• Rising electricity demand from artificial intelligence data centres is forcing the US power grid to keep old ‘peaker’ fossil fuel plants operational, reversing scheduled retirements.
• Environmental and public health concerns are growing as these plants emit high pollution and are disproportionately located in disadvantaged communities.
What happened: AI electricity demand sustains old peaker plants
The rapid growth in electricity consumption by artificial intelligence (AI) data centres in the United States has led grid operators and utilities to delay or cancel the retirements of many older fossil fuel power plants known as “peakers.” These plants, designed to run only during periods of peak demand, had been slated for closure but are being pressed back into service because electricity demand is outstripping supply, particularly in the PJM Interconnection grid covering 13 states from the Midwest to the Mid-Atlantic.
In Chicago’s Pilsen neighbourhood, for example, the rarely used eight-unit Fisk plant, built in the 1960s and owned by NRG Energy, Inc., had been scheduled to retire but remains operational after power prices surged amid growing grid load. Matt Pistner, senior vice president of generation at NRG, said that the company withdrew the retirement notice because there was “an economic case to keep them around.”
A Reuters analysis shows that out of roughly 23 oil, gas and coal power plants scheduled for retirement starting in 2025 in the PJM region, utilities and grid operators have delayed closure of 13 facilities, with 11 of those being peaker units.
Also Read: APAC Summit forecasts massive surge in AI infrastructure
Also Read: Tech giants pledge over $40B to UK AI infrastructure
Why it’s important
Peaker power plants are typically less efficient and emit more sulfur dioxide and other pollutants per unit of electricity generated than modern base-load plants or renewable energy sources. Many of these facilities lack advanced pollution control technologies, meaning increased operation can worsen local air quality and health outcomes. These impacts are often felt most by low-income and minority communities where peaker plants are disproportionately located, a pattern documented in Environmental Protection Agency data and academic studies.
Alternatives such as improved transmission capacity to balance regional supply and demand, as well as battery storage systems that can supply power during peak periods, are suggested by energy experts. These technologies could reduce reliance on peakers and support broader clean energy integration. However, current grid infrastructure remains constrained, making these transitions expensive and complex.
The Federal Energy Regulatory Commission (FERC) has recently directed PJM Interconnection to adopt new rules governing connections for large electricity consumers such as AI data centres, indicating federal interest in addressing grid reliability amid surging demand. You can review FERC’s official guidance and regulatory role on its official site.
Nevertheless, the continued use of peaker plants raises broader questions about the trade-offs between supporting rapid data centre growth and achieving long-term climate and health goals. Critics argue that maintaining these plants could slow the transition to cleaner energy sources and increase electricity costs for consumers if grid upgrades and renewable capacity expansions lag behind demand growth.
