China's Cheap Energy Infrastructure as Structural AI Advantage — 60 GW Data Center Expansion by 2030, Double US Electricity Generation Capacity

China's massive energy infrastructure is emerging as a structural competitive advantage in the global AI race, Al Jazeera reported in an investigative analysis on May 28. China generates more than twice the electricity of the United States annually and is set to double its data center capacity by 2030, with 28 GW of new data center projects coming online to add to the 32 GW already installed — reaching approximately 60 GW total by 2030. Data center rack count grew at approximately 30% annually between 2016 and 2023. China's energy cost advantage directly affects AI compute economics: running AI inference and training workloads on Chinese domestic data center infrastructure costs significantly less per token than equivalent US capacity, due to a combination of cheap coal and hydroelectric power in key provinces (Inner Mongolia, Guizhou, Sichuan) and state-subsidized electricity pricing for strategic AI infrastructure. The energy advantage compounds China's other compute cost advantages: DeepSeek V3's architecture-level training efficiency (2.788M H800 GPU-hours for ~$5.6M) and the permanent 75% price cut on V4 Pro API ($0.003625/M tokens, below Western rivals' estimated compute cost per Semafor) together with cheap energy mean Chinese AI infrastructure can serve global API demand at structurally lower costs than US equivalents. Al Jazeera notes that Western AI providers — constrained by expensive US grid electricity, data center power costs in the $0.08-0.15/kWh range versus China's $0.03-0.06/kWh — face a compute cost disadvantage that compounds over time as AI inference workloads scale. China's 15th Five-Year Plan (2026-2030) explicitly includes accelerating data center electrification and clean energy integration for AI infrastructure, with multiple provinces offering preferential electricity pricing for AI companies meeting government-aligned development criteria.