Key facts
- Alibaba unveiled the Zhenwu V900, its most powerful AI chip to date.
- The V900 chip offers three times the performance of its predecessor, the M890.
- The V900 chip features 216 GB of HBM memory and 1.2 TB/s inter-chip bandwidth.
- Mass production of the V900 is scheduled for the first quarter of 2027.
- CXMT's latest chip platform is entering mass production.
- Chinese researchers are developing advanced chips using older technology to circumvent US restrictions.
Chinese technology firms are demonstrating significant advancements in chip manufacturing, particularly in the artificial intelligence sector, despite facing US export controls that limit access to cutting-edge semiconductor technology. Alibaba recently unveiled its Zhenwu V900 chip, which it claims is the most powerful AI chip developed in China to date. This chip boasts three times the performance of its predecessor, the M890, and features 216 GB of HBM memory with 1.2 TB/s of inter-chip bandwidth. Mass production for the V900 is slated to begin in the first quarter of 2027.
In parallel, memory chipmaker CXMT is rapidly gaining market share domestically and has announced that its latest chip platform is entering mass production. Researchers at the Chinese Academy of Sciences are also reportedly making progress in developing highly advanced chips by utilizing older manufacturing technologies, potentially offering a workaround to US restrictions on advanced semiconductor equipment. The Institute of Microelectronics of the Chinese Academy of Sciences (IMECAS) has developed an experimental process for stacked-nanosheet gate-all-around (GAA) transistors using immersion DUV lithography, demonstrating functional devices without relying on EUV scanners.
These developments come as US export controls have effectively cut off Chinese firms from accessing advanced AI silicon from companies like Nvidia. Nvidia CEO Jensen Huang has described these controls as a "failure," noting a significant drop in the company's China data center revenue. Chinese GPU and AI chipmakers have consequently captured a substantial portion of the local AI accelerator server market.
