Alibaba's (阿里巴巴) AI Shift: Organizational Growing Pains and the Token War Behind It

Earnings shock, then a pivot

Alibaba (阿里巴巴) delivered a weaker‑than‑expected Q3 for fiscal 2026 and the market punished it: U.S. trading opened with the stock plunging nearly 10%, knocking the company's market cap below $300 billion. Revenue was RMB 2848.4 billion, up 2% year‑on‑year (9% on a like‑for‑like basis excluding Gaoxin Retail and Intime), but profit collapsed — net profit RMB 156.3 billion, down 66%, non‑GAAP net profit RMB 167.1 billion down 67%, and adjusted EBITA off 57%. China e‑commerce growth was effectively flat. Can an AI story re‑ignite confidence when the traditional cash engine is stalling?

ATH and the token thesis

Two days before the report, Alibaba formally established the Alibaba Token Hub (ATH) as a top‑level business group and put CEO Wu Yongming (吴泳铭) in charge, folding Tongyi Lab (通义实验室), MaaS, the Qianwen (千问) and Wukong (悟空) divisions and an AI innovation unit under one roof. ATH’s mantra — “create tokens, deliver tokens, apply tokens” — elevates the LLM token (the minimal unit a model processes) to the centre of Alibaba’s commercial calculus. It has been reported that ATH is as much a political move as a product strategy — a deliberate re‑centralization to break the “warlord” model of scattered AI teams and speed feedback from real business scenarios into model iteration. Alibaba has already pushed Qwen3‑Max‑Thinking and is betting MaaS will become its largest cloud revenue driver by monetising token consumption like utilities such as water or electricity.

The test: execution, economics, and geopolitics

The idea is simple; the execution is hard. ATH must dismantle entrenched silos, harmonise KPIs across research and products, and sustain heavy R&D spending while converting token consumption into durable revenue. Competitors ByteDance (字节跳动) and Tencent (腾讯) are consolidating their own agent and enterprise plays; industry heavyweights such as OpenAI and Microsoft have already normalised token billing. Add geopolitics: export controls on high‑end AI chips and global supply constraints make compute both strategic and costly. Management has set an ambitious target — it aims for cloud and AI commercial revenue (including MaaS) to exceed $100 billion within five years — but investors want proof that ATH will turn tokens into cash, not just a new metric. Will the Token Hub finally make Alibaba a utility for AI, or is this a rebrand of old struggles? Only execution will tell.