Alibaba Group Holding Limited (NYSE:BABA) on Wednesday introduced a new artificial intelligence chip, the Zhenwu M890, as the company accelerates efforts to develop Chinese-made alternatives to processors supplied by NVIDIA Corporation amid tighter U.S. export restrictions.
The processor was developed by Alibaba’s chip-design unit T-Head and delivers performance roughly three times greater than the previous-generation Zhenwu 810E. The chip has been specifically designed for the next generation of AI “agents,” software systems capable of handling complex multi-step tasks with minimal human supervision.
Designed for advanced AI agent workloads
Alibaba said the Zhenwu M890 is optimised for workloads requiring intensive memory use and high-speed communication, particularly for AI agent systems that must process long context windows and coordinate with other models in real time.
The group also presented a long-term semiconductor roadmap, revealing plans to launch a successor chip called the V900 in the third quarter of 2027, followed by another processor, the J900, in the third quarter of 2028.
According to Alibaba, the V900 is expected to provide another approximate threefold improvement in performance compared with the M890, highlighting the company’s intention to maintain a consistent pace of internal chip development.
China pushes for greater AI chip independence
Alibaba’s latest announcement reflects broader efforts across China’s technology sector to reduce reliance on foreign AI semiconductors as Washington continues restricting sales of advanced U.S.-made processors to Chinese companies.
The move follows similar initiatives announced by Huawei Technologies Co., Ltd. last year.
The Hangzhou-based company previously pledged to invest more than 380 billion yuan ($53 billion) over three years into cloud computing and artificial intelligence infrastructure, marking the largest technology investment commitment in Alibaba’s history.
The spending plan reflects growing expectations across China’s technology industry that demand for AI computing power will continue rising rapidly as businesses increasingly adopt agent-based AI applications.
Alibaba unveils new server system and upgraded language model
Alibaba presented the new chip during its annual Alibaba Cloud Summit alongside a server platform called the Panjiu AL128, which integrates 128 Zhenwu accelerators into a single rack system.
The platform is immediately available to enterprise customers in China through Alibaba Cloud’s domestic AI model platform, Bailian.
T-Head said more than 560,000 Zhenwu chips have already been shipped, with over 400 external customers across 20 industries — including automotive and financial services companies — currently using the processors.
Alibaba also introduced Qwen 3.7-Max, the latest version of its flagship large language model. The company said the model has been specifically engineered for advanced coding applications and long-duration AI agent operations.
According to Alibaba, the model is capable of operating continuously for up to 35 hours without any decline in performance.