Alibaba Unveils New AI Chip “Zhenwu M890” to Advance Domestic Semiconductor Push

On May 20, 2026 Alibaba Group has announced its latest artificial intelligence accelerator, the Zhenwu M890, marking a significant step in China’s drive toward semiconductor self-sufficiency amid tightening U.S. export restrictions on advanced AI hardware. According to Reuters, the new chip—developed by Alibaba’s semiconductor subsidiary T-Head, delivers approximately three times the performance of its predecessor, the Zhenwu 810E, and is optimized for modern “agentic AI” workloads that require long-context reasoning and multi-step task execution .

The chip is designed to handle both training and inference workloads, with enhanced memory capacity and interconnect bandwidth tailored for large-scale AI systems. Alibaba stated that the M890 is part of a broader roadmap, with successor chips (V900 and J900) planned for 2027 and 2028 respectively, signaling sustained investment in in-house silicon development .

This launch aligns with China’s broader strategy to reduce reliance on foreign suppliers such as Nvidia, while strengthening domestic AI infrastructure across cloud and enterprise sectors. Analysts view the move as part of an accelerating global trend toward vertically integrated AI ecosystems combining chips, models, and cloud platforms.

To understand the positioning of Alibaba’s Zhenwu M890, it is useful to compare it against NVIDIA’s H100, which remains one of the most widely adopted and influential AI accelerators in the global market. While the H100 has become the industry benchmark for large-scale AI training and inference workloads, the M890 represents Alibaba’s push toward building a competitive domestic alternative optimized for China’s AI infrastructure needs, particularly in agentic and long-context workloads. The comparison highlights both the technological ambitions of Alibaba’s semiconductor roadmap and the current gap between emerging domestic accelerators and established global leaders. 

Reference: Reuters