As ChangXin Memory Technologies gears up for a breakout public listing, the chip maker is eyeing the turf held by Micron, SK Hynix and Samsung
ChangXin Memory Technologies grew revenue by over 700% in the first quarter of 2026 amid a global DRAM shortage.
By the 2010s, China had firmly established itself as the world's largest electronics assembly hub and consumer of semiconductors.
Still, a key issue kept Beijing's top economic planners up at night. China remained dependent on U.S. and Korean companies for a component found in every phone and laptop: memory. Without memory-chip fabrication capabilities, China's domestic semiconductor industry faced a critical supply-chain chokepoint.
After over a decade of heavy investment and technical breakthroughs, China has developed a key foothold in a famously volatile market. The Chinese company ChangXin Memory Technologies is now the world's fourth-largest producer of dynamic random-access memory with an 8% market share, sitting behind the industry's "Big Three" oligopoly: Micron $(MU)$, SK Hynix (SKHY) and Samsung (KR:005930).
DRAM serves as a device's working memory, temporarily storing data that need to be accessed quickly. These standardized chips are churned out in massive quantities by producers who must compete on volume and cost efficiency.
CXMT has become the primary domestic supplier of memory chips within China, and now the company is setting its sights on global expansion as the artificial-intelligence buildout and its accompanying geopolitical tensions heat up. In what's shaping up to be Asia's largest largest initial public offering of 2026, CXMT is set to raise $8.6 billion on the Shanghai Stock Exchange STAR Market later this month.
AI workloads demand massive amounts of high-bandwidth memory, a type of advanced memory composed of multiple DRAM dies stacked on top of each other. Soaring demand for HBM has led to severe shortages and rising prices. HBM producers have diverted their efforts toward fulfilling lucrative data-center contracts, resulting in increasing component costs for consumer products.
There are signs of China's growing influence as the AI boom upends the status quo. In June, the Roundhill Memory exchange-traded fund DRAM added GigaDevice (CN:603986) - the Chinese semiconductor company that serves as a major financial backer of and strategic partner to CXMT - to its holdings. CXMT is also said to be receiving interest from Apple $(AAPL)$, historically one of Micron's most significant customers. After raising prices on its products due to climbing memory costs, Apple is reportedly lobbying the Trump administration for approval to purchase DRAM from CXMT, which currently sits on the U.S. Defense Department's list of companies suspected of having ties to the Chinese military.
Representatives from Apple did not respond to a MarketWatch request for comment.
At the same time, the U.S. is prioritizing strengthening its own domestic semiconductor ecosystem to maintain technological leadership, with Micron pledging over $250 billion in U.S. investment through 2035.
"The memory shortage is testing the limits of technology decoupling," Poe Zhao, a China tech analyst and the founder of the Hello China Tech Substack publication, told MarketWatch. "The U.S. wants to restrict China's rise in strategic semiconductors, while global companies still need adequate supply at reasonable cost."
'It is time for China to play a role'
CXMT was founded in 2016 in Hefei, a Chinese city that has transformed from an agricultural hub into a state-backed venture-capital powerhouse. There, GigaDevice founder and CEO Zhu Yiming teamed up with the Hefei government to invest roughly $2.5 billion to kick-start China's domestic DRAM initiative.
Zhu had long held a vision of China becoming a memory superpower. After spending the early 2000s as an engineer in Silicon Valley, he observed that the memory industry was slowly migrating away from its U.S. origins to Japan, then Korea and Taiwan. "It is time for China to play a role in this industry," Zhu wrote in a 2004 email pitching GigaDevice to investors. "Our long-term vision is to be the biggest memory designer and maker in China."
While Zhu would grow GigaDevice into a highly successful publicly traded chip designer in the years that followed, the company lacked the capital and expertise to become a memory juggernaut. But perhaps CXMT, bolstered by the deep pockets of Hefei's municipal-investment arm, would be able to accomplish what GigaDevice couldn't.
China had tried to enter the memory market before - first in the 1990s and again a decade later. Each attempt went bust when global economic downturns cratered the semiconductor market and sent chip prices plummeting. Unfazed by past failures, Zhu stepped down from his CEO role at GigaDevice in 2018 to become the CEO of CXMT, reportedly pledging to take no salary until the company was profitable.
CXMT faced a steep barrier with virtually no institutional knowledge of the memory-manufacturing process. To get up to speed, the company picked through the wreckage of previous boom-bust cycles. CXMT secured DRAM patents and blueprints from Qimonda, a German memory giant that went bankrupt during the 2007-09 financial crisis. CXMT also embarked on an aggressive and controversial hiring blitz, targeting talent from the bankrupt Japanese company Elpida Memory - which was acquired by Micron in 2013 - as well as Samsung and SK Hynix.
By the end of 2019, CXMT had brought its first fab online and begun commercial production of memory chips. More government support came in 2020, this time from China's National Integrated Circuit Industry Investment Fund Phase II. Dubbed the "Big Fund," the state-backed vehicle elevated CXMT from a municipal project to a critical national interest.
While China's memory industry has certainly benefited from government backing, CXMT's progress cannot be reduced to subsidies alone, according to Zhao of Hello China Tech. Export controls have increased domestic demand for CXMT's products and spurred the development of innovative workarounds.
To pack more memory onto each chip, the Big Three memory manufacturers have shrunk their process nodes down to the 11-nanometer leading edge using ASML's $(ASML)$ ultrafine extreme-ultraviolet lithography tools. Barred from accessing EUV equipment, CXMT has still managed to shrink its nodes down to 16nm, from 19nm initially, with less advanced deep-ultraviolet technology.
This feat earned CXMT significant industry recognition and has led some experts to theorize that CXMT could reach the 11nm class without EUV, a semiconductor engineer writing under the pseudonym Damnang noted in an April Substack post. Currently, CXMT currently relies on DUV lithography combined with multi-patterning, which divides a dense pattern across multiple exposure and processing steps to achieve features smaller than a single DUV exposure can reliably print, Damnang told MarketWatch.
From DRAM to HBM
CXMT turned a profit for the first time in 2025 as it capitalized on the severe consumer DRAM shortage caused by the AI boom. The memory upcycle has only continued into 2026, and the company recorded 719% year-over-year revenue growth in the first quarter of 2026. Today, CXMT operates three fabs in China, producing over 300,000 wafers per month. The company is on track to exit 2026 at 350,000 wafers per month, nearly matching Micron's estimated DRAM production.
But CXMT still has some catching up to do. Valued at around $80 billion, CXMT pales in comparison to Micron, Samsung and SK Hynix, whose market capitalizations hover near the $1 trillion mark. With the proceeds from its upcoming IPO, CXMT aims to scale up its production volume, grow a global customer base and tackle its biggest challenge yet: breaking into the HBM market with an ambitious goal of mass-producing HBM3 by the end of 2026.
The global HBM market is currently a three-way oligopoly, with Samsung, SK Hynix and Micron already shipping their next-generation HBM4 chips. Skyrocketing demand for HBM has given these companies unprecedented pricing power, putting Micron on track to become one of the most profitable companies in the world in 2027.
CXMT's products are "probably two to three years behind" mainstream DRAM technologies, with an even larger gap on HBM, according to Andrew Lu, an independent chip analyst and former head of Asia Pacific semiconductor research at Barclays and Citi.
HBM is notoriously challenging and expensive to manufacture. Specialized advanced packaging and 3D-stacking architectures are required to bind dozens of DRAM chips together with tiny vertical wires. If CXMT "wants to make it into HBM, their current technology isn't enough," Lu said. The HBM market moves rapidly, and customers will want the latest generation of chips. "You can use low-end HBM as a substitute, but the speed and everything else won't really be up to par," he added.
Designing HBM is only half the battle. The real challenge is in producing HBM at scale, according to Tom Coughlin, a memory and storage analyst and founder of Coughlin Associates. HBM yields, or the percentage of functional chips per production run, are much lower than DRAM due to their complexity.
Without access to advanced lithography tools, CXMT must once again turn to alternative workarounds with DUV technology. That could actually play to CXMT's advantage, according to Neil Shah, vice president of research at Counterpoint Research. "Limiting CXMT could help it leapfrog ahead of the incumbents who are likely to delay such innovations to protect returns on existing equipment," Shah wrote in a note earlier this week.
According to Counterpoint Research's projections, CXMT could snag a 0.4% share in the HBM market in 2027 and grow that to 3% in 2028.
MW How a homegrown Chinese chip maker became the memory industry's biggest wild card
By Christine Ji
As ChangXin Memory Technologies gears up for a breakout public listing, the chip maker is eyeing the turf held by Micron, SK Hynix and Samsung
ChangXin Memory Technologies grew revenue by over 700% in the first quarter of 2026 amid a global DRAM shortage.
By the 2010s, China had firmly established itself as the world's largest electronics assembly hub and consumer of semiconductors.
Still, a key issue kept Beijing's top economic planners up at night. China remained dependent on U.S. and Korean companies for a component found in every phone and laptop: memory. Without memory-chip fabrication capabilities, China's domestic semiconductor industry faced a critical supply-chain chokepoint.
After over a decade of heavy investment and technical breakthroughs, China has developed a key foothold in a famously volatile market. The Chinese company ChangXin Memory Technologies is now the world's fourth-largest producer of dynamic random-access memory with an 8% market share, sitting behind the industry's "Big Three" oligopoly: Micron (MU), SK Hynix (SKHY) and Samsung (KR:005930).
DRAM serves as a device's working memory, temporarily storing data that need to be accessed quickly. These standardized chips are churned out in massive quantities by producers who must compete on volume and cost efficiency.
CXMT has become the primary domestic supplier of memory chips within China, and now the company is setting its sights on global expansion as the artificial-intelligence buildout and its accompanying geopolitical tensions heat up. In what's shaping up to be Asia's largest largest initial public offering of 2026, CXMT is set to raise $8.6 billion on the Shanghai Stock Exchange STAR Market later this month.
AI workloads demand massive amounts of high-bandwidth memory, a type of advanced memory composed of multiple DRAM dies stacked on top of each other. Soaring demand for HBM has led to severe shortages and rising prices. HBM producers have diverted their efforts toward fulfilling lucrative data-center contracts, resulting in increasing component costs for consumer products.
There are signs of China's growing influence as the AI boom upends the status quo. In June, the Roundhill Memory exchange-traded fund DRAM added GigaDevice (CN:603986) - the Chinese semiconductor company that serves as a major financial backer of and strategic partner to CXMT - to its holdings. CXMT is also said to be receiving interest from Apple (AAPL), historically one of Micron's most significant customers. After raising prices on its products due to climbing memory costs, Apple is reportedly lobbying the Trump administration for approval to purchase DRAM from CXMT, which currently sits on the U.S. Defense Department's list of companies suspected of having ties to the Chinese military.
Representatives from Apple did not respond to a MarketWatch request for comment.
At the same time, the U.S. is prioritizing strengthening its own domestic semiconductor ecosystem to maintain technological leadership, with Micron pledging over $250 billion in U.S. investment through 2035.
"The memory shortage is testing the limits of technology decoupling," Poe Zhao, a China tech analyst and the founder of the Hello China Tech Substack publication, told MarketWatch. "The U.S. wants to restrict China's rise in strategic semiconductors, while global companies still need adequate supply at reasonable cost."
'It is time for China to play a role'
CXMT was founded in 2016 in Hefei, a Chinese city that has transformed from an agricultural hub into a state-backed venture-capital powerhouse. There, GigaDevice founder and CEO Zhu Yiming teamed up with the Hefei government to invest roughly $2.5 billion to kick-start China's domestic DRAM initiative.
Zhu had long held a vision of China becoming a memory superpower. After spending the early 2000s as an engineer in Silicon Valley, he observed that the memory industry was slowly migrating away from its U.S. origins to Japan, then Korea and Taiwan. "It is time for China to play a role in this industry," Zhu wrote in a 2004 email pitching GigaDevice to investors. "Our long-term vision is to be the biggest memory designer and maker in China."
While Zhu would grow GigaDevice into a highly successful publicly traded chip designer in the years that followed, the company lacked the capital and expertise to become a memory juggernaut. But perhaps CXMT, bolstered by the deep pockets of Hefei's municipal-investment arm, would be able to accomplish what GigaDevice couldn't.
China had tried to enter the memory market before - first in the 1990s and again a decade later. Each attempt went bust when global economic downturns cratered the semiconductor market and sent chip prices plummeting. Unfazed by past failures, Zhu stepped down from his CEO role at GigaDevice in 2018 to become the CEO of CXMT, reportedly pledging to take no salary until the company was profitable.
CXMT faced a steep barrier with virtually no institutional knowledge of the memory-manufacturing process. To get up to speed, the company picked through the wreckage of previous boom-bust cycles. CXMT secured DRAM patents and blueprints from Qimonda, a German memory giant that went bankrupt during the 2007-09 financial crisis. CXMT also embarked on an aggressive and controversial hiring blitz, targeting talent from the bankrupt Japanese company Elpida Memory - which was acquired by Micron in 2013 - as well as Samsung and SK Hynix.
By the end of 2019, CXMT had brought its first fab online and begun commercial production of memory chips. More government support came in 2020, this time from China's National Integrated Circuit Industry Investment Fund Phase II. Dubbed the "Big Fund," the state-backed vehicle elevated CXMT from a municipal project to a critical national interest.
While China's memory industry has certainly benefited from government backing, CXMT's progress cannot be reduced to subsidies alone, according to Zhao of Hello China Tech. Export controls have increased domestic demand for CXMT's products and spurred the development of innovative workarounds.
To pack more memory onto each chip, the Big Three memory manufacturers have shrunk their process nodes down to the 11-nanometer leading edge using ASML's (ASML) ultrafine extreme-ultraviolet lithography tools. Barred from accessing EUV equipment, CXMT has still managed to shrink its nodes down to 16nm, from 19nm initially, with less advanced deep-ultraviolet technology.
This feat earned CXMT significant industry recognition and has led some experts to theorize that CXMT could reach the 11nm class without EUV, a semiconductor engineer writing under the pseudonym Damnang noted in an April Substack post. Currently, CXMT currently relies on DUV lithography combined with multi-patterning, which divides a dense pattern across multiple exposure and processing steps to achieve features smaller than a single DUV exposure can reliably print, Damnang told MarketWatch.
From DRAM to HBM
CXMT turned a profit for the first time in 2025 as it capitalized on the severe consumer DRAM shortage caused by the AI boom. The memory upcycle has only continued into 2026, and the company recorded 719% year-over-year revenue growth in the first quarter of 2026. Today, CXMT operates three fabs in China, producing over 300,000 wafers per month. The company is on track to exit 2026 at 350,000 wafers per month, nearly matching Micron's estimated DRAM production.
But CXMT still has some catching up to do. Valued at around $80 billion, CXMT pales in comparison to Micron, Samsung and SK Hynix, whose market capitalizations hover near the $1 trillion mark. With the proceeds from its upcoming IPO, CXMT aims to scale up its production volume, grow a global customer base and tackle its biggest challenge yet: breaking into the HBM market with an ambitious goal of mass-producing HBM3 by the end of 2026.
The global HBM market is currently a three-way oligopoly, with Samsung, SK Hynix and Micron already shipping their next-generation HBM4 chips. Skyrocketing demand for HBM has given these companies unprecedented pricing power, putting Micron on track to become one of the most profitable companies in the world in 2027.
CXMT's products are "probably two to three years behind" mainstream DRAM technologies, with an even larger gap on HBM, according to Andrew Lu, an independent chip analyst and former head of Asia Pacific semiconductor research at Barclays and Citi.
HBM is notoriously challenging and expensive to manufacture. Specialized advanced packaging and 3D-stacking architectures are required to bind dozens of DRAM chips together with tiny vertical wires. If CXMT "wants to make it into HBM, their current technology isn't enough," Lu said. The HBM market moves rapidly, and customers will want the latest generation of chips. "You can use low-end HBM as a substitute, but the speed and everything else won't really be up to par," he added.
Designing HBM is only half the battle. The real challenge is in producing HBM at scale, according to Tom Coughlin, a memory and storage analyst and founder of Coughlin Associates. HBM yields, or the percentage of functional chips per production run, are much lower than DRAM due to their complexity.
Without access to advanced lithography tools, CXMT must once again turn to alternative workarounds with DUV technology. That could actually play to CXMT's advantage, according to Neil Shah, vice president of research at Counterpoint Research. "Limiting CXMT could help it leapfrog ahead of the incumbents who are likely to delay such innovations to protect returns on existing equipment," Shah wrote in a note earlier this week.
According to Counterpoint Research's projections, CXMT could snag a 0.4% share in the HBM market in 2027 and grow that to 3% in 2028.
(MORE TO FOLLOW) Dow Jones Newswires
July 18, 2026 08:00 ET (12:00 GMT)
MW How a homegrown Chinese chip maker became the -2-
"Let's see what they can do," Coughlin said. Manufacturing HBM at scale requires expertise that must be built up over multiple iterations of trial and error. "You might be able to make some, but can you make it on a consistent basis with good yields? That's when the experience comes in," he said.
Now read: Can China just steal America's AI brain that's costing trillions to develop?
-Christine Ji
This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.
(END) Dow Jones Newswires
July 18, 2026 08:00 ET (12:00 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments