[the_ad id="3024875"]
Southeast utilities may be building too much gas infrastructure for ai data centers that may never come

Power companies across the Southeast are rushing to build expensive new fossil fuel infrastructure to meet projected energy demands from AI data centers—but a new study suggests they may be dramatically overestimating how much power these facilities will actually need.
According to research from Greenlink Analytics, a clean energy nonprofit, the aggressive growth projections that utilities like Georgia Power are using to justify massive expansions have only a 0.2% chance of actually materializing. While data center energy use could grow by 2.2 to 8.7 gigawatts across the region by 2031, utilities are planning for the extreme high end of these estimates. Georgia Power alone received approval for a staggering 10-gigawatt expansion last year, largely driven by anticipated data center demand.
The consequences of this overbuilding could be severe for both ratepayers and the climate. If projected demand fails to materialize, regular electricity customers—not the data centers that never arrived—will likely bear the costs of unnecessary infrastructure. Meanwhile, these sky-high demand projections are stalling the transition away from fossil fuels, with utilities falling back on natural gas plants with 45-year lifespans instead of pursuing more efficient solutions or renewable energy sources.
The study highlights rapid improvements in AI technology efficiency that could significantly reduce future energy needs, suggesting that much of the projected data center demand could be met through existing infrastructure improvements rather than costly new gas-fired plants. As the Georgia legislature considers bills to protect ratepayers and increase transparency around data center development, the research raises critical questions about whether the Southeast’s utilities are making the right bets for the future.
This article was written by the EnviroLink Editors as a summary of an article from: Grist News







