Alibaba has installed former cloud CTO Zhou Jingren to run its AI unit, signalling a pivot away from the open‑source Qwen project. Internal friction over strategy prompted the exit of Qwen’s technical lead Lin Junyang and researcher Hui Binyuan. The move aims to steer model development toward offerings that can be directly monetised rather than freely distributed.

Alibaba’s cloud business currently leans on GPU leasing, but executives want a larger slice of AI spend by bundling proprietary models with ecommerce tools. Chief executive Eddie Wu highlighted a nascent model-as-a-service (MaaS) line, while the newly created Alibaba Token Hub consolidates training, enterprise and consumer teams. Rival ByteDance’s token‑based cloud strategy and the surge in agentic AI workloads intensify the race for compute.

Investors have shifted from chasing benchmark scores to demanding revenue‑generating AI. Model licensing remains low‑margin, yet the growing demand for multi‑step agents promises higher token consumption. Analysts liken Alibaba’s ambition to a Chinese “Google”, betting that proprietary models and in‑house chips will anchor future cloud growth. The company now banks on AI services to lift its cloud earnings overall.