Behind Changxin ringing the bell lies an ordeal nobody knows: staking an entire city, and burning through a lifetime’s renown

Author: Xie Zhifeng, Xinshi Observation

On July 27, Changxin Technology officially listed on the STAR Market.

The figures at market open, the closing market capitalization, and the fund-raising size became the buzzwords that swept through the financial world that day. Some people calculated the upside, others estimated valuations, some discussed how much liquidity it would drain, and others even chanted slogans like “a homegrown storage, epic-level victory.”

But as someone who witnessed the construction of Changxin firsthand—an old friend of Zhu Yiming and Zhao Lun—I watched the bell-ringing scene and what came to mind wasn’t any single number, but a training workshop I had in 2018 at a resort in Hefei, where I discussed with Changxin Storage’s executives. After the management training ended, everyone looked as if to say, “No matter how hard this is, we have to chew through it.”

When Zhu Yiming wrote a recommendation preface for my book “Xinshi” in 2018, he ended with: “At a new historical starting point, this could not be more timely.” Back then, he was referring to integrated circuit popular science and the cultivation of industry talent. But viewed at Changxin’s bell-ringing moment, it feels more like a footnote in the form of a prophecy—Chinese semiconductors require a whole generation, even two generations, of deep, quiet accumulation. And Changxin’s ten years have been precisely that: taking the four characters “deep, quiet accumulation” and sitting down steadily on a cold bench.

I. 2016: A bet nobody dared to take

Let’s roll the clock back to 2016.

At 44, Zhu Yiming had just taken Zhaoyi Innovation to A-share markets. He had a net worth of hundreds of millions, and along the most reliable commercial path, he could have stayed with the NOR Flash “core business,” building a light-asset, high-gross-margin chip design company—life could have been extremely comfortable. But he insisted on choosing a path that everyone thought was “suicide”: a second entrepreneurship, making DRAM.

What is DRAM? It is the “workbench” of all electronic products—phones, computers, servers, and data centers—no device can do without it. But since this technology emerged in the 1970s, the global market has been tightly held by just three companies: Samsung, SK hynix, and Micron. At its peak, together they consumed more than 95% of the shares. China started attacking the problem multiple times from the 1980s, but without success. A single 12-inch DRAM production line requires investments of more than tens of billions, and from R&D to mass production takes at least five years; the failure rate exceeds 90%. Patent barriers are also impenetrable.

I got to know Yiming in 2006. Back then, he had just returned home and started his business, while I was working at SMIC responsible for sales in China. After working together for more than a decade, I knew his temperament very well—he looks gentle, but at heart he is an extreme-axis engineering type. Once he decides on something, he has worked out the numbers, and not even ten oxen could pull him back.

He wasn’t making DRAM on a whim. From 2016 to 2017, DDR4 memory prices rose more than threefold in a year. The “three giants” joined forces to control supply and harvest. Because of memory shortages, even Xiaomi 6 had to sell one unit at a loss to sell a phone; Huawei phones were forced to mix different storage specifications. For the same model, performance could differ by as much as three times. In 2017, China imported storage chips for $88.6 billion—equivalent to paying the three giants “a $240 million memory tax” every day, and often there still wasn’t enough supply to buy.

Other industries get “choked at the neck,” but storage is held so tightly that even the neck is in their hands.

Still, the ideal couldn’t fill the real-world funding gaps. He took his proposal and ran around Beijing, Shanghai, and Shenzhen, but no one dared to make the call. With hundreds of billions invested and no way to see returns for five or six years, if the project failed, nobody could bear that responsibility.

In the end, what caught him was Hefei.

After a closed-door meeting lasting three hours, Hefei made the decision on the spot: to invest 14.4 billion, accounting for 80% of the total investment for Phase 1; to classify it as the city’s “No. 1 project.” It would not pursue short-term returns, nor interfere with day-to-day operations. Land, tax, and talent policies were all arranged accordingly.

Many people later talked about “Hefei gambling big,” but I never agreed with that framing. Hefei wasn’t gambling—it understood the industry. As early as 2013, Hefei issued a semiconductor industry plan, which came 8 months earlier than the national outline. In 2014, it set up a hundred-billion industrial fund, one month earlier than the national big fund. Before this, it bet on BOE, solving the “screen” problem. Now it bet on Changxin to fill the “chip” gap. The so-called “chips, screens, vehicles, and home appliances” is never just a slogan—it is a step-by-step industrial blueprint built piece by piece.

More importantly, Hefei’s logic was this: the government would be “junior,” shouldering the earliest and most fatal risks, while leaving all upside and autonomy to enterprises and entrepreneurs. 14.4 billion is about a quarter of Hefei’s disposable income in 2016—roughly nearly 2,000 per resident when spread out—but Hefei never intended to micromanage operations or chase short-term returns.

In “Xinshi,” I wrote specifically about the Hefei model: many local governments treat semiconductors as political performance projects, while Hefei treats semiconductors as the city’s future. That’s why the outcome is so different even though they all “throw money.”

After that meeting ended, the only thing one could say was: “The money is in hand, but the real hard battle is only just beginning.”

II. Borrowed spark, own kingdom

Changxin’s technical starting point is inseparable from Qimengda (Winbond?).

There are always people who mock it as “starting out by picking up scrap.” Those who say that don’t understand industry rules, nor do they understand the technical logic.

Qimengda was once the world’s second-largest DRAM manufacturer. In 2009 it collapsed in the price war with Samsung. But when it went under, it still held a more advanced BWL (buried word line) architecture than the three giants—burying the word lines inside the silicon wafer, fundamentally solving the leakage and interference problems of traditional stacked architectures. Theoretically, it could extend to nodes below 10nm. Before bankruptcy, Qimengda’s 46nm BWL process had already successfully taped out, while Samsung at the time was still stuck at 50nm and above.

In early 2016, Yiming legally obtained the right to use Qimengda’s roughly 7,000 patent authorizations and 2.8TB of core technical documentation through Canada’s WiLAN. He handled this extremely carefully; negotiations were kept low-key throughout, and only disclosed externally after the deal was implemented. This wasn’t “picking up scrap”—it was a legitimate technology base paid for in real money. For latecomers, catching up has never meant starting from zero to reinvent the wheel. The true wisdom is to stand on the shoulders of giants and move upward from there.

But blueprints are dead; people are alive. The core secrets of semiconductor manufacturing never lie in documents. They live in engineers’ tacit knowledge: if you adjust a certain parameter by one degree, or speed up a specific process step by ten seconds, yield can swing by more than a dozen points. These things can’t be written into patents; they can only be carried by people.

So his second step was to recruit people: he brought core engineers from Qimengda’s Xi’an R&D center to Hefei in batches. He also invited Qimengda’s former technical vice president, Kuster, to serve as a technical consultant. What these people brought wasn’t just experience—it was the engineering capability to turn “dead blueprints” into “living production lines.”

After securing the technical base, Changxin made an extremely critical decision: leapfrogging R&D generations.

In Qimengda’s documents, the technical difficulties, yield bottlenecks, and cost curves for each node—38nm, 28nm, and 20nm—were laid out clearly. After analyzing, the team believed that 38nm and 28nm were transitional nodes with no commercial value. So they skipped them and went straight to tackling 19nm. This step saved at least two to three years.

Getting the technology to “work” was only the first step. The life-or-death line was mass production, stable yields, and customers being willing to use it.

Changxin didn’t rush into branded markets at the very beginning. It chose the most pragmatic route: entering the white-label market. Counterfeit memory sticks, set-top boxes, TV boxes, and surveillance devices—these areas are price-sensitive and don’t demand extreme performance. That gave Changxin room for trial and error. It shipped products to generate cash flow, collected feedback from real scenarios, and at the same time refined processes and improved yields.

In “Xinshi,” I wrote: “The first priority for semiconductor companies is to stay alive. First take over low-end markets, use the market to obtain data, use data to achieve yields, and use yields to obtain technology. This is the only viable breakthrough path for latecomers.”

It doesn’t sound glamorous or “high-end,” but it is the only path that latecomers can truly walk.

So it ground forward year after year: yields climbed from failing grades to mainstream industry levels; products evolved from DDR4 to DDR5 and LPDDR5/5X; customers grew from white-label manufacturers to domestically branded PC companies, and then to smartphone makers, eventually entering the server supply chains of Alibaba Cloud and Tencent Cloud.

The borrowed spark ultimately burned into its own kingdom.

III. Technology is the bones; organization is the blood: management lessons I witnessed firsthand

Around 2018, Zhu Yiming would often have his friends call me and invite me to stay in Hefei for a period of time, to conduct several closed-door trainings for Changxin’s core management team. “Xie Bo, you’re one of the founders of SMIC. Back then, how did SMIC do cost control? You have to tell us.”

Those trainings were fully closed-door. The core team watched the production lines and chased schedule during the day, and at night squeezed into the conference room for lessons. I told them honestly how, in the early days when we started SMIC, we were extremely generous with equipment but extremely strict with ourselves. Printing paper was always reused on both sides. When going out for errands, if you could walk, you wouldn’t take the subway; if you could take the subway, you wouldn’t take a taxi. Every dollar saved went into the precision of the equipment. This wasn’t stinginess—it was seriousness. We knew that what we held wasn’t just a budget, but the most precious spark in the early stage of the entire industry.

This matter is rarely brought up by outsiders, but I have always believed that paying attention to cost management was a very key step for Changxin to reach where it is today.

IV. Ten years, waiting until the table

In 2025, Changxin returned to profitability. In the first quarter of 2026, revenue was 50.8 billion and net profit was 24.76 billion; in a single quarter, it almost erased the losses accumulated over the previous eight years.

Many people were shocked by “a sudden breakout,” but for those of us who watched it walk step by step, there was nothing sudden. It was ten years of accumulation in technology, capacity, and customer base—just when the AI-driven storage demand wave arrived.

The essence of this DRAM cycle is that AI has transformed storage from “consumer electronics accessories” into “computing infrastructure.” The three giants poured their best capacity into HBM and high-end server memory with higher profits, leaving a huge gap in general-purpose DRAM capacity. And at precisely this node, Changxin had capacity, qualified products, and mature customer certifications—so it naturally took over the market vacated by the giants.

Reuters reported that Changxin signed long-term supply agreements with Tencent worth more than $3 billion, and was still negotiating deep cooperation with other domestic cloud providers. This signal matters more than profit numbers: Changxin is no longer “an alternative when you can’t get supply.” It has become one of the core suppliers of China’s AI infrastructure.

From the perspective of the capital markets, Changxin’s listing also throws a benchmark anchor into A-shares. In the past, if investors wanted to gain exposure to domestic DRAM, they had to detour to buy shadow stocks of equipment, modules, and materials. Now the main asset itself has landed; funds naturally flow toward companies with real orders and real industrial status. Short-term liquidity disturbances and valuation reshuffles are inevitable, but in the long run, with Changxin as an anchor, the investment logic for domestic storage will only get stronger and stronger.

Of course, these are just footnotes for the capital market. They were never the core purpose of Changxin’s long march.

V. Ringing the bell isn’t the endpoint—it’s the first step of the long march

On the day of listing, many people said, “Congratulations—finally you’ve made it.”

But I feel the bell-ringing isn’t the endpoint. It’s an adulthood ceremony. Starting from this day on, Changxin truly stands at the global DRAM table and must face every wave head-on.

In “Xinshi,” I listed seven layers of crises Changxin would have to face. Even today, they still hold: pressure from geopolitical restrictions, the reversal of industry cycles, the gap in the high-end HBM market, being “choked” by core equipment and materials, the risk of patent litigation, the challenge of poaching talent, and pressure from short-term capital market performance assessments… none of them is easy.

But I’ve always had confidence in Changxin—not because of how much it earns now, but because from day one it has been walking the correct path: respecting the laws of industry, respecting the logic of technology, and respecting long-term value.

More importantly, Changxin’s value is never only its own. What it is meant to do in the future isn’t building an isolated island, but becoming the “leader among allies” for China’s local semiconductor supply chain—turning its massive capacity into a testing ground for domestic equipment, materials, and components. That gives the most precious opportunity for on-line validation to local manufacturers like North Huachuang and Advanced Micro-Fabrication Equipment. Only when the roots of the entire ecosystem take hold in local soil can Changxin’s supply chain security truly be as steady as Mount Tai. Only when the entire industrial chain becomes strong together can China’s storage truly stand up.

Conclusion

Ten years ago, nobody believed Chinese people could make their own DRAM.

After ten years, Changxin rang the bell and listed, firmly holding the position of fourth in the world.

During these ten years, Jin Hua in Fujian shut down; Unisplendour failed; Wuhan Hongxin was a scam; most competitors in the same track ended up with their hopes dashed. Only Changxin, step by step, went from blueprints to production lines, from white-label products to servers, from losses to profitability—until it walked into the spotlight of bell-ringing.

I’ve known Yiming for twenty years. I watched him start a business after returning from Silicon Valley. I watched Zhaoyi list. I watched him stake his entire net worth on DRAM. I watched him go from a spirited mid-career man to an industrial veteran with frost on his temples. In the preface to “Xinshi,” he wrote “history is a mirror,” and he himself is the most vivid witness to China’s storage history over the past decade.

The semiconductor industry has never had miracles—only trajectories.

All seemingly sudden successes are built on cold benches for more than ten years and investments that never counted returns.

The bell’s sound will fade, and the heat will ebb, but Changxin’s long march has only completed the first step.

China’s long march in semiconductors has only reached the point where the world can finally see it.