Alibaba Strengthens In-House AI Chip Development with New Product Unveiling

Alibaba unveiled its new AI chip Zhenwu M890 on the 20th. The announcement comes as China intensifies development of its own chips to replace Nvidia processors amid U.S. export restrictions.

Developed by Alibaba's semiconductor design subsidiary T-Head, the chip delivers 3x the performance of its predecessor, the Zhenwu 810E. It is optimized for AI "agent" workloads that perform complex multi-step tasks under limited human supervision, and handles environments requiring long context retention and real-time coordination between models well.

Alibaba also unveiled a multi-year chip roadmap:

• V900: scheduled for release in Q3 2027, offering roughly 3x the performance of the M890
• J900: scheduled for release in Q3 2028

This is part of China's effort to expand production of its own AI chips in response to strong U.S. bans on processor sales. Hangzhou-based Alibaba announced last year that it would invest more than 380 billion yuan ($53 billion) in cloud and AI infrastructure over the next three years.

T-Head has shipped more than 560,000 Zhenwu chips to date, with the chips deployed by more than 400 external customers across 20 industries including automotive and financial services.

At the annual Alibaba Cloud Summit, Alibaba also unveiled a new server system, the Panjiu AL128, which packages 128 accelerators into a single rack. This system is immediately available to Chinese enterprise customers through Alibaba Cloud's domestic model platform Bailian.

Alibaba also announced Qwen 3.7-Max, the latest version of its flagship large language model. Designed for advanced coding and long-running agent tasks, it can operate continuously for up to 35 hours without performance degradation.