CXMT to start AI memory output in China despite low chip yields
China's CXMT plans to begin HBM3E production even though only about 20% of its chips are usable, Reuters reports, citing people familiar with the matter.
- published on Newsmetry
- original source published
the public page opens no earlier than 15 minutes after we detect the story. approved stories may reach the api sooner. connect the api →
what happened
China's CXMT plans to start producing HBM3E, an advanced high-bandwidth memory used alongside AI processors, although only around 20% of its chips are usable. Reuters reported this citing two people familiar with the matter. According to them, the company has already sent samples to several domestic AI chip designers and expects limited output within the next two months.
CXMT's yields are estimated at 20% to 30%, versus about 70% at South Korea's SK Hynix, the market leader. Low yields make production costly because many chips must be discarded or reworked. CXMT is moving ahead anyway because Chinese AI chipmakers have struggled to get enough advanced memory, which has become several times more expensive through unofficial channels under US export curbs.
According to Counterpoint Research, SK Hynix held 50% of global HBM revenue in the second quarter, Samsung 33% and Micron 18%. Samsung has also reported expanded HBM4 sales.
why it matters
The signal matters for the market: China is pushing to build domestic AI memory and reduce reliance on foreign suppliers. Still, the sources say CXMT trails SK Hynix by at least one generation, so its near-term impact on the leaders of the HBM market looks limited.
market impact
HBM makers: a rising Chinese rival adds uncertainty for SK Hynix and Samsung
tags
instruments
sources
- China's CXMT to start advanced AI memory production despite poor yields, sources say — Reuters, 28 October 2026 at 04:47 UTC
prepared by the newsmetry newsroom with ai assistance from the sources listed below. not investment advice.
the public page opens no earlier than 15 minutes after we detect the story. approved stories may reach the api sooner.
get stories in real time ↗