China's Green Power Push for AI Data Centers Hits Grid Limits

China's renewable energy ambitions meet AI infrastructure reality

China's effort to power its booming artificial intelligence data center sector with renewable electricity is encountering significant operational and economic obstacles, according to industry experts speaking at a Beijing conference last week.

The country has set an aggressive target for renewables to supply 80% of data center power consumption by 2030, up from just 11% in 2023. This initiative forms part of a broader strategic priority outlined in China's 2026 government work report to strengthen integration between computing infrastructure and power supply networks.

Data center electricity demand is expected to surge by 300 billion to 500 billion kilowatt-hours between 2026 and 2030, representing 18% of China's total electricity demand growth during that period, according to Pei Shanpeng, a director at State Power Investment Corp. The lower end of that projection equals the United Kingdom's entire annual power consumption.

The forecasting problem

Despite this explosive growth, AI-focused data centers present unique challenges for renewable energy providers compared to traditional industrial customers like aluminum smelters. The core issue is unpredictable peak demand patterns.

"At least for now, they do not appear to be very flexible in managing power demand," Pei said, as first reported by Reuters. "From what we understand, they cannot really adjust power consumption load much. GPUs are very expensive, so once they are purchased, operators want to use them as quickly and as intensively as possible."

This inflexibility stems from the economics of AI infrastructure. Companies investing heavily in graphics processing units need to maximize utilization to justify costs, making it difficult to modulate power consumption in response to grid conditions or renewable availability.

Pei noted that the green power initiative is primarily aimed at emissions reduction rather than cost savings for data center operators.

Grid operator resistance

A second major hurdle involves grid operators themselves. Direct renewable power connections to data centers could reduce electricity sales through traditional distribution networks, making it harder for utilities to recover substantial investments in transmission and distribution infrastructure.

The rapid data center buildout is already straining power systems in certain regions, increasing both average and peak grid loads. Operators must balance rising demand against reliability risks, creating tension around new direct-connection models.

"If 15% of the power consumption loads can be adjusted, it will significantly reduce capacity expansion pressure on the grid over the next three to five years," said Wang Zelin, deputy director at State Grid Jibei Electric Power Research Institute.

Why it matters

China's experience reveals a fundamental tension in AI infrastructure development: the operational demands of compute-intensive workloads conflict with the flexibility required to integrate intermittent renewable energy sources. As countries worldwide race to build AI capacity while meeting climate commitments, China's struggles with demand forecasting and grid economics offer early warnings about infrastructure bottlenecks that may constrain both AI deployment and decarbonization goals. The outcome will influence how other nations approach power planning for their own AI buildouts.

These details were first reported by Reuters correspondent Eduardo Baptista.