Alibaba Unveils Zhenwu V900 and Plans 5–10 Trillion-Parameter Model

Alibaba unveiled the Zhenwu V900 AI chip at its annual flagship conference in Hangzhou. [1] Chief Executive Eddie Wu called it “the most powerful AI chip in China today” and said it delivers three times the performance of the previous-generation Zhenwu M890. [1] The chips are used in Alibaba data centers to provide computing for the company and its cloud clients, supporting both model training and inference. [1]

Alibaba plans to train a new artificial intelligence model with five trillion to 10 trillion parameters. [1][3] Its latest Qwen3.8-Max model has 2.4 trillion parameters and is described as the most powerful model in the Qwen series. [1] ABC News said the planned scale would bring Alibaba closer to the most advanced U.S. models, while Bloomberg presented the new chip as competing with Nvidia. [1][2]

Taken together, the announcements pair a higher-performing in-house chip with a planned model containing more than twice and potentially more than four times Qwen3.8-Max’s parameter count. Alibaba’s stack spans models, chips and data centers, while Bloomberg’s Nvidia framing places the hardware announcement in a direct competitive context. Parameter count and company-reported chip performance, however, do not establish how the planned model or V900 will perform in independent testing. What the sources don't address: when the larger model will be completed, whether external benchmarks validate the V900 claims, or how broadly the chip will be available outside Alibaba’s own infrastructure.