TechFlow reports that on November 19, Alibaba Group CEO Wu Yongming stated during the second-quarter fiscal 2027 earnings analyst conference that Qwen has already trained next-generation models on T-Head chips, while Bailian is already providing model inference services based on domestically produced chips.
Wu Yongming noted that T-Head's three-generation GPU roadmap is accelerating: Zhenwu M890 went fully online this year, V900 aligns with 2027, and J900 aligns with 2028, with the related manufacturing supply chain currently advancing toward mass production.
Alibaba's self-developed computing infrastructure will gradually host model training, inference, and Agentic tasks for the group and Alibaba Cloud customers, serving as the foundation for driving AI revenue growth.





