Author: Xie Zhifeng, Chip Trend Watch

On July 27, Changxin Technology officially landed on the Sci-Tech Innovation Board.

The opening numbers, closing market value, and fundraising scale became the key phrases that dominated the financial circle that day. Some calculated the increase, some assessed the valuation, some discussed how much liquidity it would withdraw, and some shouted the slogan of "a monumental victory for domestic storage."

But as someone who participated in the construction of Changxin and as an old friend of Zhu Yiming and Zhao Lun, when I watched the bell ringing, the numbers that came to my mind were not any figures, but a communication training session I had with Changxin storage executives at a resort in Hefei in 2018, where after the management training ended, everyone had the appearance of "we have to tackle this no matter how difficult it is."

When Zhu Yiming wrote a recommendation preface for my "Chip Matters" in 2018, he concluded, "At this new historical starting point, the timing is just right." What he referred to then was the popularization of integrated circuits and the training of industrial talents, and today, looking at the bell ringing for Changxin, it resembles a prophetic footnote — Chinese semiconductors require a deep and steady approach from one generation, or even two, and the past ten years of Changxin have solidly portrayed the phrase "deep and steady."

1. 2016, a gamble no one dared to take

Time rewinds to 2016.

At 44 years old, Zhu Yiming had just brought Zhaoyi Innovation to the A-shares, with a fortune worth hundreds of millions. Following the most conservative business path, he could comfortably oversee the basic platform of NOR Flash, operating a light asset, high-margin chip design company. But he had to choose a path that everyone thought was suicidal: a second entrepreneurial venture to make DRAM.

What is DRAM? It is the "workspace" for all electronic products; mobile phones, computers, servers, data centers all rely on it. However, since the technology's birth in the 1970s, the global market has been firmly grasped by Samsung, SK Hynix, and Micron, accounting for over 95% of the market share at its peak. China has attempted several breakthroughs in this field since the 1980s, all without success. A 12-inch DRAM production line requires an investment of hundreds of billions, with five years at least from development to mass production, and a failure rate exceeding 90%; the patent barriers are impenetrable.

I met Yiming in 2006 when he had just returned to China to start his business, while I was still in charge of sales in China for SMIC. After more than a decade of associations, I understood his character very well — he appears gentle, but at his core is a driven engineering student; once he sets his mind on something, no amount of persuasion can change it.

His desire to pursue DRAM was not impulsive. Between 2016 and 2017, the price of DDR4 memory more than tripled in a year, and the three giants jointly controlled supply for profits, leading to Xiaomi losing money on every unit sold due to memory shortages, while Huawei was forced to mix several specifications of flash memory in the same model, yielding performance differences of up to three times. In 2017, China spent $88.6 billion on imported memory chips, equivalent to paying the three giants $240 million a day as a "memory tax," and still often unable to buy enough even when they had the money.

Other industries are choked off, but storage is controlled at its neck.

However, ideals cannot fill the financial gaps of reality. He traversed through Beijing, Shanghai, and Shenzhen with plans, but no one dared to make a decision — hundreds of billions thrown in with no prospect of returns for five or six years, and the risk of failure was a burden no one could bear.

In the end, it was Hefei that caught him.

A closed-door meeting lasting three hours resulted in Hefei approving an investment of 14.4 billion, covering 80% of the initial total investment, labeled as the city's "number one project," with no pursuit of short-term returns, and with policies for land, taxes, and talent properly in place.

Many later referred to it as a "Hefei gamble," but I have never agreed with that description. Hefei is not gambling; it truly understands the industry. As early as 2013, Hefei released its semiconductor industry plan, eight months ahead of the national outline; in 2014, a 10 billion industrial fund was established, a month before the national big fund. Previously, it backed BOE, solving the "screen" issue; now it backs Changxin, filling the "chip" gap. The so-called "chip-screen-vehicle integration" has never been a slogan; it is a step-by-step improvement of the industrial map.

More crucially, Hefei's logic is: the government acts "last," bearing the earliest and most fatal risks while leaving returns and autonomy entirely for businesses and entrepreneurs. The 14.4 billion equaled about a quarter of Hefei's disposable income in 2016, averaging nearly 2000 yuan for each citizen, but Hefei never intended to interfere with operations or seek short-term returns.

I have specifically written about the Hefei model in "Chip Matters": many local governments treat semiconductors as performance projects, while Hefei regards semiconductors as the future of the city. This is why, despite the money being spent, the outcomes can be vastly different.

After the meeting that day, all I could say was, "The money has been secured, but the real tough battle has just begun."

2. Borrowed sparks, own kingdom

The technological starting point of Changxin cannot bypass Gigabyte.

There are always those who mock this by saying it is "built from scraps," but those who say this neither understand the rules of the industry nor the logic of technology.

Gigabyte was the world's second-largest DRAM manufacturer at the time, collapsing in a price war with Samsung in 2009, but when it fell, it held a more advanced BWL (Buried Word Line) architecture than the three giants — embedding word lines into the silicon chip successfully solved the leakage and interference issues of traditional stacked architectures, theoretically extending down to 10nm nodes. Before it went bankrupt, Gigabyte had successfully taped out a 46nm BWL process, while Samsung had still remained at 50nm and above.

In early 2016, Yiming obtained approximately 7000 patent licenses and the usage rights for 2.8TB of core technical documentation from Gigabyte through Canada’s WiLAN company, in a very stable manner, negotiating quietly and only disclosing to the public once finalized. It was not "picking up scraps," but rather a legally acquired solid technological foundation bought for real money — the pursuit of latecomer companies is never about reinventing the wheel from scratch; it is true wisdom to stand on the shoulders of giants to move further.

However, blueprints are dead, while people are alive. The core secrets of semiconductor manufacturing have always resided not in the documents but in the hidden knowledge held by engineers: adjusting a parameter by one degree or speeding up a process by ten seconds can lead to significant yield differences, these things cannot be put into patents and can only be conveyed by people.

Therefore, his second step was to recruit talent: bringing batches of core engineers from the Xi'an R&D center of Gigabyte to Hefei, along with former Vice President of Technology Kuster as a technical consultant. These individuals brought not just experience but also the engineering capability to turn "dead blueprints" into "living production lines."

After securing the technological foundation, Changxin made a crucial decision: to skip generations in development.

In Gigabyte's documents, the technical difficulties, yield bottlenecks, and cost curves for 38nm, 28nm, and 20nm nodes were clearly outlined. After assessment, the team believed that both 38nm and 28nm were transitional nodes with no commercial value, opting to skip directly to tackle 19nm. This decision saved at least two to three years.

Having successfully passed the technological obstacles was just the first step; mass production, stable yields, and customer willingness to use the products are critical survival lines.

Changxin did not rush into the branded market but chose the most pragmatic path: starting from the white brand market. Clone memory modules, set-top boxes, TV boxes, surveillance equipment... these areas are price-sensitive and do not demand extreme performance, providing Changxin with space for trial and error. While shipping goods to recover funds, they gathered feedback from real-use scenarios while refining processes and improving yields.

I wrote in "Chip Matters": "The first priority for semiconductor companies is to survive. First, occupy the low-end market, exchanging market share for data, data for yields, and yields for technology. This is the only feasible breakout path for latecomer companies."

This path may not sound glamorous, nor "high-end," but it is the only one late players can successfully navigate.

Thus, year by year, yield rates climbed from failing below standard to industry mainstream levels; products evolved from DDR4 to DDR5 and LPDDR5/5X; customers transitioned from white brand manufacturers to domestic PC brands and finally to mobile phone manufacturers, gaining entry into the server supply chains of Alibaba Cloud and Tencent Cloud.

The borrowed sparks ultimately kindled their own kingdom.

3. Technology is bone, organization is blood: The management lessons I witnessed

Around 2018, Zhu Yiming often had friends call me, inviting me to Hefei for some time to conduct several closed-door training sessions for Changxin’s core management team. "Xie Bo, you were one of the founders of SMIC; you should tell us how SMIC controlled costs initially."

Those training sessions were completely closed-door, with the core team monitoring production lines and timelines during the day and cramming into a conference room for classes at night. I shared my experiences from those years at SMIC, how we were extremely generous with equipment while being extremely frugal with ourselves: printing paper must be reused on both sides; if we could walk instead of riding the subway, we wouldn’t take the subway; if we could take the subway, we wouldn’t take a taxi. Every penny saved rolled back into the precision of the equipment. It wasn't about being strapped; it was serious. We knew that what we held wasn’t just a budget but the most precious spark of early industry.

This matter is rarely mentioned publicly, but I have always believed that paying attention to cost management was a crucial step for Changxin to arrive at today’s position.

4. Ten years to reach the table

By 2025, Changxin turned a profit; in the first quarter of 2026, revenue reached 50.8 billion, and net profit was 24.76 billion, nearly erasing the losses of the past eight years in one quarter.

Many exclaimed a "sudden outbreak," but to those of us who have watched it progress step by step, it is hardly sudden. It is the accumulation of ten years of technology, production capacity, and customers, perfectly coinciding with AI reshaping storage demand.

The essence of this round of DRAM market is that AI has transformed storage from a "consumer electronics accessory" into "computing power infrastructure." The three giants have directed all their best production capacity toward the more profitable HBM and high-end server memory, leaving a huge gap in general DRAM capacity. Changxin just happened to be at this juncture with production capacity, qualified products, and mature client approvals, naturally capturing the market vacated by the giants.

Reuters reported that Changxin signed a long-term supply agreement worth over thirty billion dollars with Tencent and is also negotiating deep cooperation with other domestic cloud vendors. This signal is more significant than profit numbers: Changxin is no longer a "backup option when goods are unavailable," but has become one of the core suppliers of China’s AI infrastructure.

From the perspective of the capital market, Changxin’s listing has also thrown an anchor into A-shares. Previously, investors wanting to position in domestic DRAM had to go around to buy shadow stocks of equipment, modules, and materials; now, the core asset has landed, and funds will naturally concentrate on companies with genuine orders and industry positions. Short-term liquidity disturbances and valuation restructures are inevitable, but in the long run, with Changxin as an anchor, the investment logic for domestic storage will only become more solid.

Of course, this is merely a note for the capital market; it has never been the primary目的of Changxin’s long march.

5. The bell ringing is not the end; it is the first step of the Long March

On the day of the listing, many said, "Congratulations, you finally made it."

But I felt instead that ringing the bell isn’t the end; it's a rite of passage. From this day forward, Changxin has truly stood at the global DRAM table and must face all the storms head-on.

I have listed the seven crises Changxin must face in "Chip Matters," which are still valid today: pressures of geopolitical controls, reversals in industry cycles, generational differentials in the high-end HBM market, choke points in core equipment and materials, risks of patent lawsuits, challenges of talent poaching, and the pressures of short-term evaluations from the capital market... each hurdle poses its own difficulties.

However, I hold confidence in Changxin, not because it is currently earning significant profits but because it has been on the right path since day one: respecting industry rules, respecting technological logic, and respecting long-term value.

More crucially, Changxin’s value has never been limited to itself. What it needs to do in the future is not to be an isolated island but to become the "leader" of the entire domestic semiconductor supply chain in China — transforming its immense production capacity into a testing ground for domestic equipment, materials, and components, providing the most precious opportunity for local companies like Northern Huachuang and Zhongwei to verify production increases.

Only when the entire ecosystem’s roots penetrate the local soil can Changxin’s supply chain security truly become as stable as a mountain; only when the entire industry chain strengthens can Chinese storage truly stand tall.

Conclusion

Ten years ago, no one believed that Chinese people could produce their own DRAM.

Ten years later, Changxin rings the bell for listing, firmly ranking fourth globally.

In these ten years, Fujian Jinhua has halted, Unigroup has collapsed, Wuhan Hongxin was a scam, and most players in the same field have fallen to failure. Only Changxin has walked step by step from blueprints to production lines, from white brands to servers, from losses to profits, ultimately reaching the spotlight of the bell ringing.

I have known Yiming for twenty years, watching him return from Silicon Valley to start a business, seeing Zhaoyi go public, watching him risk all his wealth to make DRAM, and seeing him transform from a spirited middle-aged man into a gray-haired veteran of the industry. He wrote in the preface to "Chip Matters," "History is a mirror," and he himself is the most vivid witness to the ten-year history of Chinese storage.

The semiconductor industry has never had miracles, only trajectories.

All seemingly sudden successes are built on the foundations of ten years of sitting on the cold bench and unmeasured investment.

The bell will fade away, the heat will recede, but Changxin's Long March has just completed its first step.

The Long March of China's semiconductors has only just begun to reach a point visible to the world.