ChangXin Memory Technologies (CXMT) is moving forward with limited production of advanced HBM3E high-bandwidth memory chips despite facing severe manufacturing inefficiencies and low functional yields.
CXMT—China’s largest maker of dynamic random-access memory (DRAM), the conventional memory used in computers and servers has sent HBM3E samples to several domestic AI chip designers.
It expects to begin limited production over the next two months, one person said.
CXMT’s current low output efficiency sit between a stark contrast to global memory leaders.
Yields, or the share of chips coming off a production line that meet specifications, are only about 20% to 30%, compared with yields of around 70% at South Korea’s SK Hynix, the market leader, they said.
This low yield means a vast majority of manufactured chips are defective and must be scrapped, significantly inflating per-unit production costs.
The latest push stems from strict Washington-led export controls implemented to block Chinese firms from purchasing advanced foreign-made semiconductors.
Cut off from standard channels, local tech companies have faced extreme supply squeezes, driving Beijing to prioritize domestic self-reliance regardless of short-term economic losses.
Essential for powering high-performance AI accelerators used to train large language models, HBM technology enables rapid data transfer between memory and processors.
Overcoming domestic hurdles is vital if Chinese designers like Huawei and Cambricon hope to field competitive counterparts to Nvidia hardware.
Moreover, early testing reveals that CXMT’s HBM3E performance tracks closer to older global generations (such as standard HBM3), leaving the firm roughly a generation behind current global market benchmarks.
Whether CXMT can bridge this technological gap and scale commercial viability will depend heavily on stabilizing its fabrication lines.
The decision highlights China’s intense shortage of the AI memory chips, which are essential to a government-backed push to develop domestic alternatives to Nvidia’s AI processors at scale amid US curbs on access to foreign-made technology.
