Deep Tide TechFlow, November 19 — At Alibaba Group’s analyst call for its second-quarter fiscal 2027 results, CEO Eddie Wu said that Qwen has trained its next-generation model on T-Head chips, and Bailian has provided model inference services based on domestically produced chips.
Wu said that T-Head’s three-generation GPU roadmap is accelerating: Zhenwu M890 will be fully rolled out this year, V900 is slated for 2027, and J900 for 2028. The related production supply chain is advancing and moving into large-scale mass production.
Alibaba’s self-developed computing infrastructure will gradually support model training, inference, and agentic tasks for both the group and Alibaba Cloud customers, and will serve as a foundation for driving AI revenue growth.
Wu said that T-Head’s three-generation GPU roadmap is accelerating: Zhenwu M890 will be fully rolled out this year, V900 is slated for 2027, and J900 for 2028. The related production supply chain is advancing and moving into large-scale mass production.
Alibaba’s self-developed computing infrastructure will gradually support model training, inference, and agentic tasks for both the group and Alibaba Cloud customers, and will serve as a foundation for driving AI revenue growth.
