenue reached USD 97 billion, up 80% quarter-on-quarter and 260% year-on-year. DRAM contract prices rose 90% to 95% quarter-on-quarter in the first quarter, setting the largest quarterly increase in history. Research institutions within the industry expect contract prices to continue rising by about 50% in the second quarter, and conventional DRAM contract prices to climb a further 13% to 18% quarter-on-quarter in the third quarter.
The drivers of this round of price increases differ structurally from those in the past. Historically, cyclical fluctuations in the DRAM industry were mainly driven by changes in consumer electronics demand such as PCs and smartphones, with the supply-demand relationship showing a cyclical rotation of 3 to 4 years. The current price upturn, however, mainly stems from the accelerated construction of AI computing power infrastructure.
The training and inference processes of large language models require loading large volumes of parameters into memory, and the DRAM capacity per AI server unit is more than 10 times that of a traditional server. According to institutional forecasts, AI-related DRAM demand will exceed 53% of total demand in 2026, and the average DRAM capacity per server will jump from 1,032 GB in 2025 to 1,432 GB in 2026. At the same time, Samsung and SK Hynix have allocated 80% to 90% of their advanced capacity to high-bandwidth memory, while Micron has shifted about 70% toward HBM and high-end DDR5 products. The three major original manufacturers’ combined capital expenditure for 2026 is expected to reach USD 53.5 billion, but the new capacity is mainly concentrated in high-end product lines, and the supply of general-purpose DRAM has been systematically compressed.
The mismatch on both the supply and demand sides has formed a pattern of “structural tight balance.” Institutions judge that the current tight state of memory supply and demand will last at least until the second half of 2027, and UBS points out that the supply gap may persist until 2028. Samsung has stated that customers have already begun reserving 2027 capacity. These judgments constitute the market consensus on the current high prosperity of the DRAM industry.
But the DRAM industry has never in its history seen a cycle that only rises without falling. Some analyses hold that the second half of 2027 to early 2028 may become a phased peak of this memory cycle. If AI-related demand growth slows at that point while new capacity is released in a concentrated manner, the industry may re-enter a downward channel. This cyclical risk is a factor that cannot be ignored when assessing CXMT’s long-term value.
CXMT’s Market Position and Financial Performance
CXMT was founded in 2016 and is headquartered in Hefei; it is the only IDM enterprise in mainland China to achieve large-scale, integrated in-house design, manufacturing, and packaging-and-testing mass production of DRAM. The company owns three 12-inch wafer fabs in Hefei and Beijing, with monthly capacity of about 280,000 to 300,000 wafers as of early 2026, and capacity utilization stable at above 95%. Its workforce is nearly 20,000 people, with more than 6,259 R&D personnel, accounting for over 30%.
On the product side, in September 2019 CXMT launched an independently designed and produced 8Gb DDR4 product, achieving a breakthrough from zero for mainland China’s DRAM industry. It subsequently mass-produced mainstream series products such as DDR4, DDR5, LPDDR4X, and LPDDR5/5X. As of December 31, 2025, the company held 3,929 domestic patents and 3,043 overseas patents.
In terms of market share, according to industry-institution data, CXMT’s global DRAM market share rose from 3% in the first quarter of 2025 to 8% in the first quarter of 2026, ranking fourth globally by sales revenue. By comparison with the top three, in the first quarter of 2026 Samsung’s DRAM sales revenue was USD 38.2 billion, with a 40.5% market share; SK Hynix’s was USD 27.9 billion, with a 29.6% share; and Micron’s was USD 18.8 billion, with a 22.4% share.
In terms of financial data, from 2022 to 2024, CXMT recorded cumulative losses of about RMB 31.8 billion (approx. USD 4.69 billion) due to large-scale capacity construction and R&D investment. In 2025, benefiting from industry price increases, it achieved revenue of about RMB 61.799 billion (approx. USD 9.11 billion) and net profit of RMB 1.875 billion (approx. USD 277 million), realizing its first annual profit. Entering 2026, the scale of growth widened: first-quarter revenue was RMB 50.8 billion (approx. USD 7.49 billion), up 719% year-on-year, with net profit attributable to the parent of RMB 24.762 billion (approx. USD 3.65 billion); first-half revenue is projected at RMB 110 billion to 120 billion (approx. USD 16.2 billion to 17.7 billion), with net profit attributable to the parent of RMB 50 billion to 57 billion (approx. USD 7.37 billion to 8.41 billion). The half-year profit essentially covers the previously accumulated losses.
It should be noted that CXMT’s profitability is highly correlated with DRAM prices. If the industry cycle reverses, its revenue and profit may contract in tandem. In HBM, a key sub-segment of the AI era, CXMT’s HBM3 is still at the sample-delivery and customer-validation stage, with a 2-to-3-year gap behind internationally leading enterprises in stacking processes, yield, and high-end product iteration. This technology gap is difficult to eliminate in the short term.
Valuation Logic: A Dual Structure of Cyclical Pricing and Strategic Premium
CXMT’s offering price was RMB 8.66 per share (approx. USD 1.28), corresponding to a market capitalization of RMB 579.2 billion (approx. USD 85.4 billion). The static price-to-earnings ratio calculated on the basis of 2024 net profit is 308.92 times, significantly higher than the semiconductor industry’s average P/E of 76.32 times over the most recent month. If annualized on the basis of the lower limit of the forecast profit for the first half of 2026, the dynamic P/E is about 5 to 6 times.
By comparison with international peers, as of July 13, 2026, the forecast P/E ratios of Samsung Electronics, SK Hynix, and Micron Technology were 5.02 times, 5.64 times, and 8.2 times respectively, and their forecast price-to-book ratios were 2.22 times, 3.73 times, and 5.73 times respectively. CXMT’s dynamic P/E is essentially in the same range as its international peers.
The composition of the RMB 579.2 billion (approx. USD 85.4 billion) valuation contains two layers of logic. The first layer is the current profitability brought by the industry’s boom cycle — the high profit in the first half of 2026 provides immediate financial support for the valuation. The second layer is a strategic-position premium — as the only DRAM IDM manufacturer in mainland China, CXMT plays a specific role in the process of domestic substitution, and this factor is reflected in the pricing. The two together form the basis of the current valuation, but their respective weights adjust dynamically as the industry cycle changes.
Some market views hold that the prudent pricing given in the early stage of listing both avoids over-drawing on future expectations and leaves room for subsequent value growth. But it should be pointed out that the DRAM industry is strongly cyclical, and the current high prosperity may not necessarily be sustainable over the long term. If future AI demand falls short of expectations or new capacity is released in a concentrated manner, the industry may re-enter a downward cycle, at which point the valuation level will face repricing pressure.
Equity Structure: Strategic Synergy of Multiple Capital Parties
CXMT’s equity structure exhibits characteristics of multi-party checks and balances, with no controlling shareholder and no actual controller. The top five shareholders collectively hold about 58%, but no single shareholder can determine the selection of more than half of the board members.
The Hefei state-owned assets system is the shareholder camp with the highest shareholding ratio. Hefei Qinghui Jidian holds 21.67%, making it the largest shareholder; ChangXin Integrated (Changxin Jicheng) holds 11.71%; Hefei Jixin holds 3.27%; and Anhui Provincial Investment holds 7.91%. Hefei state-owned assets hold about 36.79% in total. The National Integrated Circuit Industry Investment Fund Phase II holds 8.73%. In addition, Alibaba-affiliated entities collectively hold close to 5%, and enterprises such as Tencent, Xiaomi, and Midea also hold small stakes. GigaDevice holds about 1.8%; that company was likewise founded by Zhu Yiming.
The formation of this equity structure is related to the DRAM industry’s asset-heavy, long-cycle investment attributes. State-owned-background shareholders provide long-term capital support, industrial capital forms upstream-downstream synergy, and market-oriented institutions participate in price discovery. But a dispersed equity structure may generate coordination costs when the company faces an industry-cycle reversal or needs to make rapid decisions.
In terms of strategic placement, 36 institutions subscribed for a combined amount of RMB 14.4 billion (approx. USD 2.12 billion). Enterprises in the semiconductor equipment, materials, and packaging-and-testing segments — such as AMEC, Montage Technology, Anji Microelectronics, Tongfu Microelectronics, and Piotech — participated in the subscription. Downstream mobile-phone manufacturers, cloud service providers, and smart-vehicle enterprises also appeared on the list. This “equity + supply chain” binding relationship provides CXMT with customer stability to a certain extent, and also reflects downstream enterprises’ demand for supply chain security.
Subsequent Challenges: Technology Catch-up, Capacity Expansion, and Cyclical Risk
The listing provides CXMT with a new financing channel, but the challenges the company faces have not decreased as a result.
In terms of technology catch-up, CXMT has achieved mass production of mainstream products such as DDR5 and LPDDR5X, and the gap with the industry’s mainstream level is gradually narrowing. But in the HBM field, the company has a generational gap of about 2 to 3 years with Samsung, SK Hynix, and Micron. HBM is the most core memory product category in current AI computing power demand, with a gross margin more than 3 times that of ordinary DDR memory. CXMT plans to mass-produce 12-layer HBM3E in 2027, and if this materializes on schedule, the technology gap is expected to narrow somewhat.
In terms of capacity expansion, industry institutions expect CXMT’s monthly capacity to climb to about 350,000 wafers by the end of 2026, approaching Micron’s level of about 385,000 wafers in the same period; the 2028 target is 500,000 wafers per month, accounting for about 17% of global DRAM supply. Ranked by wafer capacity, CXMT is expected to surpass Micron to become the world’s third-largest by the end of 2026. But capacity expansion requires sustained large-scale capital investment, and the yield ramp-up of new production lines takes time, so there is uncertainty in the process of converting capacity into effective output.
In terms of supply chain security, CXMT has clearly set out to raise the procurement proportion of domestic equipment and materials to above 45%, with the domestic-production rate of production-line equipment rising from 22% in 2023 to 31% in 2025, and expected to exceed 50% in 2027. But advanced equipment and some key materials still rely on external supply, and this constraint will persist over the medium and long term.
In terms of cyclical risk, the DRAM industry has historically experienced multiple rounds of alternating booms and downturns. CXMT’s consecutive losses before 2025 were precisely the result of an industry downturn cycle overlapping with a capacity ramp-up period. If the industry enters an adjustment after this boom’s peak, whether the company can maintain profitability and market share will depend on its cost-control level, customer stickiness, and speed of technology iteration.
CXMT Chairman Zhu Yiming stated during the IPO roadshow that the company will continue to adhere to independent innovation, accelerating process upgrades, expanding capacity, and enriching its product matrix; at the same time, he also cautioned that the DRAM industry is distinctly cyclical, and that if future AI demand falls short of expectations and new capacity is released in a concentrated manner, the industry may still re-enter a downward cycle.
CXMT completed the zero-to-one breakthrough for mainland China’s DRAM industry over the course of ten years. After listing, the proposition it faces shifts from “can it make DRAM” to “can it continuously operate and expand its market share in DRAM, the most fiercely globally competitive semiconductor sub-segment.” The subsequent evolution of this process depends both on the company’s own execution capability and on changes in the global DRAM industry’s supply-demand landscape and the geopolitical policy environment.
[Disclaimer]: The above content reflects analysis of publicly available information, expert insights, and BCC research. It does not constitute investment advice. BCC is not responsible for any losses resulting from reliance on the views expressed herein. Investors should exercise caution.