While the country's overall power prices remain relatively low, industrial rates for large users including artificial intelligence (AI)-related data centers are higher than in countries such as China and the U.S.
According to data released by the U.S. Energy Information Administration (EIA) on Tuesday, the average industrial rate applied to data centers in the U.S. was 8.62 cents per kilowatt-hour as of the end of 2025, nearly 40 percent lower than South Korea’s 182 won per kWh, meaning that operating the same data center for artificial intelligence (AI) in the U.S. would be far more cost-effective.
Industrial electricity rates are applied to large, around-the-clock users such as factories and data centers, and are typically set below commercial and residential rates. In some U.S. states like Texas and North Dakota, industrial power prices fall below 7 cents per kWh.
The gap is even wider for renewable energy power purchase agreements (PPAs) used by AI data centers. Average U.S. solar PPA prices were 6.45 cents per kWh in the first quarter, and average wind PPA prices were 7.94 cents. In South Korea, the average PPA price is about 13 cents per kWh or around 200 won, higher than the country's average industrial rate.
Supposing a cluster of 1,000 Nvidia H100 chips, annual electricity costs in South Korea would total about 2.67 billion won, compared with about 2.02 billion won in the U.S. China remains cheaper in hub regions such as Gansu, Guizhou and Inner Mongolia, where electricity prices average about 5.2 cents per kWh, bringing annual costs down to roughly 1.21 billion won.
Virginia and Oregon in the U.S. have begun applying separate rates for large loads of 25 megawatts or more. Japan subsidizes up to 50 percent of investment costs for data center facilities using decarbonized electricity, while Singapore and Ireland require a share of green energy as a condition for grid connection permits.
But South Korea, which has declared its ambition to become an AI powerhouse, has no dedicated electricity rate for data centers.
Relatively high operating costs for data centers could push up prices for AI services in South Korea, as electricity accounts for a large share of overall costs and providers using domestic data centers would have little choice but to set higher per-token prices
"If electricity costs start to widen, domestic AI models will ultimately lose price competitiveness, which will work against them in the global AI market," said Yoon Sung-ro, a professor at Seoul National University.
Copyright ⓒ Aju Press All rights reserved.