In Jinjiang, Fujian, a super unicorn in the storage industry is developing quietly.

In recent days, Changxin Memory Technologies based in Hefei debuted on the A-share market, and its closing price on the listing day made it the highest-priced stock on the A-share market, even surpassing Tencent to become the listed company with the highest market value in China for a period of time. Yangtze Memory Technologies Co., Ltd. (YMTC) based in Wuhan is also in the IPO counseling period.

In fact, in Jinjiang, Fujian, a small city famous for sports shoes, there is another top-tier memory company hidden — Fujian Jinhua Integrated Circuit Co., Ltd. (Fujian Jinhua). 

At one point, Fujian Jinhua was one of the three major memory chip bases alongside YMTC and Changxin Memory Technologies. Even in the early stage of its establishment, it once outpaced Changxin Memory Technologies and was the most promising domestic memory manufacturer recognized by the industry.

However, the development path of Fujian Jinhua was far from smooth. What arrived earlier than its products were the Entity List issued by the U.S. Department of Commerce and criminal charges filed by the U.S. Department of Justice. These drastic external changes instantly brought Fujian Jinhua's production line to a complete halt, and it almost disappeared from public view for the following five years. 

On February 27, 2024, a not-guilty ruling from the Federal Court of San Francisco, the United States, brought this long-silent company back to the spotlight. It has been exactly eight years since Jinhua was founded. Over the past eight years, the global DRAM market has also gone through cyclical shifts and entered an AI-driven super-prosperity cycle. Samsung and SK Hynix, needless to say, have already reaped huge profits.

On July 27, Changxin Memory Technologies became the highest-priced A-share stock, while Jinhua at that time had only been cleared of the charges for more than two years.

An unofficial data released last year showed that the valuation of Fujian Jinhua was about 80 billion yuan. Since it has not raised funds in the capital market for a long time, the accuracy of this valuation cannot be verified. However, although Fujian Jinhua's main business is DRAM, which are mostly customized DDR4 chips rather than the DDR5 and HBM that AI applications rely on, under the situation that Changxin Memory Technologies has set a positive example after its listing and memory chip prices continue to rise, Fujian Jinhua can at least share the benefits brought by this industry boom.

Dreams and Controversies

Just as the story of Changxin Memory Technologies cannot be told without mentioning Zhu Yiming, Jinhua also has a core figure — Chen Zhengkun.

He is low-profile and rarely gives interviews. There is not even a detailed verified resume of Chen Zhengkun available to the public.

Limited information shows that he was born after 1965. He showed great interest in semiconductors in his teenage years, studied successively at the Department of Physics of Tsing Hua University in Chinese Taiwan and the University of California, Berkeley. After graduation, Chen Zhengkun stayed in Silicon Valley and worked as a chip engineer. Later, he returned to Chinese Taiwan and joined an electronics company called Rexchip. In 2013, Rexchip was acquired by Micron, and Chen Zhengkun became the president of Micron's Taiwan subsidiary, which was responsible for manufacturing 25nm DRAM chips for Micron.

During his tenure at Micron, Chen Zhengkun increasingly realized a fact: no matter how high a management position a Chinese engineer could reach, he would never have access to the real core technology.

In 2015, he officially resigned from Micron and joined United Microelectronics Corporation (UMC).

On the other hand, on February 26, 2016, Fujian Jinhua Integrated Circuit Co., Ltd. was registered and established in Jinjiang City, with a registered capital of 34.477 billion yuan, which was jointly funded by provincial, municipal and county-level state-owned enterprises including Fujian Electronics Information Group, Quanzhou Financial Holding Group Co., Ltd. and Fujian Jinjiang Industrial Development Investment Group Co., Ltd. 

This is an enterprise born with a national mission. It was included in China's "13th Five-Year Plan" Major Productivity Layout Plan for Integrated Circuits, with only one goal: to make a breakthrough in the DRAM field and break the decades-long monopoly of three overseas giants, Samsung, SK Hynix and Micron. 

At that time, there were three memory projects across the country: Wuhan-based Yangtze Memory Technologies focused on NAND flash memory, Hefei Changxin independently developed DRAM, while Fujian Jinhua planned to develop niche DRAM through cooperation with UMC from Chinese Taiwan. 

Therefore, UMC and Fujian Jinhua signed a technical cooperation agreement: Jinhua was responsible for investing funds to build production lines and purchase equipment, while UMC dispatched a core technical team to jointly develop 32nm process DRAM technology. Jinhua provided 300 million US dollars for purchasing R&D equipment, and paid 400 million US dollars to UMC in stages according to the progress of the project. The technical achievements developed through the cooperation are jointly owned by both parties.

In February 2017, Chen Zhengkun took the position of General Manager of Jinhua. According to reports from multiple media outlets, Chen Zhengkun hardly hesitated when he received the invitation. It is worth noting that what he gave up was not only the high annual salary at Micron. At that time, Jinhua did not even have a decent R&D building, and its temporary office was set in an idle factory building next to the industrial park, with a construction site under construction outside the window. 

When asked why he joined the DRAM technology cooperative R&D program between UMC and the Chinese mainland, Chen Zhengkun said that the acquisition of Rexchip by Micron had a huge impact on him back then, and independently developing DRAM technology had always been his long-cherished dream.

With Chen Zhengkun in charge, the cooperation between the two sides advanced faster than expected, and the plant construction was completed in less than two years.

According to the plan, the total investment of the first-phase project is 5.3 billion US dollars, which is scheduled to be officially put into production in the third quarter of 2018. The monthly production capacity of 12-inch wafers with 32nm process will reach 60,000 pieces. The final goal is to launch 20nm products, and the monthly capacity will reach 240,000 pieces after the completion of the fourth phase in 2025.

Everything seemed to be going on the right track.

Then the storm came. In September 2017, Micron sued UMC in Chinese Taiwan, alleging that employees who moved from Micron to UMC stole Micron's DRAM trade secrets and were suspected of helping UMC develop 32nm DRAM. In December of the same year, Micron sued Jinhua and UMC in the Federal Court of California, the United States. Jinhua quickly fought back by countersuing that Micron's products sold in China infringed its patents, and finally won the lawsuit.

What no one expected was that all this was just a prelude. At the end of October 2018, the U.S. Department of Commerce announced that it would add Fujian Jinhua to the export control Entity List on the grounds of national security. This means that not only U.S. enterprises are prohibited from doing business with Jinhua, but also global suppliers that use U.S. technology are forbidden to supply goods to Jinhua. More seriously, a newly completed production line of Jinhua that relied on the global supply chain was abruptly cut off by the ban. 

Only one month later, the U.S. Department of Justice formally prosecuted Fujian Jinhua, UMC, Chen Zhengkun and other relevant personnel, upgrading the original civil dispute to a criminal case, with charges including conspiracy to commit economic espionage.

In October 2020, UMC chose to compromise and reached a settlement with the U.S. Department of Justice, admitting that it had infringed one trade secret, agreeing to pay a fine of 60 million US dollars, and promising to cooperate with the U.S. government's investigation of Jinhua during the three-year probation period. Two months later, the U.S. court issued a global arrest warrant for Chen Zhengkun.

However, since neither the Chinese mainland nor Chinese Taiwan has signed an extradition treaty with the United States, Chen Zhengkun has always stayed in China to continue leading the R&D work.

The turning point came at the end of 2023, when Micron announced that it had reached a global settlement agreement with Fujian Jinhua. Both sides withdrew their lawsuits against each other around the world and terminated all litigations. Two months later, on February 27, 2024, the Federal Court of San Francisco, the United States ruled that the prosecution failed to prove that Jinhua misappropriated Micron's proprietary data, all charges were dismissed, and Fujian Jinhua was fully cleared of accusations.

Although it finally obtained a not-guilty outcome, it must be admitted that these six years of charges have seriously slowed down Jinhua's progress. Meanwhile, in January 2025, Chen Zhengkun stepped down as General Manager of Jinhua due to the expiration of his employment contract and turned to be the company's technical consultant, when he was already over 60 years old.

The Most Challenging Chip to Manufacture

To understand the situation Jinhua is facing, it is necessary to first understand the DRAM industry. 

The essence of DRAM is a super-large array of memory cells. Each cell contains one transistor and one capacitor. The transistor can be made smaller and smaller, but it is extremely difficult to shrink the capacitor. Once the capacitor becomes too small, it cannot hold the charge stably, and the charge is prone to leakage or interference from adjacent components. 

In order to make DRAM more efficient, manufacturers can only try various increasingly complex structures, which makes the manufacturing process extremely costly. Nowadays, building the most advanced DRAM wafer fab often requires an investment of 15 to 20 billion US dollars, not to mention the billions of dollars needed to purchase equipment such as lithography machines and etching machines. 

More critically, DRAM is a commodity, which means that it follows a unified set of industry standards, and chips manufactured by any manufacturer can be directly installed into the equipment of any other manufacturer.

Looking back at the development history of DRAM, it is a series of life-and-death cycles from prosperity to recession.

For example, Intel dominated the market in the 1970s, with a market share as high as 82.9% at one point. In the 1980s, Japan's NEC, Hitachi and Toshiba staged a counterattack with low prices, and their combined market share exceeded 50% in 1985, leading to a total defeat of U.S. enterprises in this field. In the same period, South Korea's Samsung acquired Korean Semiconductor and officially entered the DRAM market in 1983. 

Another example is that after 1990, the industry entered a downward cycle. In the harsh winter when the whole industry suffered losses and DRAM prices plummeted, Samsung instead frantically raised debts to expand production, and pushed the cost below the cash flow depletion line of its competitors through the scale effect. In 1998, the combined market share of South Korean enterprises surpassed that of Japan for the first time. In 2001, Hyundai was split and renamed SK Hynix, while Micron gradually grew stronger through continuous mergers and acquisitions.

When the time came to the 2008 financial crisis, the memory industry ushered in a major reshuffle again. Germany's Qimonda and Japan's Elpida declared bankruptcy one after another, which marked the successive exit of European and Japanese enterprises from the market. After that, the global DRAM market finally formed an oligopoly pattern dominated by three giants: Samsung, SK Hynix and Micron, with a combined market share of over 95%.

The three surviving giants have reached a subtle tacit understanding as oligarchs, no longer easily launching full-scale price wars, and the focus of competition has shifted to technical positioning. For example, HBM, which has been very popular in the past two years, not only involves process shrinking, but also tests advanced stacking and packaging technologies, including 3D stacking and Through-Silicon Via (TSV) technology. At present, Samsung and SK Hynix have established technical barriers in these cutting-edge fields.

Looking at the domestic market, when Qimonda went bankrupt, local state-owned capital in China once tried to acquire all its assets, but under the intervention and game of various internal and external forces, the process lasted for several years and finally failed to complete the acquisition.

Another reality is that around 2015, chips surpassed crude oil procurement for the first time and became China's largest imported commodity. At that time, when China needed to import memory chips, the price was completely dominated by South Korean and American manufacturers, and the procurement was also subject to quota restrictions. The domestic memory chip production was almost zero, and there was no autonomy in the entire industrial chain.

The reason why Fujian Jinhua attracted Micron's attention is that on the one hand, many top technical experts and chip engineers joined Jinhua, and on the other hand, more than 40% of Micron's total revenue of 12.4 billion US dollars in 2016 came from memory procurement in the Chinese mainland. DRAM and NAND flash memory together contributed more than 90% of the company's total sales. It is no exaggeration to say that the Chinese market has become the fundamental support for Micron's performance.

The highly single market dependence also laid a commercial motivation for Micron to launch cross-border intellectual property litigation and promote the Entity List to restrict Jinhua later.

In contrast, Changxin Memory Technologies also recruited many employees from Samsung in the early stage, and some of the related cases triggered judicial intervention in South Korea. But at the same time, it also spent huge sums of money to acquire Qimonda's patent portfolio, and signed a patent license agreement with U.S.-based Rambus, obtaining the implementation license of a large number of DRAM technology patents.

Although the purchase price seemed high at that time, it successfully avoided the technical patent risks from Micron, Samsung and SK Hynix. Later, Zhu Yiming revealed in 2019 that Changxin Memory Technologies modified part of the technologies to completely remove U.S.-related technical components.

Reconstruction in Adversity

After being added to the Entity List, Jinhua faced an almost impossible task: to re-establish an operable DRAM production line without the supply of U.S. equipment and materials. 

Here we have to mention a past experience of Chen Zhengkun.

During the 2008 financial crisis, Chen Zhengkun served as the General Manager of Rexchip Electronics in Chinese Taiwan. At that time, the company had in-depth cooperation with Micron and owned a DRAM production line with advanced technology. However, the yield of wafers produced by this production line dropped to just over 60%, which meant that the company lost money on every wafer produced. The senior management was already discussing whether to shut down this production line directly and sell the equipment as scrap.

Chen Zhengkun led a technical team to stay on that production line for four months, re-optimized various process parameters, and finally raised the yield to more than 80%.

The production line that used to lose 3 million US dollars per month began to make a profit of 2 million US dollars per month, which helped the company get through its most difficult period. Rexchip was later acquired by Micron, and this experience also became a legendary part of his career. Inside the company, his name has since been associated with the nickname "Yield Magician".

There are rumors that nearly 70% of Jinhua's original equipment relied on U.S. supply. After being added to the Entity List, Chen Zhengkun led the team to boldly transform domestic equipment on the front line, and then reconstructed the entire production process logic. This is a long and painful process. However, it is with this tenacity that Jinhua has greatly reduced the proportion of U.S. technology in its equipment.

There is a page on Jinhua's official website that clearly records "Jinhua Memorabilia", where we cannot see noisy controversies and disputes, but only product progress and revenue breakthroughs.

Behind these figures is the slow recovery of an enterprise under extreme pressure.

Of course, these data still have a large gap compared with Changxin Memory Technologies and YMTC, which were established in the same period. Changxin Storage achieved mass production of 19nm DRAM in 2019, and its revenue in the first quarter of 2025 has reached 50.8 billion yuan; YMTC took the lead in achieving mass production of 232-layer 3D NAND in 2022, and its market share is expected to reach 9% in 2025.

Fortunately, Jinhua was only slowed down, not eliminated from the market. In addition, Jinhua is deeply engaged in the niche DRAM market, and its products are widely used in smart TVs, set-top boxes, printers, routers, industrial control equipment and other fields. The product life cycle of such products is generally as long as 5 to 10 years, and once they enter the supply chain, they have extremely high stability.

At present, the monthly production capacity of Jinhua's 12-inch wafer fab