Wu Yongming (吴泳铭) makes cuts — Alibaba (阿里) wants to be the power grid of the AI era

New hub, new bet

Alibaba (阿里) has announced a sweeping reorganization that elevates AI to the same formal status as its e‑commerce and cloud businesses. In an internal letter that reportedly reverberated across China’s tech community, Wu Yongming (吴泳铭) unveiled the Alibaba Token Hub (ATH) as a full‑level core business group he will personally lead. ATH’s stated mission — “create Token, deliver Token, apply Token” — frames tokens produced by large models as the primary medium for the coming era of billions of AI agents. In plain terms: Alibaba aims to stop selling models and start selling the fuel those models consume.

From platforms of trade and compute to a token supply chain

To Western readers: Alibaba long ago platformized commerce and later compute — think of marketplaces and cloud as utilities. Damo Academy (达摩院) and Tongyi Lab (通义实验室) gave Alibaba research heft; now ATH bundles model R&D, MaaS (model‑as‑service) delivery, consumer app 千问 (Qianwen), enterprise arm 悟空 (Wukong) and an AI innovation unit into a single “supply chain” for tokens. The logic is simple. If model capabilities converge and raw compute becomes commoditized, whoever can generate, route and meter token consumption at scale will control pricing power — effectively becoming the power grid of the AI era.

Execution risks, internal friction and geopolitical headwinds

The gamble is large. Longstanding “department walls” and differing team incentives could blunt ATH’s intended synergy — history shows Alibaba’s 2018 “1+6+N” reform did not eliminate internal siloing. It has been reported that Tongyi Qianwen’s figurehead Lin Junyang (林俊旸) recently departed, a signal of the tension between scientific leadership and industrial engineering priorities. ATH also rests on a fragile premise: that Alibaba’s models will stay ahead. If open‑source projects or nimble rivals leapfrog core capabilities, ATH risks building an expensive distribution network for an obsolete product. There is a geopolitical overlay too: export controls and chip sanctions constraining China’s access to top‑end accelerators raise the stakes for domestic consolidation and for strategies that prioritize efficient token use over raw horsepower.

A platform play with a simple question

ATH is as much an organizational bet as a technical one. If Alibaba can clear internal frictions, keep Tongyi competitive, and convert token consumption into predictable revenue, it could reshape how Chinese enterprises buy AI — and force competitors to consolidate. Can Alibaba become the AI “power grid”? Execution will tell, and the industry is already watching to see whether this is a historic platform shift or another costly experiment in a long history of corporate reorganizations.