Energy for AI: Data center power demand, grid constraints, and the carbon challenge in South Korea

A new opinion piece in The Korea Times examines the growing tension between AI data center expansion and energy supply. Global data center capacity is projected to hit 220 GW by 2030, six times 2020 levels, with power consumption expected to exceed 1,200 TWh by 2030. That is more than Japan or Russia consumed in 2025. The author, Kim Sung-woo, warns that up to half of announced data center builds could face delays due to power shortages, transformer bottlenecks, and long lead times for gas plants and small modular reactors. The article reviews the tradeoffs between grid power, gas, SMRs, fuel cells, and renewables. Solar and wind are intermittent and cannot guarantee 24/7 supply even with storage. Fuel cells currently rely on fossil fuels and high clean hydrogen costs. The author suggests a hybrid approach: meet primary demand with renewables and storage, cover auxiliary load with gas or grid power. The piece notes that 93% of planned U.S. generation capacity in 2026 is solar, wind, and storage, and that roughly 10 GW of fuel cells are expected at U.S. data centers over five years. For readers tracking grid decarbonization and AI energy demand, this article lays out the real tradeoffs. It does not pretend there is a silver bullet. The tension between AI companies' carbon targets and the need for reliable 24/7 power is the core problem. Worth reading for the concrete numbers on capacity, consumption, and the practical limits of each generation option.