By Habib Ur Rehman
Publication Date: 2026-09-22 22:14:00
Alibaba used its September 22 developer conference to show how quickly China’s AI stack is filling the holes left by restricted access to U.S. chips. The company unveiled the Zhenwu V900 accelerator, which it says delivers roughly three times the performance of its predecessor and is scheduled for mass production in early 2027. Alibaba Group Holding Limited (NYSE:BABA) also discussed future Qwen models scaling to 5 trillion to 10 trillion parameters and its plans to expand global data-center capacity above 20 gigawatts by 2032. For NVIDIA Corporation (NASDAQ:NVDA), the issue is not today’s China revenue. It is whether a protected local ecosystem becomes good enough that customers stop waiting for Nvidia to return.
Alibaba is building the full stack at once
Alibaba Group Holding Limited can use a homegrown accelerator across its cloud, models and internal applications, which gives it a captive workload before it ever sells the chip broadly. That vertical integration is the bull case. If Zhenwu performance improves while Qwen demand rises, Alibaba can capture cloud revenue without paying an outside GPU vendor for every incremental unit of compute. The 20-gigawatt capacity target suggests the company is preparing for a much larger infrastructure footprint.
Pixabay/Public Domain
The obvious limitation is that the V900’s performance claims come from Alibaba and mass production is still expected in 2027. A chip can benchmark well and still struggle with yields, software…