Reporter |
Editor | Wen Shuqi
On the morning of July 27th, ChangXin Memory Technologies (688825), China’s DRAM (Dynamic Random Access Memory) leader, officially rang the bell on the STAR Market. Its share price surged a whopping 471.59% at the open to 49.5 yuan per share, pushing its total market cap past 3.31 trillion yuan. This “chip behemoth,” which set a new fundraising record on the STAR Market, made a capital market debut that was nothing short of spectacular.
The market’s frenzy is rooted in solid logic. During the retail subscription phase, ChangXin saw an incredible 212 times oversubscription, with over 9.4 million valid orders. The combination of tight supply and sky-high sentiment made a doubling of the share price a near-consensus view.
Just seven months ago, this trillion-yuan chip giant was barely at the starting line of its IPO journey.
On December 30, 2025, ChangXin’s STAR Market IPO application was accepted by the Shanghai Stock Exchange. Thanks to a “pre-review” mechanism, regulators had already done a lot of legwork before the formal filing. On May 27, 2026, the listing committee gave its approval; on June 12, the CSRC issued its registration; and on July 27, the company officially started trading. From application to registration approval in just 165 days, and from acceptance to listing in under seven months, ChangXin smashed the record for the fastest large-scale IPO on the STAR Market.
ChangXin’s story so far is a tale of a bold counter-cyclical gamble paying off big time. That said, ChangXin still lags behind global leaders in high-end products like HBM (High Bandwidth Memory), and it’s an open question whether the current massive surge in DRAM prices is sustainable.
But this strategic window of opportunity isn’t waiting for anyone. With its successful listing, ChangXin is set to get a massive capital injection. This cash will be crucial for expanding production capacity, refining its manufacturing processes, and attracting top talent—all key to making China’s homegrown memory industry bigger and stronger.
The “Counter-Cyclical Gamble” That Broke the Iron Curtain
ChangXin’s journey began as a high-stakes bet, picking up the pieces from a patent graveyard.
The mastermind who put this gamble on the table is Zhu Yiming.
A Tsinghua University graduate, Zhu Yiming worked in the semiconductor industry in the US, focusing on network processors and memory chips. In 2005, he returned to China to found GigaDevice, spending over a decade transforming a tech startup into a publicly listed chip company.
In 2018, Zhu stepped down as GigaDevice’s general manager to focus on ChangXin, taking on roles as chairman, CEO, and later chairman of the parent company. For him, ChangXin wasn’t just a shift from one type of memory product to another. It was a leap from the asset-light world of chip design into the capital, technology, and organizationally intense realm of wafer fabrication—a “second startup” with much higher stakes.
When ChangXin’s predecessor was registered in Hefei back in 2016, the global DRAM market had been locked down by Samsung, SK Hynix, and Micron for over two decades. Together, they held more than 95% market share. ChangXin’s team spent billions of dollars to acquire over 6,200 DRAM patents left behind by the bankrupt German memory giant Qimonda in 2009. The main point of this deal wasn’t to get the technology directly. It was to build a legal moat, sidestepping the patent barriers set up by Samsung and Hynix, to give their own R&D a fighting chance.
At the same time, they began recruiting overseas Chinese engineers who had worked at Hynix, Micron, and TSMC, pulling together ChangXin’s first technical team.
Around Zhu Yiming, ChangXin quickly built a core team with a mix of experience in industrial management, product development, and fab operations. Cao Kanyu, who joined in 2017, brought over two decades of semiconductor experience and later led the company’s R&D strategy and technology roadmap. Zhao Lun, who also joined around the same time, was a former general manager of Datang Microelectronics and knew the ins and outs of managing large-scale domestic semiconductor projects.
From there, ChangXin kept absorbing technical talent from international memory companies: Li Hongwen, a former design manager at Micron Shanghai; Chen Dehong, a long-time Micron Singapore veteran; Wang Dan, a former senior engineer at Samsung; and Zhu Wenju, who had worked at Intel and Micron subsidiaries. This team covered everything from chip design and process development to factory operations and mass production, plugging the critical gaps needed to take homegrown DRAM from the lab to the production line.
On the technology front, ChangXin took an unusual “skip-generation” strategy. When its first-generation product went into mass production in 2019, ChangXin jumped straight from the 18nm process to the more challenging 17nm, hoping to leapfrog from DDR4 (Double Data Rate 4) straight into the DDR5 era.
In the semiconductor world, process nodes usually advance step-by-step, with each step requiring years of effort and billions of dollars. ChangXin’s gamble was based on the logic that playing it safe would mean forever trailing the global leaders, and the market window for DDR5 wasn’t going to wait for latecomers.
An even bigger bet was placed on production capacity.
From 2022 to 2023, the global memory industry went through its worst downturn in nearly 15 years. Samsung’s operating profit in Q1 2023 plummeted 96% year-over-year, hitting its lowest point since the 2008 financial crisis. SK Hynix posted its biggest quarterly loss ever. Micron reported a net loss of $5.8 billion for its fiscal year 2023, laid off 5,000 employees, and suspended all bonuses.
NAND flash prices collapsed 30% to 40% for the year, and DRAM contract prices fell for eight straight quarters. The big three memory makers all cut production to support prices, and industry inventories soared to a historic high, equaling three to four months of supply.
But ChangXin did the exact opposite.
With continued funding from Hefei’s state capital and China’s Big Fund Phase II, ChangXin aggressively expanded production from 2022 to 2024. Its monthly wafer output jumped from under 100,000 to 300,000. At one point, its product prices were just half of what its overseas competitors were charging. Under immense cost pressure, this meant ChangXin was likely losing money on every chip it sold.
Its prospectus shows that from 2022 to 2024, ChangXin accumulated losses of over 30 billion yuan, with a net loss of 7.145 billion yuan in 2024 alone.
“At the time, a lot of people in the industry thought we were committing suicide,” a former ChangXin employee recalled. “But looking back now, it was actually a strategic window. Samsung, Hynix, and Micron were all shifting their capacity to HBM, so supply of traditional DRAM was shrinking, while demand from AI and data centers was just exploding. If we didn’t fill that gap, we would never have another chance.”
And as it turned out, ChangXin’s moment finally arrived.
The Strategic Window of Opportunity
ChangXin’s strategy was based on a different playbook: the memory chip business is a classic cyclical industry. “Expand during the downturn, harvest during the upswing” is the exact same path Samsung used to overtake its Japanese and American rivals back in the 1980s.
ChangXin was betting on the policy and market space created by domestic substitution, as well as the timing of the cycle turning around.
That bet is starting to pay off, but the window might be narrower than anyone thought.
Since 2025, the global DRAM industry has seen a massive reallocation of production capacity. NVIDIA’s GPUs created an explosive demand for HBM, which offers profit margins several times higher than traditional DRAM. SK Hynix was the first to convert traditional DRAM lines to HBM, with Samsung and Micron quickly following suit.
According to TrendForce, from 2025 to 2026, the big three’s wafer input for HBM grew over 80% year-over-year, while capacity for traditional DRAM, including DDR4 and DDR5, grew less than 10%.
This means the supply of traditional DRAM suddenly slowed down. But demand hasn’t stopped. AI servers need DDR5, domestic PC and phone makers need LPDDR5, and data centers need RDIMMs. The capacity that ChangXin built up during the 2022-2024 downturn was perfectly positioned to fill this structural supply-demand gap.
Data from industry research firm IC Insights clearly shows this turnaround. In Q2 2025, ChangXin held just 3.97% of the global DRAM market. By Q4, its share had jumped to 7.67%, eating into the market share of international players in some traditional DRAM segments and solidifying its position as the world’s fourth-largest player.
On the other hand, policies pushing for domestic substitution in government and enterprise servers, as well as the computing power bases of Chinese internet giants, are accelerating. In July 2026, seven ministries including the MIIT and NDRC jointly issued the “2026-2028 National Computing Infrastructure Construction Action Plan,” which took effect from September 2026.
“The cycle for memory chips is simple: expand in the down cycle, harvest in the up cycle,” a ChangXin employee explained. “But ChangXin’s expansion isn’t just about the cycle. There’s also the domestic substitution logic. The big three moving to HBM has objectively opened up market space for ChangXin in traditional DRAM. If ChangXin had cut production along with everyone else in 2022, that market would belong to someone else today.”
TrendForce data shows that in Q3 2025, DRAM contract prices rose about 40% quarter-over-quarter, followed by another 13% to 18% increase in Q4. The pace accelerated in 2026: Q1 saw a 90% to 95% sequential jump, and Q2 added another 58% to 63%.
A report from Counterpoint noted that memory prices surged 80% to 90% quarter-over-quarter in Q1 2026, with all product categories hitting record highs.
ChangXin’s financial performance has tracked this price surge. In 2025, the company turned profitable for the full year, posting a net profit of 1.875 billion yuan. In Q1 2026, it recorded 50.8 billion yuan in revenue, a staggering 719.13% year-over-year increase, with a net profit of 24.76 billion yuan.
By shipment volume, ChangXin is now China’s largest and the world’s fourth-largest DRAM manufacturer. The company’s prospectus reveals partnerships with core clients like Tencent, Alibaba Cloud, ByteDance, Lenovo, and Xiaomi.
Behind this dramatic comeback in capacity and market share, ChangXin’s rise is more than just one company’s success story. It’s become a crucial lever for upgrading the entire Chinese semiconductor supply chain.
In the old days of “lacking chips and core technology,” domestic memory companies were too small to demand custom R&D from giant overseas equipment and materials suppliers. The domestic supply chain was stuck in a chicken-and-egg loop: “we have products, but no production line to test them on.”
But as ChangXin ramped up its monthly output from under 100,000 wafers to over 300,000, its massive production needs started to give a real boost to the domestic supply chain.
In the core equipment sector, the expansion of leading companies like ChangXin has accelerated the introduction and iteration of domestic equipment.
For example, in a key “laser-induced crystallization” step used in advanced 3D NAND flash manufacturing, Chengdu Laipu Technology, in collaboration with a domestic industry leader, developed a laser annealing tool that has achieved global mass production first. This broke the exclusive monopoly of overseas companies and effectively supports the mass production of next-gen memory products.
In the field of precision semiconductor components, companies like Sinoceramics Electronics have overcome material and process bottlenecks for core products like electrostatic chucks and ceramic heaters. Some technical specs of these products have already reached international standards, accelerating the domestic substitution of foreign parts.
The synergy within the domestic chain is also showing up in basic materials and packaging.
Companies like Zhuhai Kelong Technology are now able to stably produce various high-end wet electronic chemicals, some reaching the G5 grade, meaning metal ion impurities are controlled to parts per trillion. These products have broken the decades-long monopoly of overseas companies and filled some gaps in the semiconductor materials industry in South China.
In advanced packaging, companies like Zhuhai Tiancheng Advanced Semiconductor are tackling core 2.5D and 3D integration technologies. The company has built its own “Nine Layers” technology system, with over 85% of its core equipment sourced domestically, providing a solid foundation for high-end domestic chips.
Once ChangXin had established self-controlled manufacturing capacity, the orders it generated started flowing down to the equipment, materials, precision components, and packaging & testing segments. This has helped a cluster of specialized and innovative suppliers across the supply chain achieve breakthroughs together.
Looking back, ChangXin’s decision to list on the STAR Market, getting from application to registration in just 165 days, sends a clear signal: this strategic window of opportunity is not going to wait.
Ramping Up Production and Hiring for the Next Battle
Right now, ChangXin’s capacity expansion plan is accelerating rapidly.
ChangXin has established a dual-base layout for 12-inch DRAM wafer manufacturing in Hefei and Beijing. By the end of 2026, the company’s monthly wafer input capacity is expected to exceed 300,000, making it roughly three times larger than the second-tier domestic DRAM makers.
As of early 2026, ChangXin operates three 12-inch DRAM fabs in Hefei and Beijing, with a monthly capacity of around 280,000 to 300,000 wafers, and capacity utilization consistently above 95%. By the end of 2026, monthly capacity is projected to rise to about 350,000 wafers, closing in on the estimated capacity of one of the big three, Micron, which is around 385,000.
With the construction and progress of a new fab in Shanghai’s Lingang area, ChangXin plans to boost monthly capacity to about 420,000 wafers by 2027, and target 500,000 wafers per month by the end of 2028. By then, its share of global DRAM capacity could rise to around 17%.
ChangXin’s long-term goal is to reach a monthly capacity of 600,000 wafers. At its current expansion rate, the company could surpass Micron in overall production scale around 2030.
In the high-bandwidth memory space for AI computing, ChangXin has already started mass production of HBM2. The company plans to convert some capacity at its Hefei and Beijing fabs to produce 60,000 HBM wafers per month. Construction of a “super fab” in Shanghai, primarily for high-end memory products like HBM3, began in March 2026. ChangXin also plans to build a new HBM packaging and testing facility in Shanghai, expected to start production by the end of 2026.
Industry analysts predict that by 2028, ChangXin’s monthly HBM wafer capacity could reach 100,000, accounting for about 12% of global HBM capacity.
Besides expanding production, ChangXin is also on a massive hiring spree.
At the end of June, when reporters visited the area around ChangXin’s technology park in Hefei’s Economic Development Zone to check out rental prices, they noticed a clear spike in rents.
“In March and April, apartments were really cheap. You could rent a two-bedroom for 15,000 yuan for a whole year. After May, the rent just shot up. Now a similar apartment costs over 20,000 yuan a year,” a local rental agent, Sister Gao, told us.
Behind this rental boom is a clear flow of talent.
As ChangXin pushes its capacity from 300,000 wafers per month towards even higher targets, it needs the engineering team to match.
One industry insider explained that a process integration engineer on a 12-inch wafer line typically needs two to three years of training before they can independently manage a module. The hiring push happening today is actually about stockpiling talent for the capacity ramp-up that will happen two years down the line.
ChangXin’s 2026 campus recruitment page on its website currently lists 45 open positions, covering locations in Hefei, Beijing, Shanghai, Xi’an, and even Japan.
These roles span circuit design, process integration, semiconductor data science, and AI application development. Most positions require at least a master’s degree.
ChangXin’s hiring spree is also directly poaching talent from foreign memory makers. Before, because there was no domestic DRAM company of a similar scale, engineers leaving Hynix’s Wuxi fab or Samsung’s Xi’an fab often had to switch to chip design or move to foundries.
ChangXin’s rise has changed that dynamic. Since 2025, significantly more engineers have been moving from Hynix’s Wuxi fab to ChangXin.
An employee at Samsung’s China division told us: “In the last few years, ChangXin has become the top choice for many colleagues looking to change jobs. A lot of them got options. After ChangXin announced its IPO this year, another wave of people jumped ship.”
According to the prospectus, ChangXin’s headcount was 9,605 in 2023, 13,858 in 2024, and 19,298 at the end of 2025. The company’s workforce doubled in just two years.
The prospectus also shows that Hefei’s state capital, through platforms like Qinghui Jidian and Hefei Jixin, holds a combined stake of about 30%, making it the most critical strategic investor after the national Big Fund Phase II.
With ChangXin’s listing on the STAR Market and its subsequent valuation surge, the book value of the Hefei state capital’s stake has grown by more than an order of magnitude over its initial investment.
The essence of this investment return is a textbook example of a city using state capital as leverage to bet on a national strategic industry and then reaping the rewards during a super-cycle.
More important than the financial return is this: as ChangXin’s fabs in Hefei and Beijing run at full capacity, China, for the first time, has self-controlled production capacity in a core semiconductor sector that was monopolized by three international companies for three decades.
But the ringing of the listing bell isn’t the end of the story. ChangXin still faces some serious challenges.
Memory chips are a highly cyclical industry. The “price hike → expansion → oversupply → price crash” script has played out many times before.
Right now, there’s a clear split in market opinion on when this cycle will turn. Goldman Sachs, Bank of America, UBS, and others believe this super-cycle could last until 2027 or 2028.
Goldman Sachs, for instance, uses a supply-demand gap model to predict a 4.9% global DRAM supply deficit in 2026 and a 5.1% deficit for HBM—both the highest levels since 2011. They expect the price hike trend to continue until 2028, with memory chip supply potentially being even tighter in 2027 than in 2026, and the real inflection point for capacity release coming in 2028.
Bloomberg analysts, on the other hand, believe the memory chip shortage will ease significantly in Q1 2027 and could even lead to an oversupply as early as 2028. If large cloud service providers cut capital spending, an oversupply could happen even sooner, in 2027.
For ChangXin, the key is to use this one-to-two-year window to complete its IPO fundraising, rapidly expand capacity, and solidify its scale, technology, and customer relationships. If it doesn’t, when overseas capacity flows back and price wars restart, a lack of scale will leave it with no room to survive.
ChangXin’s net book value of fixed assets is a staggering 183 billion yuan, and its annual depreciation charges exceed 10 billion yuan. Once product prices enter a downward cycle, the company’s income statement will come under immense pressure.
On the technology front, while ChangXin is already mass-producing DDR5 and continuously catching up, it still lags behind SK Hynix and Samsung in the HBM space for AI computing.
HBM is the next battlefield, and it’s a fortress ChangXin must conquer. But the opportunity is just as clear.
The domestic substitution of China’s computing infrastructure is accelerating. ChangXin plans to use the nearly 60 billion yuan raised from its IPO primarily for HBM production line construction and advanced process R&D. This capital is both the ammunition to defend its position in the traditional DRAM market and the ticket to charge into the high-end market.
From a patent graveyard to the STAR Market, ChangXin’s journey has taken a decade. But today’s listing bell is just a comma in the middle of its long march.
