According to the study "Rewiring China's grid for the AI era," reported by energynews, electricity consumption by Chinese data centers could quadruple by 2030 to roughly 774 terawatt-hours (TWh) due to surging AI workloads. The research estimates that data centers would account for about 6 percent of China’s total electricity use in 2030, rising to as much as 17 percent by 2060.
Network limits and renewables push investment outward
The analysis finds that eight designated national data-center hubs would retain a leading role, supplying more than 70 percent of China’s data-center capacity through 2060. However, network constraints and the continued expansion of renewable generation are likely to shift new development away from those hubs and toward sites outside them—particularly western regions of China that are rich in renewable resources.
Historically, data centers were sited close to computing demand: in large cities and technology centers to be near customers, skilled labor and low-latency networks. The AI expansion adds a new priority: access to abundant, reliable and lower-carbon electricity. For latency-tolerant tasks, this favors a model in which compute follows power rather than power following compute.
Compute follows power for latency-tolerant workloads
The authors note that for tasks less sensitive to latency—such as training large AI models—there is an advantage to locating computing where cheap, plentiful and low-emissions power is available. Applications that demand rapid response would continue to rely on eastern hubs near major demand centers.
Data centers as flexible grid participants and renewables absorbers
New energy architectures, battery storage and intelligent scheduling can turn data centers into flexible loads. Non-urgent processing can be shifted to periods of abundant renewable generation or lower electricity prices, while latency-critical services stay near high-demand zones. This shifts data centers from passive baseload consumers to active elements of China’s power system, including absorbing curtailed renewable output in oversupplied regions.
The study projects that demand for green power from data centers will rise steadily through 2060, while their carbon emissions could peak as early as 2035 because of the growth of renewables.
Expert remark
Wanting Zhao, an analyst for Wood Mackenzie’s Asia–Pacific power and renewables research, says that access to reliable, competitively priced and lower-emission electricity will increasingly determine where and how new capacity is built.
Why this matters
The scenario sketched by the study has implications for the geographic distribution of technology investment, local grid upgrades and China’s emissions trajectory. Policymakers and investors need to consider network constraints, the spatial distribution of renewables and latency requirements when planning new data-center projects.
(Image: illustration. Cover image source: Getty Images)
