Alibaba Launches Token Hub事业群 Under CEO Wu Yongming
Alibaba Token Hub (ATH)事业群 will operate as a peer-level organization alongside Alibaba Cloud Intelligence Group and E-commerce Group.
Strategic Reorganization for the AI Era
Alibaba has initiated a foundational restructure to position itself at the forefront of the next-generation AI economy — where Token is no longer just a technical unit, but the essential infrastructure: the “water, electricity, and coal” of AI productivity.
According to exclusive reporting by Huxiu, on March 16, 2026, Alibaba formally established the Alibaba Token Hub (ATH)事业群, a new, top-tier business group directly overseen by CEO Wu Yongming. This marks the most consequential organizational shift since the consolidation of its e-commerce business units.
The name “ATH” carries dual significance:
– Alibaba Token Hub — reflecting its functional mission;
– All Time High — symbolizing aspirational growth in valuation, influence, and technological leadership.
Organizational Architecture: The 4+N Framework
ATH operates as an independent, parallel事业群 — equal in stature to Alibaba Cloud and the E-commerce Group. Its core structure follows a 4+N model: four foundational pillars anchored by continuous innovation streams.
Core事业群 Units under ATH:
| Unit | Focus | Key Mission |
|---|---|---|
| Tongyi Lab | Foundational Model R&D | Develop cutting-edge multimodal foundation models (e.g., Qwen series); push the frontier of base model capabilities for Alibaba and global industry. |
| MaaS Business Line | Enterprise Model-as-a-Service | Build an open, high-efficiency model platform and technical stack to empower AI ecosystems across industries. |
| Qwen Division | Consumer AI Assistants | Deliver the world’s best personal AI assistant — intuitive, reliable, and deeply integrated into daily life. |
| Wukong Division | B2B AI-Native Work Platform | Embed AI natively into enterprise workflows via DingTalk; transform how teams collaborate, plan, and execute. |
| AI Innovation Division | Experimental Applications | Rapidly prototype and validate novel AI use cases across verticals — from creative tools to domain-specific agents. |
✅ Key Insight: ATH unifies previously siloed AI efforts — breaking down “BU walls” to enable seamless collaboration between Tongyi (model creation), MaaS (token delivery), and Qwen/Wukong (token application).
The Token Lifecycle: Creation → Delivery → Application
ATH is explicitly architected around the full lifecycle of Token — the atomic unit of AI processing and cost:
🔹 Creation
- Led by Tongyi Lab: Produces high-performance base models (e.g., Qwen) — the upstream source of semantic tokens.
🔹 Delivery
- Managed by MaaS: Translates models into production-ready APIs, SDKs, fine-tuning tools, and deployment frameworks — distributing token capacity to developers and enterprises.
🔹 Application
- Executed by Qwen (C-end), Wukong (B-end), and Innovation Division: Builds end-user products where tokens are consumed — powering chat, coding, automation, and domain-specific reasoning.
This end-to-end orchestration transforms AI from isolated experiments into an industrial-scale, feedback-driven ecosystem — where model improvements directly accelerate application performance, and usage data fuels further model refinement.
Why Now? The Token Demand Inflection Point
The ATH launch responds to explosive real-world demand:
- During the recent OpenClaw (“Lobster”) AI wave on Bilibili and Xiaohongshu, users repeatedly voiced: “Tokens are running out”, “Great model — too expensive per token.”
- In Chinese AI systems, ~1 Chinese character ≈ 1 token; pricing and quotas are universally tied to token volume.
- Alibaba Cloud’s Coding Plan became the fastest-selling product in company history during Spring Festival 2026 — so popular that its “first-purchase discount” was retired after just two weeks.
CEO Wu Yongming articulated three strategic imperatives in a post-holiday internal address:
- AI Coding has crossed the inflection point: Most application systems can now be built or maintained by models — shifting cloud users from human engineers to AI agents. The future cloud is not just for humans, but for AI.
- AI Agents drive exponential token consumption: Always-on, autonomous agents generate massive, sustained token demand — requiring scalable, cost-efficient infrastructure.
- Democratization of AI: Beyond large enterprises, small businesses (“small B”) and super-individuals will adopt AI tools — demanding accessible, intuitive, and affordable offerings.
Global Context: Learning from Google DeepMind
Alibaba’s move mirrors broader industry consolidation trends:
- In 2023, Google merged Brain and DeepMind, eliminating internal fragmentation between product engineering and frontier research. The unified “Google DeepMind” accelerated Gemini’s development and deployment.
However, while Google unified R&D and productization, Alibaba’s ATH unifies token economics and workflow integration — emphasizing practical usability, cross-functional agility, and production-grade scalability.
Leadership & Challenges Ahead
- ATH is led by CEO Wu Yongming, signaling its strategic centrality.
- Notably, former Qwen technical lead Lin Junyang reportedly advocated for Qwen’s independence — highlighting tensions between open-research autonomy and cross-group operational synergy.
- Success hinges not only on structure, but on aligning incentives, shared KPIs, and agile governance to sustain momentum across research, platform, and product teams.
Internal Memo: Wu Yongming’s ATH Announcement
To all Alibaba colleagues,
We stand at the dawn of the AGI era. Billions of AI Agents will soon power digital work — each fueled by tokens generated by models. Tokens are becoming the primary medium through which humans interact with the digital world.
To seize this historic opportunity, Alibaba officially establishes the Alibaba Token Hub (ATH)事业群, organized around three pillars: Create Token, Deliver Token, Apply Token.
ATH reports directly to me and includes:
– Tongyi Lab: Create leading multimodal models; maximize foundational capability.
– MaaS: Build efficient, open model platforms — empowering the entire AI ecosystem.
– Qwen Division: Build the world’s best personal AI assistant.
– Wukong Division: Build a B2B AI-native workplace platform deeply embedded in enterprise workflows.
– AI Innovation Division: Explore breakthrough applications — validating new markets and models.Effective immediately. Let’s build the future — together.
— Wu Yongming
March 16, 2026
Article originally published by Huxiu APP.