China's Grid Operators Resist Renewables Mandate for AI Data Centers, Exposing Hard Infrastructure Limits

Chinese grid operators are pushing back against national plans to power the majority of AI data center demand with renewables by 2030, citing the sector's inflexible, high-intensity load profiles that complicate forecasting and risk management. Industry experts quoted by Reuters highlight that data centers cannot easily adjust consumption, as expensive GPUs drive operators to maximize utilization once deployed, straining transmission planning and raising reliability concerns amid rapid capacity expansion. This resistance emerges even as China advances flagship projects like the 24 MW Shanghai Lingang offshore wind-powered underwater data center, developed by HiCloud Technology and state-owned China Communications Construction, which leverages seawater cooling and direct renewable supply to cut energy and land use. An IEA analysis confirms China's data center electricity mix remains coal-dominated at nearly 70% as of 2025, with renewables at ~20%, while projecting solar and wind to add nearly 90 TWh by 2030 through grid mix improvements, co-location mandates, and western siting policies. Yet experts note data centers' poor fit for variable renewables due to unpredictable peaks, potentially forcing continued reliance on dispatchable sources. This constraint reveals deeper limits on AI scaling in the world's manufacturing hub, where grid inflexibility and demand opacity intersect with energy policy in ways rarely foregrounded in broader decarbonization debates—potentially slowing hyperscale deployments unless storage, demand response, or hybrid baseload solutions advance rapidly.