Energy of Computing: Alibaba Builds 10T Parameter AI Model and 20 GW Data Centers

Chinese Big Tech has moved to a scale inaccessible to European competitors. On September 22, 2026, at the Apsara conference, Alibaba Corporation announced a plan to create a Qwen neural network with 5 to 10 trillion parameters. Synchronously, a new AI accelerator, the `Zhenwu V900`, was unveiled.

The main macroeconomic shock lies in the company’s infrastructural goal: bringing the capacity of Alibaba Cloud's global data centers to over 20 GW by 2032. This is the equivalent output of a dozen nuclear power plants. Alibaba is closing its architectural loop: from its own silicon to frontier models and megawatt capacities. Western sanctions have definitively lost their meaning. China has not only achieved independence in building AI factories but is also preparing to reshape the global B2B cloud market, providing an unprecedented scale of computing on custom hardware.

Source: Alibaba / Reuters / The Business Times