PADUCAH — The Tennessee Valley Authority is considering creating a new power rate classification for AI data centers, with a formal proposal expected to reach its board as early as August. The initiative aims to adjust the utility’s rate structure to reflect the unique energy demands of data centers, which are currently grouped with traditional industrial users such as steel mills.

In February, the TVA sent a letter to the 153 local power providers it partners with to begin discussions about potential rate opportunities, including options tailored to AI data centers. The authority noted that data centers now account for approximately 18% of its industrial load—up from single digits five years ago—and projects that share will double by 2030. Industrial customers make up about 60% of the TVA’s total electricity load.

TVA spokesperson Scott Brooks said, "The TVA has a number of rate classes, one of which is residential and business, that's the one that is reflected on everybody's power bills. Then there's a large industrial and a large manufacturing, and so there's an opportunity there to take a look at that structure and get feedback from the local power companies on how we can offer them more flexibility and — most importantly — to ensure that consumers or other rate classes are not shouldering the burden of the growing market of data centers." He added that 99% of data center load is served by local power companies, with only a handful of facilities connected directly to the TVA system, and that local providers "are the ones who are in the best position to tell us what they're seeing — what their challenges are, what the opportunities are, and how the TVA can help."

Brooks said data centers factor into most scenarios in the TVA’s integrated resource plan, which looks 10 to 20 years ahead. "Part of our mission is economic development, so we — as a public power provider — are doing everything we can to be able to say yes anytime a business wants to locate in the valley." The TVA emphasized that the discussion centers on rate classification, not on raising or lowering rates specifically due to AI data centers.