Some Thoughts on AI, Data Centers, and Our Grid

Published by Terry Rubenthaler on Wednesday, July 2, 2025 in Electric

I was at the doctor’s recently, and he asked if he could use his Artificial Intelligence (AI) app to transcribe our visit. I was a little surprised, especially given that the office still requires faxed medical records, but I guess that’s a sign AI is definitely a part of our everyday lives.

Regardless of how you feel about AI, two things are undeniable: 1. It will have a dramatic impact, and 2. The data centers that make AI (aka machine learning) possible require a lot of electricity. I’ve read that ChatGPT uses 10x the amount of electricity as a standard web search. That’s because when you ask ChatGPT something, it performs several web searches at once and then creates new custom content based on the data from the searches. Put simply: It takes more work.

In fact, data centers in Virginia consumed 33,851,122 megawatt-hours (MWh) of electricity in 2023, according to the Electric Power Research Institute. That’s enough to power about 2.8 million homes.

As a CEO of an electric cooperative, you probably think this increased electric demand has me dancing in the streets. Not entirely.

That power has to come from somewhere, and our grid is already taxed due to growing demand and shrinking supply, even without the surge of AI. The potential risk to the reliability of the grid due to a lack of supply is very real, and it’s something MEC has been educating customers on for years. Plus, it takes about 10 years to build a new power plant, thanks to permitting and regulatory constraints. AI developers and engineers don’t want to wait that long, and since these data centers use electricity 24/7, they don’t want to rely on weather-dependent energy sources, like renewables, that only generate electricity in certain conditions.

That’s why on a recent trip to Washington, D.C., thousands of leaders from cooperatives across the nation, including myself and others representing MEC, met with legislative staff to educate them on concerns around electric reliability and permitting. It’s up to them to reduce the red tape that bogs down the building of critical new power sources. The good news is that it seems our representatives understand the urgency of this issue, and I remain cautiously optimistic that real change will happen. The grid operators are also doing their part by partnering with Big Data to streamline new generation sources. It’s not hard to predict the future if supply does not match the demand for power.

AI has exciting potential for our industry. There is the possibility for it to enable grid operators to better predict and prepare for peak demand, making the grid more efficient. There is also potential for AI to someday notify us when equipment is going to fail before it actually does and to analyze and direct our tree crews to areas that need attention instead of relying on our usual rotation. On the consumer level, machine learning is already helping people use less energy with smart home devices. For example, smart thermostats can sense when you are not home and automatically turn down the temperature. It’s exciting to think about all the possibilities.

Not even ChatGPT can tell us what the future holds. However, the entire industry will keep a close watch on how AI will influence our grid. For our part, MEC will continue meeting with legislators and industry partners to help them make informed decisions about how we get and consume energy. I will remain dedicated to ensuring you have safe, reliable power.

Just don’t be surprised if you visit your doctor soon and AI is in the waiting room with you.