Alibaba Unveils In-House AI Chip 'Zhenwu M890' to Reduce Dependence on Nvidia

by AJP Posted : May 21, 2026, 12:18Updated : May 21, 2026, 12:18
Photo from Getty Images
[Photo from Getty Images]

Alibaba has unveiled its in-house artificial intelligence (AI) chip, the 'Zhenwu (젠우) M890,' as part of a strategy to reduce reliance on foreign high-performance AI semiconductors like those from Nvidia amid ongoing U.S. export restrictions on advanced semiconductors to China.

On May 20, during the Alibaba Cloud Summit held in Hangzhou, the company showcased the next-generation AI chip, which was developed by its semiconductor design subsidiary, T-Head. According to the company, the performance of the Zhenwu M890 is three times greater than that of its predecessor, the Zhenwu 810E. The Wall Street Journal reported that this chip features 144GB of GPU memory and supports both AI training and inference tasks, unlike previous models. Alibaba stated, "The new chip is suitable for agent-type AI tasks that require processing multiple stages and utilizing various AI models simultaneously."

Alongside the chip, Alibaba also introduced a new server and a large language model (LLM). The 'Panjiu (판지우) AL128' server is a high-performance system that combines 128 Zhenwu M890 chips. Reuters reported that this system will be offered to corporate clients through Alibaba Cloud's AI model platform in China.

The next-generation LLM, 'Qwen3.7-Max,' was also launched. This model is designed for complex coding, intricate reasoning, and multi-stage tasks that require extended processing time. The company noted that the Qwen3.7-Max and the infrastructure based on the Zhenwu M890 can handle complex AI tasks for up to 35 hours without performance degradation.

Alibaba also presented its AI chip roadmap, planning to release a follow-up product, the V900, in the third quarter of 2027, and the J900 in the third quarter of 2028. The V900 is expected to achieve approximately three times the performance of the Zhenwu M890. T-Head has reportedly shipped over 560,000 units of the Zhenwu product line, with more than 400 clients across 20 industries, including automotive and finance, adopting the technology.

This announcement aligns with the trend of China's big tech companies striving for semiconductor independence. As the U.S. continues to limit exports of high-performance AI chips to China, companies like Alibaba and Huawei are accelerating their in-house chip development. Previously, Alibaba announced plans to invest 380 billion yuan (approximately $53 billion) over three years in cloud and AI infrastructure.





* This article has been translated by AI.