Can the AI trend propel ChangXin Memory Technologies to a trillion-yuan market value?

How long does it take to transform from a "money shredder" into a "money printer"?

It might take about half a year.

Changxin Technology, the largest domestic DRAM storage manufacturer, submitted its prospectus to the STAR Market at the end of 2025. The rate at which it burned money was alarming. The performance showed that the company had a net loss of 5.98 billion yuan in the first nine months of 2025.

Due to the expiration of its financial data, Changxin was put on hold by the STAR Market. Coincidentally, during these months, there was a super price surge in storage chips ignited by AI. In the first quarter of 2026, the contract price of DRAM soared by 93% - 98% quarter - on - quarter, and NAND increased by 55% - 60%, reaching a 15 - year high.

When Changxin reappeared in the capital market, the updated prospectus showed that the company's net profit in the first quarter of 2026 was close to 25 billion yuan, which means it was making a net profit of 270 million yuan per day on average, truly becoming a "money printer".

However, if we take a long - term view, it actually took Changxin ten years to reach this harvest period.

Now, Changxin stands at the door of the STAR Market and also faces a bigger test: Can China's storage chips really break the monopoly of Samsung, SK Hynix, and Micron?

The answer is far from as easy as making 270 million yuan a day. There are issues such as the process generation gap, the non - mass production of HBM3, and the core technologies still being held by giants. Whether Changxin's "money printer" can keep running depends on whether it can win the race against time in the next more brutal technology competition.

The Big Bet Behind a Decade of Money - Burning

‍The story of Changxin Storage starts with its founder, Zhu Yiming.

Zhu Yiming was born in Yancheng, Jiangsu in 1972. After completing his undergraduate and master's degrees at Tsinghua University, he went to the United States to pursue a dual master's degree in physics and electronic engineering. After graduating in 2000, he stayed in Silicon Valley and worked for several semiconductor companies successively.

In the summer of 2004, he presented his entrepreneurial idea to Li Jun, the initiator of the Tsinghua Entrepreneurs Association, in a café in Silicon Valley. He planned to build the largest memory design and manufacturing enterprise in China. This idea was supported by Li Jun, which led to the first round of investment. In the same year, Zhu Yiming founded GigaDevice Semiconductor Inc. in the United States.

In 2005, he resigned from his position in Silicon Valley and returned to China to found Beijing Xinjijiaoyi Technology, starting with Nor Flash storage chips. In 2010, the company was renamed Zhaoyi Innovation and went public on the A - share market in 2016, with its market value once exceeding 100 billion yuan.

In this year, Zhu Yiming turned his attention to DRAM. However, this is an extremely money - burning game. Building a wafer factory, researching and developing advanced processes, and breaking through technological blockades all require tens of billions of investment.

Finally, with the support of the Hefei municipal government, Zhu Yiming founded Changxin Technology. In 2018, at the critical node when DRAM products were about to enter mass production, Zhu Yiming resigned from the position of general manager of Zhaoyi Innovation, only retaining the position of chairman. He devoted himself entirely to Changxin Storage and served as the CEO. At the same time, he promised not to receive any salary or bonus until the company achieved profitability.

In 2019, Changxin's DRAM factory was officially put into production, marking the first step in the domestic production of DRAM in China.

Looking back at the development of Changxin Storage, it clearly features "state - owned capital setting the stage and entrepreneurs performing on it", and the same is true for its equity structure. Although the prospectus shows that the company has no controlling shareholder or actual controller, state - owned capital plays an important role in the company. The largest shareholder, Qinghui Jidian, is ultimately dominated by Hefei state - owned assets after penetration. In addition, other state - owned assets directly hold more than 36% of the shares. Zhu Yiming indirectly holds shares through channels such as Qinghui Jidian.

Data source: Changxin Technology's prospectus, Tianyancha

The state - owned capital mainly comes from local capital in Hefei.

Hefei became the location for Changxin Storage because of the local government's long - term "investment - driven production" strategy. When Changxin decided to build a factory, Hefei Industrial Investment, as the core partner of Qinghui Jidian, took on the risk of heavy - asset investment, while Zhu Yiming's team focused on technology R & D and enterprise operation.

Changxin Technology plans to raise 29.5 billion yuan this time, and the proportion of the total share capital after the issuance will be no less than 10%. Based on this, the estimated post - issuance valuation is about 295 billion yuan. Among them, 7.5 billion yuan will be used for the technical transformation of the wafer manufacturing mass - production line, 13 billion yuan for DRAM technology upgrading, and 9 billion yuan for forward - looking technology R & D. This fundraising scale is second only to SMIC in the history of the STAR Market.

Recently, the A - share storage sector has risen collectively. As of May 20 this year, the Wind Memory Index has increased by 71.1%. The two memory leaders, Deminli and Jiangbolong, have increased by 200.4% and 133.4% respectively since the beginning of this year. On May 20, the A - share sector continued to strengthen and became the leading force in the semiconductor track on that day.

Referring to the market enthusiasm and the performance of new semiconductor stocks on the STAR Market, the market has an expectation of a trillion - yuan market value for Changxin.

On December 30, 2025, Changxin Technology submitted its prospectus. In March, due to the expiration of the financial data, the IPO review was suspended and could only be resumed after the updated performance data was supplemented. On May 17, the company released the latest prospectus. These more than four months were a period of drastic changes in the storage industry.

During this period, driven by the explosive demand for AI servers, the prices of storage chips climbed rapidly.

In the first quarter of 2026, the contract price of DRAM increased by more than 90% quarter - on - quarter. The demand for DRAM in a single AI server is 8 to 10 times that of a traditional server. Leading cloud providers and AI companies locked in a large amount of production capacity in advance, resulting in a serious shortage of supply of general - purpose storage chips. At the same time, storage giants shifted their production capacity to higher - profit products to meet AI demand, which also made the traditional DRAM in short supply. Changxin accurately seized the opportunity.

This also affected Changxin's revenue and profit situation in the first quarter. The prospectus shows that Changxin Storage's revenue in Q1 2026 was 50.8 billion yuan, a year - on - year increase of 719.13%; the net profit attributable to the parent company was 24.762 billion yuan, achieving net profit again after Q4 2025.

In the past two years, Changxin Storage's revenue has also maintained a year - on - year growth of more than 150%.

In fiscal year 2025, the revenue reached 61.8 billion yuan, of which 66.43% came from LPDDR products (i.e., the cache of mobile phones, tablets, and cars), and 31.87% came from DDR products (i.e., the cache of computers or servers).

It is worth noting that the growth of the server business is mainly affected by AI.

According to Omdia data, in the global DRAM market in 2025, server demand accounted for about 50%. The expansion of cloud computing and data centers behind it is the core driving force for the growth of the DRAM market, especially the demand for AI computing power.

The AI Storm Behind Earning 270 Million Yuan a Day

The demand for storage from AI is mainly divided into two levels: the bottom layer is NAND Flash/SSD, which is like the "hard drive" of a computer or the "storage space" of a mobile phone. It is responsible for storing data for a long time and providing "ammunition" for training and inference; the upper layer is DRAM, which is like the "memory" of a computer. It is a "short - term memory" that will be lost once the power is cut off. When the chip is working, data must be loaded into DRAM first to be processed.

Among them, HBM (High - Bandwidth Memory) is a special form of DRAM. It integrates the memory directly beside the GPU and determines the data scale for large - model training.

Changxin Storage mainly produces DRAM and is far ahead in the domestic market. In terms of the customer base, Changxin Storage's users include major domestic terminal and cloud - computing manufacturers such as Alibaba Cloud, ByteDance, Tencent, Lenovo, Xiaomi, OPPO, and vivo. In 2025, the revenue from the top five customers accounted for 68.08%.

However, compared with international giants, Changxin Storage still has a gap. According to Omdia's estimate, in the global DRAM market in 2024, the combined market share of Samsung, SK Hynix, and Micron exceeded 90%. Although Changxin Storage ranks first in China, its market share is only about 3% - 5%, and it increased to about 7.6% in the fourth quarter of 2025.

In the high - end enterprise - level AI server market, it currently mainly plays the role of a "domestic backup" rather than the first choice in terms of performance. In terms of the process, Changxin Storage has achieved large - scale mass production of 17nm DRAM, with a gap of about two generations compared with international giants (1β/1γnm, about 12 - 13nm).

Advanced - process DRAM is a prerequisite for producing HBM. In the storage cost structure of AI servers, HBM accounts for more than 40%, and it is also the most bottleneck part of current AI computing. Currently, the global HBM market is completely monopolized by SK Hynix, Samsung, and Micron, and the HBM4 production capacity in 2026 has been booked until 2028, with a supply gap continuously exceeding 10%.

Chinese enterprises have no experience in mass - producing HBM. Changxin Storage currently only has the process node level to manufacture HBM2, with a significant generational gap compared with HBM3E/HBM4.

Technologically, manufacturing HBM involves vertically stacking multiple DRAM chips (below 1βnm), connecting them with copper wires (TSV) and then welding, and finally encapsulating the whole beside the GPU. The loss rate in this process is extremely high. The stacking yield of SK Hynix's 12 - layer HBM3E is only about 75%, which means that one out of every four wafers is scrapped. Moreover, producing 1 bit of HBM consumes about three times the wafer area of DDR5.

This means that for every additional HBM wafer produced, three to four DRAM wafers will be produced less. The more vigorous the demand for HBM, the tighter the supply of DRAM. There is a paradox here: for the three storage giants, the gross profit margin of HBM is much higher than that of standard DRAM. Therefore, they will inevitably tilt their limited advanced - process production capacity towards HBM. Then the production capacity of general - purpose DRAM and NAND will be squeezed, and the prices are more likely to rise due to scarcity.

AI servers need both HBM and a large amount of standard DDR5 as the main memory, as well as a large amount of NAND to store data. The triple demand is superimposed, while the supply side is in a situation of ebb and flow, and there is always one layer in short supply, especially with the further popularization of AI.

This also explains the recent sharp rise in the stock prices of NAND manufacturers such as SanDisk and Western Digital. The price of enterprise - level SSDs (i.e., solid - state drives, the finished products of NAND chips) increased by more than 40% in the first quarter, and the price of consumer - level 1TB SSDs rose from about $45 at the end of 2025 to nearly $90.

When the data flood generated by AI training requires a large amount of flash memory to receive, and storage manufacturers shift their NAND production capacity to higher - profit enterprise - level products, NAND also becomes scarce.

The most well - known NAND enterprise in China is Changjiang Storage, which is called the "two domestic storage powerhouses" together with Changxin. It is expected to reach a 15% market share in the global NAND market this year and recently started IPO counseling.

How Long Can the "Money Printer" Keep Running?

Of course, the demand for HBM cannot be ignored because solving the problem of HBM is crucial for the localization of large models. This is difficult to achieve independently by a single enterprise. The support of national industrial policies, the breakthrough of enterprise chip processes, and the improvement of industry - wide compatibility are all indispensable.

However, the time window is extremely urgent. The mass production of domestic HBM may not be achieved until 2028 - 2030. During this period, China's AI computing infrastructure will still have to rely on imported HBM.

For Changxin, the more fundamental problem lies in the storage industry itself. The storage industry is one of the most capital - intensive industries. The time and cost required to set up production lines before production and sales are extremely high, and it is highly cyclical, with shortages and surpluses occurring repeatedly.

The price of DRAM has experienced many sharp rises and falls in history. For example, in 1993, the popularization of computer graphical interfaces led to a sharp increase in memory demand and price, but then the memory price dropped quickly due to too many newly built production lines; in 2010, the popularization of smartphones brought a short - term price increase in storage; in 2020, the epidemic drove the demand for storage again, and after the wave receded, there was an inventory backlog, resulting in a sharp drop in storage prices in 2022.

This round of storage price increase started in 2025, driven by AI.

Generally speaking, the storage industry experiences a cycle of "explosive demand - shortage - price surge - aggressive expansion - oversupply - price collapse" every few years due to new demands. Changxin's financial data illustrates this point. Since 2023, the company's gross profit margin and net profit have been in a state of drastic change. The improved profit situation in 2025 is closely related to the growth of AI demand.

However, the demand for AI also has risks. At the end of March 2026, a paper