CXMT Is Eating Micron’s Lunch in China

American China Hawks love to believe that we can beat China by withholding superior American technology.

What they have missed repeatedly is how well China can do with “good enough” tech made in China. The latest painful example is how ChangXin Memory Technologies is eating Micron’s (NASDAQ: MU) noodles.

In 2023, Micron derived 14% of its revenue from China. By 2025, that had declined to 7%. In dollar terms, Micron generated approximately $2.64 billion from China in 2025, down by roughly $2.4–2.6 billion from its 2023 level. This decline occurred despite continued growth in global memory demand and rising revenues in other regions. China remains a major consumer of DRAM (Dynamic Random Access Memory) across PCs, servers, mobile devices, and embedded systems.

During this same period, CXMT’s reported revenue rose sharply and by an amount roughly similar to Micron’s decline. CXMT generated RMB 24.178 billion ($3.36 billion) in 2024 and RMB 32.084 billion ($4.46 billion) in the first nine months of 2025, implying an annualized run rate of approximately $5.9–6.0 billion. CXMT’s revenue growth explains a meaningful portion of Micron’s China decline, though CXMT remains unprofitable and technologically behind the global leaders.

CXMT grows into the China market

Between 2023 and 2025, CXMT’s DRAM wafer capacity tripled from 90,000 wafers per month to 260,000 wafers per month, establishing CXMT as an industrial-scale supplier rather than a pilot operation. Growth moderated but continued after 2025 as CMXT‘s capacity grew faster than Micron’s and is projected to pull nearly even by 2027. CMXT cannot compete globally against technology leaders Samsung (KRX: 005930), SK hynix, (KRX: 000660), or Micron, but it does not need to build share outside of China. Its capacity needs only to be sufficient to support domestic demand within China. The figures in Table 1 show that this threshold has been reached.

First bigger, then better

CXMT’s capacity growth slowed after 2025 as the company’s goals shifted from fab construction to yield improvement, node migration, and product qualification. CXMT is no longer focused on maximizing square footage. It is focused on converting installed capacity into usable output.

Once baseline capacity is in place, putting CXMT on the right side of the learning curve, further progress comes from improved process control, defect reduction, and product mix.

Good enough to go

CXMT remains one generation behind leading-edge DRAM suppliers. But leading-edge parity is not required for domestic substitution. CXMT has successfully manufactured 16-nanometer-class DDR5 (Double Data Rate 5) DRAM, producing a 16-gigabit device aligned with the current generation of mainstream memory products. DDR5 is expected to dominate mainstream DRAM markets through at least 2027. Qualification at current-generation DDR5 is sufficient to displace imports in many applications.

HBM on the way

CXMT does trail the leading suppliers in high-bandwidth memory (HBM) by approximately three to four years. HBM is indispensable to leading artificial intelligence chips. With their high margins, HBM devices have been driving revenue and profit growth at Samsung, SK hynix, and Micron. But CXMT has not been idle, vigorously pursuing process integration, packaging development, and customer qualification. Expect it to be able to fill some domestic demand by late 2026.

Success sans profit

CXMT generated revenue of RMB 8.287 billion in 2022, grew to RMB 24.178 billion in 2024, and RMB 15.438 billion in the first half of 2025. The company still loses money. In the first nine months of 2025 alone, it recorded a loss of RMB 5.28 billion.

No worries: the Chinese government is picking up the bill. CXMT was an early pilot for “Made in China 2025.” It has received billions in government investment. And its Chinese customers know the Chinese government wants them to shop local.

In political and policy terms, CXMT is a huge success and well worth the government support it gets, which makes that support likely to continue.

Made in America

Micron’s China troubles were made in America. Absent U.S. export controls and other trade war collateral damage, China would likely be buying more each year from the big three in memory, not less. China would be an opportunity for Micron rather than a dying market.

CXMT’s position in HBM further reinforces this dynamic. While CXMT is not participating in the leading-edge HBM4 market, it will soon be a player in HBM3 and HBM3E. These are “good enough” for inference workloads in data centers, where power efficiency, bandwidth, and availability matter more than absolute performance leadership. CXMT does not need to match Micron technologically at the HBM frontier. It needs only to remain qualified, available, and aligned with domestic policy.

Micron’s Chinese revenue will continue to decline as long as the China Hawks drive U.S. policy.