At the 2026 Yunqi Conference, Pingtouge under Alibaba unveiled the domestic AI chip Zhenwu V900, with compute power increased to three times that of the prior-generation Zhenwu M890. A single cluster built on the Zhenwu V900 can be expanded to 500,000 cards, providing large-scale compute power for training and inference of cutting-edge AI models. Alibaba Group CEO Wu Yongming said that the company’s self-developed M890 AI supernode currently can support inference for trillion-parameter (20 trillion) large models, and that it has been scaled and deployed on Alibaba Cloud data centers this quarter. At the same time, Alibaba open-sourced the Qianwen 2.7B model, launched a Qianwen mobile AI solution, and is simultaneously advancing chip development, ultra-large-scale clusters, cloud infrastructure, and edge model ecosystem building—delivering end-to-end compute coverage across the full pipeline from training and inference to terminal deployment.

Article author, source: Yunqi Conference

At the 2026 Yunqi Conference, Pingtouge under Alibaba unveiled the domestic AI chip Zhenwu V900, with compute power increased to three times that of the prior-generation Zhenwu M890. Built on the Zhenwu V900, a single cluster can scale to 500,000 cards, supporting large-scale compute for training and inference of cutting-edge AI models. Alibaba said that as the Pingtouge chip product line gradually matures and sees broad customer adoption, the company expects annual chip shipments to increase significantly.

Alibaba Group CEO Wu Yongming revealed that Alibaba’s self-developed M890 AI supernode currently has the capability to support inference for trillion-parameter (20 trillion) large models. The supernode has already been scaled and made available on Alibaba Cloud data centers this quarter. While continuing to advance its chips and cloud infrastructure, Alibaba has also further rolled out an edge AI ecosystem: it open-sourced the Qianwen 2.7B model and officially launched a Qianwen mobile AI solution.

From the Zhenwu V900 to the M890 AI supernode, Alibaba is advancing AI chips, ultra-large-scale clusters, cloud infrastructure, and edge-side model ecosystem development in parallel—covering different compute needs for cutting-edge models across training, inference, and terminal deployment.