
Chinese technology company Alibaba Group announced Wednesday the launch of its newest artificial intelligence processor, the Zhenwu M890, as the firm accelerates efforts to create homegrown alternatives to processors amid stricter U.S. trade restrictions.
The processor, created by Alibaba’s chip design division T-Head, provides triple the performance capabilities compared to its earlier version, the Zhenwu 810E. The chip was specifically designed for the next generation of AI “agents” — computer programs capable of executing complicated, multi-phase operations with minimal human intervention.
According to Alibaba, the latest processor excels at managing the substantial memory requirements and communication needs of agent-based computing tasks, where AI models must maintain extensive contextual information and work together simultaneously.
The technology giant also revealed its long-term processor development strategy, announcing plans to release a follow-up chip named the V900 during the third quarter of 2027, followed by another processor called the J900 in the third quarter of 2028. The V900 is projected to provide approximately three times the performance improvement over the M890, demonstrating Alibaba’s commitment to continuous internal chip advancement.
This strategy highlights China’s expanding initiatives to develop domestically-produced AI processors as Washington prohibits sales of the most advanced U.S. chips to Chinese buyers, following a comparable announcement by another Chinese company last year.
The Hangzhou-based technology firm committed last year to investing more than 380 billion yuan ($53 billion) in cloud computing and AI infrastructure over a three-year period, representing its largest financial commitment to this sector.
This investment demonstrates a wider industry belief across China’s technology sector that AI computing demand will continue growing as businesses implement agent-based software solutions.
Alibaba revealed the chip during its yearly Alibaba Cloud Summit, along with introducing a new server configuration called the Panjiu AL128, which combines 128 of these processors into one rack system.
The server system became immediately available to Chinese business customers via Alibaba Cloud’s domestic platform, called Bailian.
T-Head reported delivering more than 560,000 Zhenwu processors so far, with over 400 external clients spanning 20 different sectors, including automotive manufacturers and financial services companies, currently using these chips.
Alibaba additionally announced Qwen 3.7-Max, the newest iteration of its primary large language model, which the company stated was built for sophisticated programming tasks and extended agent operations. The firm claimed this model can function continuously for up to 35 hours without any decline in performance.







