A global policy question emerges: who should bear the financial burden of upgrading electricity grids to support AI data-center power demand at gigawatt scale?
As artificial intelligence consumes ever-growing amounts of electricity, the world is grappling with who should bear the costs of building power grids to transport this energy. Imposing the burden on ordinary citizens sparks fairness debates, while heavily taxing businesses risks driving AI investment abroad. In South Korea, Korea Electric Power Corporation (KEPCO) currently shoulders these costs, but concerns persist that the expense could eventually return to households through higher electricity bills. With the global "AI attraction race" intensifying, calls are growing for urgent societal consensus on how to fairly distribute these costs.
On September 16, the U.S. House of Representatives passed the "Electric Bill Payer Protection Act" with 417 votes in favor and 3 against. The bill prevents costs from new power plants and grids built for AI data centers from being passed to household electricity bills, instead requiring state regulators to ensure large data centers cover the expenses of additional infrastructure their operations require. President Donald Trump discussed the bill's progress with Senate Republican Leader John Thune on September 18. Australia's New South Wales is similarly reforming its system to prevent grid costs from shifting to general households and businesses.
Power grids, like roads, are traditionally shared public infrastructure. Expansion costs were typically distributed across household, commercial, and industrial electricity bills rather than being charged to specific entities. However, the AI era has changed this dynamic. A single large data center or semiconductor cluster now consumes as much electricity as a small city, necessitating multibillion-dollar investments in transmission lines and substations—effectively for one entity.
Studies show residents are already paying higher bills due to AI. Ireland, a European data center hub, exemplifies this trend. Research by Spain's Universitat Autònoma de Barcelona (UAB) estimated households paid a cumulative average of 360 euros (approximately 580,000 Korean won) more between 2015 and 2023 due to surging data center demand. After 2025, this burden could rise to 644 euros (approximately 1.04 million Korean won) per household. In the U.S., 4.4 billion dollars of grid investment costs in seven states—including Illinois, Maryland, and Virginia—were allocated to general consumers in 2024.
Shifting all grid costs to businesses is equally problematic. Data centers are critical AI infrastructure that drive massive investments and jobs; excessive costs could deter companies or push them to other countries. With nations fiercely competing to attract AI investment, imposing higher costs risks losing ground in the race. The common challenge is finding a balance: covering costs tied to data centers without burdening citizens or deterring corporate investment.
South Korea faces the same dilemma. KEPCO, responsible for grid investments, carried 210.7 trillion Korean won in debt as of mid-2026, paying 11.5 billion Korean won daily in interest. If KEPCO funds grid construction for AI data centers or semiconductor clusters, costs could ultimately be passed to citizens through rate hikes or mounting debt. Conversely, overburdening businesses risks driving high-tech investments overseas. KEPCO proposed that Samsung Electronics and SK Hynix prepay portions of future electricity bills to fund grid investments, but the companies rejected the proposal. This too reflects the core challenge: how to fairly distribute costs.
The U.S.-based nonprofit Regulatory Assistance Project (RAP) recently emphasized in a report: "Data centers drive economic growth and tax revenue but also raise electricity costs and grid expenses. Governments, utilities, and communities must collaboratively establish principles for cost-sharing."