Physix Frontier · News Briefing Card (IT Home · Nov 19, 2026)

Alibaba's Wu Yongming: Next-gen Qwen model trained on T-Head chips

KEY FACTS

  • Wu Yongming disclosed at the fiscal 2027 second-quarter earnings call that the next-generation Qwen model has been trained on T-Head chips.
  • The Bailian platform is already providing external model inference services based on domestic chips.
  • Wu Yongming said that co-designing in-house chips with in-house models can lower cost per unit of compute.
  • T-Head confidentially filed an A1 listing application with the Hong Kong Stock Exchange in September 2026, and the spin-off has been confirmed by the exchange.
  • The Zhenwu V900 chip is planned for mass production in the first quarter of 2027, with three times the performance of the previous generation.

KEY DATA

271.519 billion yuanGroup revenue this quarter
260%Adjusted EBITA year-over-year
67%Alibaba Cloud external recurring revenue year-over-year
216GBZhenwu V900 memory

PHYSIX OBSERVATION

With both training and inference moved onto in-house chips, Alibaba has pushed the model-chip closed loop from slogan to production line; cost per unit of compute is the earnings story it really wants to tell. With the T-Head spin-off IPO imminent, this synergy narrative also directly supports its valuation; but whether the domestic supply chain can withstand the training intensity of the next-generation large model still awaits real-world validation by Qwen4.

Source: IT Home report