Alibaba launches next-generation AI chip to take on Nvidia

Alibaba Group on Wednesday introduced a new artificial intelligence chip, the Zhenwu M890, as the Chinese tech giant steps up efforts to develop domestic alternatives to processors made by Nvidia amid stricter U.S. export restrictions, News.Az reports, citing Reuters.

The chip was developed by Alibaba’s semiconductor design arm, T-Head, and offers three times the performance of its previous-generation Zhenwu 810E chip.

It has been specifically designed for the growing generation of AI “agents” — software systems capable of performing complex, multi-step tasks with minimal human supervision.

Alibaba said the new processor is designed to manage the demanding memory and communication requirements of agent-based workloads, where AI models must maintain extended context and coordinate with each other in real time.

The company also revealed a multi-year chip development roadmap, stating that the M890 will be followed by a next-generation processor called the V900 in the third quarter of 2027, and another chip, the J900, in the third quarter of 2028. According to Alibaba, the V900 is expected to deliver another roughly threefold performance increase over the M890, highlighting the company’s long-term commitment to advancing its in-house chip technology.

The announcement reflects China’s broader push to develop locally produced AI chips as Washington continues to restrict the sale of the most advanced U.S. processors to Chinese companies. The move follows a similar initiative announced by Huawei last year.

The Hangzhou-based company previously pledged to invest more than 380 billion yuan (around $53 billion) in cloud computing and AI infrastructure over a three-year period, marking its largest-ever investment commitment in the sector.

The investment signals a wider expectation across China’s technology industry that demand for AI computing power will continue rising as businesses increasingly adopt agent-based AI applications.

Alibaba unveiled the chip during its annual Alibaba Cloud Summit, where it also introduced a new server platform called the Panjiu AL128. The system combines 128 AI accelerators into a single rack configuration.

Alibaba said the system is available immediately to enterprise customers in China through Alibaba Cloud’s domestic AI model platform, Bailian.

T-Head added that it has shipped more than 560,000 Zhenwu chips so far, with over 400 external customers across 20 industries — including automotive and financial services companies — already deploying the processors.

Alibaba also announced Qwen 3.7-Max, the newest version of its flagship large language model, which the company said is optimized for advanced coding and long-duration agent tasks. According to Alibaba, the model can run continuously for up to 35 hours without experiencing performance degradation.

News.Az 

By Nijat Babayev