China's memory makers are playing a long game and it's working

YMTC and CXMT are scaling fast, pricing sharp, and backed by Beijing. The global memory hierarchy is shifting.

China's two dominant memory chipmakers are not waiting for permission to reshape the global market. Yangtze Memory Technologies Co (YMTC) and ChangXin Memory Technologies (CXMT) are moving on parallel tracks, expanding production capacity, sharpening their pricing edge, and preparing to flood the market with both NAND and DRAM just as an AI-driven demand super cycle creates the ideal conditions to do so.

The headline numbers tell part of the story. CXMT posted 2025 revenue that surged 130% from a year earlier, exceeding 55 billion yuan (approximately US$8 billion). That kind of growth does not happen without structural tailwinds and the company has several working simultaneously in its favour.

On the NAND side, YMTC exceeded a 10% shipment share in the global NAND Flash market in the first quarter of 2025, rising to 13% by the third quarter. The pace of that climb is notable. YMTC is now preparing its next move: mass production of cutting-edge NAND products at a new production line in Wuhan is expected to begin in the second half of 2026, at which point analysts project the company will overtake SK Hynix and Micron to become the world's third-largest NAND producer, behind Samsung and Kioxia.

On the other hand, CXMT's trajectory in DRAM mirrors this. The company's monthly DRAM production capacity grew from 100,000 wafers at the start of 2024 to 290,000 wafers by end-2025. It is simultaneously targeting high-bandwidth memory (HBM) production in Shanghai by year-end–the segment that matters most for AI inference workloads, and one where US export restrictions have effectively blocked Chinese buyers from accessing Samsung or SK Hynix supply.

The pricing dynamic adds a further dimension. In a report by South China Morning Post, Arisa Liu, chief director and research fellow at Taiwan Industry Economics Services, said Chinese manufacturers hold a price advantage of over 15% for equivalent specifications, making them particularly attractive to cost-sensitive buyers in general-purpose server and consumer segments.

But that pricing gap is narrowing, and the competitive story is evolving accordingly. In the same SCMP report, MS Hwang, a research analyst at Counterpoint Research, argues Chinese players are now gaining share not solely through discounting, but because they have volume that competitors cannot currently match.

Experts reckon that volume is only going to increase. UBS estimated China's combined memory capacity expansion could reach between 120,000 and 140,000 wafers per month in 2026, with further significant increases projected for 2027. The caveat is timing: new production lines typically take nine to twelve months to reach meaningful yield rates, meaning the real supply impact from current expansions will be felt more acutely in 2027 than in the immediate term.

Funding the scale-up is where China's government strategy becomes a structural advantage rather than a temporary subsidy story. CXMT has earmarked 7.5 billion yuan (approximately US$1.1 billion) from its planned IPO on Shanghai's Star Market specifically for technical upgrades to mass production lines.

YMTC, too, according to Nikkei Asia, is planning to go public in the second half of 2026, with proceeds directed toward equipment and R&D investment. Beyond the capital markets play, Hwang points to Beijing's domestic subsidy programme, incentivising smartphone and device manufacturers to use locally produced memory, as a factor that will compound cost competitiveness over time.

The market context makes all of this matter more than it otherwise might. Memory prices are in a sharp upswing. TrendForce projects DRAM contract prices will rise 58 to 63% in the second quarter of 2026 compared to the previous quarter, with NAND contract prices expected to climb 70 to 75% over the same period.

Some analysts have flagged that customer pre-stocking may be amplifying the apparent demand signal, but others push back. Hwang notes that stockpiling has not yet produced inventory excess, and that multi-year supply agreements between GPU manufacturers, cloud providers, and memory suppliers point to sustained demand rather than a bubble.

Where this leaves Samsung, SK Hynix, and Micron is a harder question than the market share numbers suggest. Price competition from Chinese suppliers is manageable when the volume is limited. It becomes structurally disruptive when the volume is not–and UBS's estimate of 120,000 to 140,000 additional wafers per month, with further expansion in 2027, suggests the volume question is close to being answered.

CXMT still trails by roughly three years in advanced DRAM nodes, and yield on new lines remains the execution risk. But the incumbents built their market positions in a period when Chinese memory was a future concern. That period is ending.