Alibaba (阿里) launches "Token Mobilization" with new Alibaba Token Hub

What changed

Alibaba (阿里) has created a new business group, Alibaba Token Hub (ATH), to centralize its AI strategy around "creating tokens, delivering tokens, and applying tokens." ATH will fold in foundational model R&D at Tongyi Lab (通义实验室), the MaaS line, the consumer AI assistant Qianwen (千问), the enterprise AI work platform Wukong (悟空), and a separate innovation unit. It has been reported that the ATH abbreviation is meant to carry a double meaning, with a wink to "All Time High" as well as the technical token focus.

Why it matters

The immediate aim is operational: resolve chronic tension over scarce compute and improve coordination between model teams, platforms and product groups. Senior exits and internal complaints about resource allocation — reportedly including friction around compute quotas that contributed to Lin Junyang's (林俊旸) departure — made the fragmentation visible. Who controls the C‑end AI entry point matters. By bringing Qianwen and Wukong under one roof led by Wu Yongming (吴泳铭), Alibaba is consolidating both control and incentives to reduce future cross‑business "tug‑of‑war."

Business and strategic implications

The move effectively shrinks the remit of Alibaba Cloud (阿里云) from an AI front‑end player to a pure infrastructure provider in the group: ATH will define model and token strategy while Alibaba Cloud supplies the compute, storage and platform services. That reallocation of responsibilities will reshape internal power and resource flows — who reports to whom, who gets prioritized compute, and who owns the primary consumer and enterprise AI entrances. For example, Qianwen’s transfer into ATH reduces the strategic heft of the team that previously managed it, and raises the likelihood that Wukong will receive similar resource priority on the B‑end.

Wider context

This reorganization comes as China doubles down on domestic AI ecosystems amid global tech competition. It has been reported that, after OpenClaw's surge, OpenRouter saw domestic model token usage edge past U.S. models in February 2026 — a signal of local commercial momentum. At the same time, export controls and geopolitics have tightened access to top‑end chips, making on‑premise and internal compute allocation a strategic constraint for Chinese cloud groups. ATH is Alibaba’s answer: a single, group‑level command for tokens, models and downstream applications, designed to cut internal friction and position the company as a more unified AI competitor.