For decades, the global memory chip market has been a three-way oligopoly, controlled with quiet determination by Samsung, SK hynix, and Micron. The barriers to entry were measured not in years but in billions of dollars and decades of process expertise. In 2026, that fortress is showing its first real cracks. Two Chinese companies—ChangXin Memory Technologies (CXMT) and Yangtze Memory Technologies (YMTC)—have moved from being distant imitators to genuine challengers, closing the technology gap to a matter of years and, in some segments, months. The question is no longer whether China can make competitive memory chips. It is whether it can make them profitably, at scale, and without the tools that the incumbents have spent a generation perfecting.
The evidence of progress is undeniable. CXMT, founded in Hefei in 2016, has become the world’s fourth-largest DRAM producer. Its share of global DRAM revenue reached 9 percent in the third quarter of 2026, up from 5 percent a year earlier. Its quarterly revenue rose more than sevenfold to 118.7 billion yuan ($17.8 billion), with a net profit of 54.4 billion yuan, reversing years of losses. YMTC, meanwhile, ranked third in NAND shipments in the second quarter of 2026, surpassing Micron and Kioxia on a shipment basis and trailing only Samsung and SK hynix.
The Gap: One Year, Two Years, Three Years
The most striking measure of how far China has come is the shrinking technology gap. According to the Korea Semiconductor Industry Association, as of September 2026, the gap between Korea and China has narrowed to one year in NAND flash, two years in DRAM, and three years in HBM. Not long ago, that gap was hovering around five years.
The progress in HBM—the high-bandwidth memory that powers AI accelerators—is particularly significant. CXMT has begun small-scale production of HBM3E, narrowing the product-generation gap with Samsung and SK hynix to just one generation. Alibaba’s chip subsidiary T-Head and AI chipmaker Cambricon are already testing CXMT’s HBM3E with their own processors, and if plans proceed on schedule, Chinese AI chips incorporating domestically made HBM could reach the market as early as 2027.
In DRAM, CXMT has successfully mass-produced 16-nanometer DDR5 and achieved a 12-layer stacked HBM3 milestone. It has also validated prototypes using a sixth-generation 4F² design, a cell architecture that could shrink memory cells by roughly one-third compared with conventional designs. YMTC’s Xtacking architecture, which builds memory cells and control circuitry on separate wafers before bonding them, has enabled denser storage and faster data transfer without relying on the most advanced lithography equipment.
The Financial Engine: IPOs, Subsidies, and a Captive Market
China’s memory challengers are not fighting this battle with their own balance sheets alone. They are backed by a deliberate state strategy to achieve semiconductor self-sufficiency, and the capital is flowing.
CXMT raised 66.6 billion yuan ($8.6 billion) in a Shanghai IPO in July 2026—described as Asia’s largest listing of the year—giving it the financial firepower to invest in advanced DRAM and HBM development. YMTC is reportedly targeting an IPO as early as the first quarter of 2027. Government funding helped finance CXMT’s entry into the capital-intensive industry, and the memory boom is now funding its ramp-up.
The captive domestic market provides a crucial advantage. Beijing has reportedly raised the mandatory use rate of Chinese-made AI chips in new government or state-funded AI data centers from 50 percent to 80 percent. The domestic AI accelerator market is projected to see Chinese chips exceed 50 percent share for the first time in 2026, up from 35 percent in 2024. Huawei is projected to capture 62 percent of China’s domestic AI accelerator market in 2026, followed by Cambricon at 14 percent. This guaranteed demand gives CXMT and YMTC a foundation to scale, refine yields, and build customer relationships before taking on the global market.
The Obstacles: Yields, Tools, and the HBM Challenge
For all the progress, the gap that remains is the one that matters most: the ability to manufacture at competitive cost and quality.
Yields are the critical weakness. CXMT’s HBM3E pilot production is reportedly achieving yields of only around 25 percent, and its 12-layer HBM3 yields are estimated between 50 and 80 percent. By contrast, Samsung and SK hynix have spent years optimizing their HBM processes to achieve the yields necessary for profitable mass production. Low yields mean higher cost per usable chip, and in a commodity market, cost is everything.
Equipment restrictions are a structural constraint. US export controls have limited China’s access to the advanced manufacturing equipment used by the leading chipmakers, particularly extreme ultraviolet (EUV) lithography machines. CXMT and YMTC have responded with workarounds—YMTC’s Xtacking architecture and CXMT’s use of YMTC’s copper-to-copper hybrid bonding for HBM stacks are both designed to manufacture competitive products without EUV. But these workarounds come with trade-offs in complexity, yield, and scalability.
HBM is the hardest frontier. While CXMT has reached product-generation parity on HBM3, the leading Korean manufacturers are already moving toward HBM4. The actual technology gap in HBM is still estimated at roughly three to five years. And the transition from pilot production to mass production is not a given. “Developing a technology and mass-producing it are two entirely different things,” said Park Jae-geun, a professor of semiconductor engineering at Hanyang University.
The Incumbents Are Not Standing Still
It is also important to remember that Samsung and SK hynix are not passive targets. They are generating extraordinary profits from the AI boom—SK hynix’s third-quarter operating profit rose more than sevenfold to 83.3 trillion won ($62 billion), an operating margin of almost 80 percent. That financial firepower allows them to invest in next-generation technologies, defend their market share, and compete on cost in ways that a challenger with low yields cannot easily match.
Samsung has already demonstrated a working 16-gigabit 4F² DRAM prototype using vertical transistors and wafer-to-wafer hybrid bonding, a design that could yield more than 20 percent more chips per wafer under identical design rules. SK hynix has also presented research on vertical-gate 4F² memory. The incumbents are not ceding the technological high ground; they are preparing to defend it.
The bigger risk, as one analysis put it, is not that CXMT will out-innovate Samsung tomorrow. It is that CXMT is using the current boom to build the capacity and customer relationships it will need to challenge them when the market turns. Capacity expansion, not technological leadership, may be the lever that shifts the balance of power.
Closing the Gap, Not Yet Closing the Deal
China’s AI chipmakers are catching up. CXMT and YMTC have moved from being marginal players to genuine challengers, closing the technology gap to one to three years and building the financial, industrial, and customer foundations for a long-term challenge. The progress is real, and it is faster than almost anyone predicted.
But catching up is not the same as overtaking. The remaining obstacles—low yields, equipment restrictions, the complexity of HBM, and the incumbents’ entrenched advantages in cost and process—are formidable. The most likely outcome in the near term is not a Chinese takeover of the global memory market, but the emergence of a parallel supply chain: a domestic Chinese ecosystem that serves Chinese AI chipmakers, insulated from US export controls, while Samsung and SK hynix continue to dominate the global market for the most advanced products.
That parallel ecosystem matters. It reduces China’s vulnerability to sanctions, it gives Huawei and Cambricon a domestic memory source, and it creates a foundation for future competition. But it does not yet threaten the Korean giants’ core business. The gap has closed to years, not months. The next phase of the competition will be decided not by who can make the most advanced chip in a lab, but by who can make it profitably, at scale, and with the yields that make the economics work.
China has proven it can build the chip. Now it has to prove it can build the business.
