Alibaba’s new chip puts cloud ambitions in focus

Alibaba Group Holding has unveiled the Zhenwu V900 AI accelerator chip, which it says delivers roughly 3x the performance of its predecessor. The company also plans to scale Qwen models to up to 10 trillion parameters and build more than 20 gigawatts of global data center capacity by 2032, linking chips, models and cloud services into a broader commercial platform.

The strategy signals an effort to keep more AI economics inside Alibaba by reducing reliance on outside suppliers while expanding infrastructure across Europe, the Middle East and Asia. Management is backing the buildout with RMB 380b of AI and cloud investment, while Alibaba is committing more than US$53b over three years to support expansion. Citi is assuming higher capital spending from fiscal 2027 to 2029 as AI workloads scale.

Analyst expectations already reflect stronger activity, with consensus revenue forecast to rise 12.5% per year over the next three years. Margins are modeled to improve from 7.0% today to 11.9% in three years, while earnings are expected to reach CN¥176.3b by 2029, up from CN¥73.3b today. The key test is whether cloud and AI revenue can grow quickly enough to offset heavy capex, quick commerce spending, softer China consumption and continuing regulatory and legal risks.