Alibaba (阿里巴巴) major AI consolidation: waging a new battle centered on tokens
A new hub, a new metric
It has been reported that Alibaba (阿里巴巴) has created a new Alibaba Token Hub (ATH) business group, announced internally on March 16 and placed directly under CEO Wu Yongming (吴泳铭). ATH consolidates Tongyi Lab (通义实验室), the MaaS business line, the Qianwen division (千问事业部), the Wukong division (悟空事业部) and the AI Innovation division, with a single stated goal: “create tokens, deliver tokens, apply tokens.” Short sentence: tokens become the organising principle. Why? Because in large-language-model systems the smallest unit of computation and billing is the token — roughly one Chinese character per token — and commercial scale increasingly maps to token production and consumption, not just model size.
Building an AI “power grid”
ATH is being pitched internally as an end‑to‑end AI supply chain. Tongyi Lab is the “power plant,” producing the Qwen series models and therefore the capacity to generate tokens; MaaS functions as the “grid,” distributing model services via APIs to developers and enterprises; and Qianwen, Wukong and the AI Innovation unit are the “appliances” where tokens are burned in real user scenarios from personal assistants to enterprise workflows. It has been reported that a recent internal product — Coding Plan Tokens — became one of Alibaba Cloud’s fastest‑selling offerings, a sign that token demand is accelerating as AI Agent use cases move from Q&A to task execution.
Why the reorg matters — and what could limit it
On paper this is an organisational move to reduce duplication and concentrate resources; in practice it’s a strategic bet that owning token flows will be a new form of infrastructure in the AI era. Alibaba’s fragmentation — model teams, cloud platform, enterprise apps spread across units — created duplication and fuzzy product identity. The ATH move recalls Google’s earlier recombination of research and product teams aimed at tighter coordination. But geopolitical realities also matter: with U.S. export controls on advanced chips and broader trade frictions, Chinese firms have incentives to focus on software scale, platform distribution and token economics even as hardware constraints persist.
Who controls the tokens may determine competitive advantage. ATH centralises model R&D, platform distribution and application monetisation under one roof. The question now is execution: can Alibaba turn organisational design into sustained API call volume, developer adoption and enterprise deployments fast enough to make tokens its new currency of AI power?