Alibaba said on July 8 that it is merging its Tongyi LLM business unit with Future Life Lab to form a new Token Foundry business unit, reporting directly to Group CEO Eddie Wu. The move marks the company’s third major AI reorganization since the launch of ATH in March, showing a rapid pace of internal reshuffling.
From ATH to Token Foundry, Alibaba redraws AI responsibilities
According to the timeline disclosed by the company, Wu announced the creation of ATH on March 16 in an internal letter, combining assets such as Tongyi Lab, the MaaS business line, and the Qwen app. Alibaba also said it would use “Token” as a unified metric for measuring AI business ROI. On April 8, Tongyi Lab was upgraded into the Tongyi LLM business unit under Zhou Jingren. With the latest change, the company is separating commercialization from research more clearly, assigning Token Foundry to AI product deployment and ecosystem building.
The new structure places the commercialization side of Alibaba’s AI push closer to the core management system. Token Foundry will focus on product execution and ecosystem expansion, while frontier research will be handled through a newly established institute led by a different executive setup.
Zhou Jingren becomes chief scientist, Zheng Bo’s team joins
Zhou Jingren, who has long led the Qwen model line, will become Alibaba’s chief scientist and take charge of building the Alibaba AI Future Research Institute. The source describes this title as the highest academic role within Alibaba’s technology system. Zhou studied at the University of Science and Technology of China’s special class for gifted youth, earned a PhD in computer science from Columbia University, and previously worked at Microsoft and Alibaba Cloud before becoming an Alibaba partner last year.
Alibaba vice president Zheng Bo will bring his team into Token Foundry as well. That includes projects such as Happy Horse and Happy Oyster. The source identifies Happy Horse as a multimodal model that drew attention in the AI community in April, while Happy Oyster is described as an in-house open world model capable of generating continuous video content in real time from text, voice, and images.
Qwen-3.7 ranking and earnings disclosure set the backdrop
The reorganization comes as Alibaba says its AI business is beginning to show stronger market traction. The latest Qwen-3.7 ranked among the global top three in coding capability and placed first in the Chinese market, according to the source. Alibaba’s latest Q4 earnings report also disclosed for the first time that its AI business had moved beyond the initial investment stage and entered a commercialization return cycle.
Across the three changes made since March, the pattern is consistent: Alibaba has been consolidating AI resources, standardizing internal measurement around token-based ROI, and tightening the link between model development and commercial execution. What is clear from the announced structure is that Tongyi, lab resources, and selected innovation teams are now being regrouped under a CEO-led operating framework.