Manager’s Comments — Artificial Intelligence: The end of our world as we’ve known it!

Jon Cullimore is general manager of Coosa Valley Electric Cooperative.

I was born in 1968, which puts my formative years smack-dab at the beginning of the digital revolution. In the 1980s, technology advanced quickly and our imaginations ran wild.
I can still remember when I first saw movies like “War Games” and “The Terminator” and being absolutely blown away. “The Terminator” in particular tells the story of an artificial intelligence (AI) network called Skynet that declared war on humans and was trying to wipe them out. Many other movies since have postulated similar story lines about AI working to exterminate humans before we ruin our planet. It makes you wonder if Hollywood has a collective guilty conscience for its own overindulgences.
Hollywood nightmares aside, AI is now a reality, and the race is on to make the most of its potential. As your electric cooperative, however, what’s most important to us is not how it’s used. It’s AI’s hunger for energy.
You might think that a large power user is good for Coosa Valley Electric, but AI data centers need enormous amounts of electricity.
For some context, CVEC serves more than 18,000 consumers. On average, those consumers demand between 60 and 80 megawatts (MW) of electricity. On our coldest winter mornings, demand can climb to more than 110 MW. And, out of all those consumers, our single largest, one industry, demands less than 5 MW on its own.
For more context, consider Honda Manufacturing of Alabama in Lincoln. While we don’t serve that facility, we know its total demand tops out at less than 200 MW, which is a very big load.
Now let’s look at AI data centers. The average data center demands 1,000 MW! That’s 1 gigawatt, or five times the Honda plant, or nearly 10 times the peak demand of our entire distribution system. On top of that, it’s consistent, “always on” demand that never shuts off and never slows down. So, where do we get the energy to run something like that?
Our power provider, PowerSouth Energy Cooperative, can generate close to 2,400 MW in total and is currently building another 450 MW of additional capacity. That will be close to 2,850 MW of total generation, but none of it is intended to supply a data center. This is only to meet the growing needs of its current distribution partners, like CVEC. The story is similar across the country.
The energy consumption of data centers is unlike anything we’ve seen, and the race to meet the world’s appetite for AI’s capabilities is only heating up. Several $1 billion-plus projects are already under construction in Alabama with dozens of companies exploring more possibilities. PowerSouth doesn’t have the generating capacity to meet the needs of these facilities, and Alabama Power’s capabilities are very limited as well.
New generation is needed, but who’s going to pay for it? Facilities capable of supplying the capacity will cost more than $2 billion each. To complicate the risk even more, the technology driving AI is becoming more efficient even as it’s being developed. This means a data center that demands 1,000 MW initially, may demand half that in just a few years. By that time, there may or may not be additional loads needing the excess generating capacity. In the end, the likelihood of having very large, underutilized power plants is concerning and expensive.
With all these uncertainties, several things are becoming clear. AI has the potential to change our energy landscape forever as well as impact our personal and work lives in ways we can barely imagine. It may not be the end of our world, but its evolution is going to be the end of our world as we know it. I’m not saying it’s a bad thing. I’m only saying that our world is going to change.
We are looking forward to some warmer weather and pray that you have a pleasant March.