There was no issue in my recent campaign more controversial than data centers — which led to Public Service Commission action to make sure ratepayers get lower bills instead of higher ones.
These large warehouselike buildings need massive quantities of water, backup generators and electricity to power their operations. The endless rows of computers running inside their chilly corridors provide streaming for our movies, favorite apps and even artificial intelligence chatbots. But no matter the cost, it cannot fall on the backs of our families.
Let’s start with the money. While there are plenty of counties who want nothing to do with data centers, many do. The data centers bring immense capital and contributions to the tax base, offer premiums for the right property, support community infrastructure upgrades and often give financial support to community initiatives.
Douglas County, Fayetteville, Porterdale and many others have welcomed these power-hungry companies. With the tax revenue, roads get paved, libraries built, ball fields improved and so much more.
Some jurisdictions have even provided a tax abatement to lure data centers to their area. But they need power.
In a perfect world, Big Tech would pay for nuclear plants
Credit: handout
Credit: handout
That is where public utility commissions, power companies and rural energy cooperatives come in.
In 2024, our Public Service Commission took emergency action to acquire 3,300 megawatts of additional capacity, making sure the cost was borne by those asking for that power. The benefit to residential customers — almost $3 less per month.
On Dec. 19, we approved an additional 9,885 megawatts for this growing category of customers. Georgia Power agreed to financially backstop the expected new power load this time, preventing the possibility of ratepayers having to pay if the AI bubble bursts.
This backstop also created $8.50 per month of downward pressure on the average household using 1,000 kilowatts of electricity.
Not building the resources would be a missed opportunity to lower all rates. How? Because the data centers are paying all incremental costs for this power.
But this massive increase will require the construction of five new power plants.
Virtually all these plants are powered with natural gas — a just-in-time resource. Think Texas, North Dakota, New Mexico, Oklahoma and Pennsylvania — all with massive shale deposits sending gas our way.
Using gas for power plants has an opportunity cost, though, because the pipeline does not have unlimited capacity. Gas used to generate power can’t be used to build a poultry processing plant or car manufacturing facility. Expansion of those pipelines are underway to allow for Georgia’s needs and continued economic development for now. But there is a better way.
In a perfect world, every one of these data centers should be run by nuclear energy — paid for by Big Tech. Amazon, Microsoft, Google and their pals need to build nuclear on or adjacent to their property, working with local utilities and federal authorities to ensure they carry their own load, metaphorically and literally.
Data centers should pay the bulk of the cost for power
The technology is ready. Small modular reactors, or SMRs, as they are called, can provide up to 300 megawatts, and the safety protocols required are far less than the traditional massive reactors.
Data centers and their artificial intelligence experiment — not ratepayers — need to take the financial risk for building out America’s nuclear future.
Data centers are using the bulk of the power, so they should pay the bulk of the cost — and their investment in this technology will pay the learning curve, allowing all our states to advance nuclear at a much cheaper price in the future.
The benefits will mean that a carbon-free resource is providing this massive electricity requirement, and our natural gas resources are preserved for those who really need them. With the Federal Energy Regulatory Commission’s recent vote to allow massive energy users to plug directly into power plants, we are one step closer for data centers to make these deals and get huge quantities of carbon-free power.
Unconscious bias can seep into AI systems
But there is more at stake than just money. Anyone who has used artificial intelligence knows there may be a bias there with results produced. The developers and data annotators who create AI systems make choices about what data to use and how to label it.
Their own unconscious biases can seep into the system during this process. Or worse. The powerful algorithms themselves might be designed in a way that, although seemingly neutral, can produce discriminatory outcomes when processing certain data.
According to Pew Research, although some students use AI to help with schoolwork, others have developed relationships with digital companions or counselors that have at times encouraged sexual immorality or suicide, leading to a handful of deaths.
Evangelical Christians like myself are quite concerned there may be a bias against conservative or religious worldviews among AI creators that can lead to a narrow range of perspectives, meaning biases may go unrecognized or unaddressed during development and testing.
Not all the glitters is gold, as the expression goes. As a regulator, my job was to provide gas and electricity to whoever wanted it and would pay for it. Now, with my private citizen hat on, I have other concerns.
Data centers promise to transform how people live and work. On the other side are Georgia residents who worry the construction spree will have dire consequences for the environment, power prices and surrounding communities — and their children.
Now, Congress needs to act and quickly allow the learning curve for SMRs to be paid by hyperscalers. And leaders need to evaluate all of these matters and make good decisions for their communities.
Tim Echols finished a 15-year stint on the Georgia Public Service Commission on Dec. 31. He is now an adviser for United Consulting in Norcross.
About the Author
Keep Reading
The Latest
Featured
