HappyHorse Launches 'Counterstrike' — Alibaba's First Shot After a Major Restructure

The reveal: HappyHorse (快乐马) now wears the Alibaba badge

Alibaba (阿里巴巴) has publicly claimed the hotly discussed video model HappyHorse (快乐马), it has been reported that the project sits inside the newly formed Alibaba Token Hub (ATH, 阿里AI核心事业群). Observers had already suspected the link. It has been reported that two threads of evidence seeded that view: HappyHorse's technical path mirrors Alibaba’s Tongyi Lab (通义实验室) work, and the return of model architect Zhang Di (张迪) allegedly accelerated development — though other sources say ATH’s Zheng Bo (郑波) team actually built the product. Which is it? The answer is less important than the signal: ATH is now executing.

A rapid organizational shove toward integration

ATH’s emergence came alongside a second, faster move — a new Group Technology Committee charged with top-level design and resource coordination, chaired by Wu Yongming (吴泳铭) and including Zhou Jingren (周靖人), Wu Zeming (吴泽明) and Li Feifei (李飞飞). The appointments are deliberate. Zhou is positioned to “own” model sovereignty; Li Feifei combines cloud and AI infrastructure leadership; Wu Zeming is tasked with embedding inference into core businesses like Taobao and Alipay. Short sentence: Alibaba is trying to stop teams from pulling in different directions.

Why this matters — commercial closure, not just engineering

This is about turning models into money. Alibaba has been strong in cloud, e‑commerce and research, but those strengths were siloed. What changes now is the push to convert model capacity and compute into a token‑based commercial loop — create tokens, move tokens, monetize tokens — and to do so at group scale. It has been reported that CEO Wu Yongming has set a five‑year target for cloud and AI commercial revenue to exceed $100 billion. Ambitious? Yes. Achievable only if the group can translate this reorg into faster model iteration, tighter cloud integration and repeatable business deployments.

Bigger picture: competition, geopolitics and execution risk

Globally, cloud vendors are bundling models and infrastructure; China’s internet giants are racing to build their own closed loops. That drive is amplified by U.S. export curbs on advanced chips and broader tech tensions, which have pushed Chinese firms toward greater self‑reliance in models and compute. But strategy is one thing; execution is another. If ATH and the Group Technology Committee can break down old department walls, Alibaba could recast itself from an e‑commerce platform into a platform that sells models, compute and applications at scale. If they fail, this will join a long list of well‑intentioned but ineffective restructurings. Either way, HappyHorse is the first, visible test.