Alibaba’s CEO Wu Yongming personally leads the establishment of the Token Foundry division, integrating the Tongyi Large Model division and the Future Life Lab to further focus on native model development and multimodal model production. Meanwhile, Zhou Jingren, head of the Tongyi Large Model division, has been appointed Alibaba’s Chief Scientist and will lead the formation of the AI Future Research Institute. Vice President Zheng Bo joins with his team, bringing the Happy Horse video generation large model and the Happy Oyster open-world model. The Qwen-3.7-Max model ranks second globally in the Code Arena evaluation. Last quarter, AI product revenue reached RMB 8.971 billion, accounting for over 30% of external commercial revenue, with an annualized revenue exceeding RMB 35.8 billion, marking AI’s entry into a large-scale commercial return phase.
Author and source: Zhixidong
Alibaba has established Token Foundry, with Wu Yongming personally leading the initiative.
Alibaba has announced the merger of the Tongyi Large Model Division and the Future Life Lab to form the Token Foundry Division, directly overseen by Wu Yongming.
This reorganization involves multiple AI initiatives. Zhou Jingren, Alibaba Partner and Head of the Tongyi Large Model Division, has been appointed as Alibaba’s Chief Scientist and will lead the establishment of the Alibaba AI Future Research Institute, focused on exploring cutting-edge AI technologies. Previously, on April 8, Li Feifei succeeded Zhou Jingren as CTO of Alibaba Cloud. The title of Chief Scientist is the highest academic position in Alibaba’s technology structure; this appointment recognizes his contributions and represents a strategic move toward the future.
Meanwhile, Zheng Bo, Vice President of Alibaba, has joined the Token Foundry division alongside teams including Happy Horse and Happy Oyster. Zheng joined Alibaba in September 2017, after working at Google from 2006 to 2017. He was promoted to M6 (Vice President) level in January 2022 and previously served as Vice President of Technology, Head of Algorithm Technology, and CTO of Alimama.
Just three months after establishing the ATH (Alibaba Token Hub) business group in March this year, Alibaba has once again restructured its AI organization. Led personally by Wu Yongming, the ATH business group integrates five key areas—Tongyi Lab, MaaS Platform, Qwen, Wukong, and AI Innovation—forming a new organization with the core mission of “creating tokens, delivering tokens, and applying tokens.”
If the ATH division focuses on horizontally coordinating token circulation and resource allocation across the entire group, the newly established Token Foundry division is more akin to a specialized hub for native model development and multimodal model production—serving as Alibaba’s core “foundry” for model capacity.
By merging two core R&D units, Alibaba will more efficiently allocate computing resources, prioritizing the training and iteration efficiency of models, thereby optimizing computing resource allocation for model development and avoiding redundant construction and resource waste.
This is Ali's continued investment in the field of AI. On May 20, the Alibaba Cloud Summit unveiled a comprehensive Agent-oriented blueprint spanning "chip-cloud-model-inference." On the same day, Zhou Jingren announced Alibaba's latest Qwen-3.7-Max model, which ranks second globally and first among Chinese models in the Code Arena benchmark, and leads in LiveCodeBench v6, earning widespread recognition.
The Happy Horse and Happy Oyster models introduced by Zheng Bo’s team are also among the pioneers of this integration. Happy Horse, a large video generation model, leads in multiple AI video evaluations; Happy Oyster is Alibaba’s first open-world model, built on a native multimodal architecture that enables the generation of a continuously running physical world from text or images, allowing users to interact in real time and switch perspectives with superior consistency compared to similar products.
Last quarter’s financial report showed that Alibaba’s AI products generated quarterly revenue of RMB 8.971 billion, with external commercial revenue accounting for over 30% for the first time, translating to an annualized revenue exceeding RMB 35.8 billion. Wu Yongming stated on the earnings call: “AI has moved beyond its initial investment phase and entered a period of large-scale commercial returns.” From this perspective, the significance of Token Foundry may extend beyond merely consolidating computing resources—it also lies in packaging Alibaba’s foundational model capabilities into directly commercializable token products. The model-as-a-product paradigm is becoming a trend.
Globally, AI competition is shifting from model rivalry to token economy competition. OpenAI, Google, and Microsoft are all exploring ways to transform model capabilities into scalable, on-demand services, with tokens emerging as the core medium connecting computing power, models, agents, and applications. Through Token Foundry, Alibaba is consolidating R&D resources and optimizing computing allocation, potentially accelerating model iteration and commercialization through a more streamlined organizational structure.
This article is from the WeChat public account "Zhi Dongxi" (ID: zhidxcom), authored by Li Shuiqing, republished with permission from 36Kr.