As the U.S. requires increasing amounts of electricity to power the ongoing AI race, it is getting harder to deliver that energy, The Wall Street Journal writes.
Investors and developers are putting financing decisions on hold as they try to determine how much more their projects will cost to complete, partially due to tariffs and other economic uncertainties.
About 28% of planned wind, solar and battery projects have been delayed or canceled, according to an Atlas Public Policy analysis of government energy data. Delays and cancellations had been growing but recently accelerated, according to the data, which provides a snapshot of existing and planned power generation across the country. Only about 10% of projects, or 12,000 megawatts, were delayed or canceled in 2022.
Efforts to build natural-gas-fired power plants face similar challenges—wait times for turbines can stretch beyond three years.
While many companies say AI could be transformational for business, searches on AI platforms can use sometimes 10 times the amount of energy as a single Google search. Trends suggest that by 2030, data centers for training the largest AI models will require more than five gigawatts of electricity—or roughly enough to power about 5,000 Walmart stores.
