A public-data decision-support framework for carbon-aware workload allocation in AI data-center hubs: evidence from China

AI data centers are becoming important electricity loads, yet public evidence on their operational flexibility remains limited. This article develops a public-data decision-support method for estimating how carbon-aware spatial workload allocation could affect emissions in AI datacenter hubs. The method combines public-source inputs, author-defined workload mobility assumptions, a network-flow allocation model, and claim-tier rules that separate main scenario evidence from exploratory stress tests.

China's computing hubs are used as an illustrative case because the setting combines rapid digital infrastructure growth, regional carbon-intensity differences and planning interest in coordinated computing resources. The current baseline is estimated at 20.30 Mt CO2. PUE-only improvement gives a 3.63% reduction, a verified 10% flexibility case gives 4.47%, and the main workload-aware allocation scenario gives 11.78% under the stated assumptions.

An upper-bound renewable and storage sensitivity reaches 29.23%, but it is not treated as a headline result. The method clarifies what public evidence can support, what it cannot support, and which additional operational data would most improve confidence.