US Gas Plant Costs Surge as Data Center Demand Strains Supply

The hidden cost of the AI boom is starting to show up in the price tag for U.S. gas plants. New combined-cycle gas projects cost an average of $2,157 per kilowatt in 2025, up from less than $1,500 in 2023, while build times stretched 23% longer as utilities and developers chased power for data centers and other fast-growing loads.

Natural gas already dominated the U.S. grid before this surge began. It provided 43% of electricity generation in 2023, according to the U.S. Energy Information Administration, and gas-fired plants ranked second among new generating-capacity additions that year, behind solar. But the latest rush has exposed a system under strain, where nearly every part of the supply chain is tighter and more expensive than planners expected.

Turbine delays are now one of the biggest bottlenecks. Industry sources told S&P Global in May 2025 that wait times ranged from one to seven years depending on the model, and Bobby Noble of EPRI said some original equipment manufacturers were quoting five to seven years for orders as of February 2025. S&P Global also said costs had risen as much as 2.5 times in some markets, underscoring how quickly the economics of new gas generation have shifted.

The demand shock is being driven largely by data centers. S&P Global said in August 2025 that grid power demand from hyperscale, leased and crypto-mining data centers was set to roughly double between 2024 and 2028. Utilities have responded by moving toward more gas, not less. Jon Rea of RMI said utilities reviewing plans since the end of 2023 had added 52 gigawatts of gas capacity, compared with just 4 gigawatts of wind and solar. Gas utilities have also advanced supply deals for data-center campuses in Ohio, Pennsylvania and Texas.

That scramble has broad implications for consumers, grid planners and climate targets. A September 2025 analysis by GridLab, Energy Futures Group and Halcyon put recent combined-cycle costs near $2,000 per kilowatt and warned that public datasets may be underestimating current market prices. That matters because utilities recover much of these costs through ratepayers, and expensive gas assets can lock in higher bills for years. It also leaves planners trying to reconcile faster AI-driven load growth with slower, cleaner alternatives that are still constrained by interconnection queues, supply shortages and long development timelines.