Original Exclusive | Alibaba (阿里) makes another big move, forms a Token-focused business group personally led by Wu Yongming (吴泳铭)

The move

Alibaba (阿里) has reportedly reorganized major AI assets into a new, token-focused business group called Alibaba Token Hub (ATH), and CEO Wu Yongming (吴泳铭) will directly lead it. This is being framed internally as the company’s deepest integration since the last big e‑commerce reshuffle. Short sentence: tokens are now core infrastructure. Longer sentence: ATH is said to sit alongside Alibaba Cloud Intelligence and the e‑commerce group as an independent business group, unifying model R&D, model serving (MaaS), consumer and enterprise AI products under one management chain.

Structure and purpose

Reportedly ATH will comprise five divisions: Tongyi Lab (通义实验室, the Qwen large‑model team), Bailian MaaS (百炼MaaS), Qianwen (千问, C‑end AI products), Wukong (悟空, the DingTalk/钉钉‑centric enterprise AI unit) and an AI innovation division. Internally this is described as a “4+N” architecture where Tongyi, Bailian, Qianwen and Wukong form the base, and additional AI projects flow into the N. ATH’s stated mission is to treat “Token” as the key production element — to create tokens (models), deliver tokens (MaaS/platforms) and apply tokens (apps and agents) across B2B and B2C scenarios.

Why Tokens matter — and the stakes

What exactly is a token? In large language models it’s the minimal semantic unit used for understanding and generation (in Chinese input roughly one character ≈ one token). It has been reported that internal demand for tokens — driven by AI agents that can run 24/7 — has already spiked, and Alibaba’s Coding Plan tokens were its fastest‑selling cloud product recently. If cloud users become “AI agents” rather than human engineers, token consumption could explode. Will ATH be the industrial system to feed that demand? That is the bet leadership is making. It has been reported that some senior engineers, including former Qwen lead Lin Junyang (林俊旸), preferred keeping Qwen more independent rather than folded into a horizontally integrated hub.

Context: competition, chips and historical echoes

This reorg matters beyond Alibaba’s org chart. In an era of US export controls on advanced chips and intensified US‑China tech rivalry, Chinese cloud and internet companies are racing to own more of the AI stack and the production ecosystem — from models to platforms to end applications. Google’s 2023 merger of Google Brain and DeepMind is a useful precedent: tech giants are consolidating to reduce internal friction and focus on “industrial” AI competitiveness. Reportedly, ATH is Alibaba’s answer to that same logic — centralize token creation, delivery and consumption to compete in the next phase of AI.