Table of Contents
In this age of artificial intelligence, could Kentucky’s elected officials be making some dumb decisions with taxpayer dollars when it comes to data-center subsidies? Rather than driving economic growth, could some of Frankfort’s economic development policies risk leaving Kentuckians worse off than before?
Thanks in part to new tax breaks created last year, Kentucky has seen a slew of plans and proposals for giant data centers – mammoth buildings full of computers that run our cell phone apps, search engine queries, social media posts, credit card transactions and every other tool of the modern information age.
There’s a huge demand for these facilities across the country, especially to meet the needs of new artificial intelligence tools. But just because they’re important doesn’t mean it’s intelligent for Kentucky’s governments to be shoveling millions – or eventually billions – of dollars in subsidies toward them.
To be clear, government “economic development” subsidy programs are almost always a terrible idea. A vast amount of real-world evidence backed by decades of independent research from experts across the political spectrum makes it clear: governments don’t grow the economy when they hand out taxpayer-funded subsidies to a select few companies or projects.
Companies make decisions about where to move and what to build based on a host of fundamental business factors. Government subsidies have little power to turn a bad site into a good one, create a workforce out of thin air or make it worth a company’s while to suffer through bad environmental rules, labor laws or other stifling regulations.
Within this world of bad deals, data centers stand out for being especially abysmal tools for an economic-development strategy. In many ways, they’re what you might get if you sit down with a clean sheet of paper to design the worst-possible project to subsidize in the name of economic progress.
To begin with, they’re unimpressive at job creation since they employ few people onsite. The high-value work enabled by data centers is being done by programmers in places like Silicon Valley, Shanghai and Mumbai, not in onsite offices.
Data centers offer limited benefits to surrounding communities. They’re self-contained, high-security facilities that don’t have a lot of visiting customers or vendors who might patronize local restaurants, gas stations, convenience stores, dry cleaners or other businesses that benefit from such spillover traffic.
They also do little business with local vendors. Almost all their expensive computers and networking equipment is imported from places like Taiwan, China and Mexico.
Finally, and perhaps most importantly, data centers have an almost-insatiable demand for energy, to the point where their growth is the biggest issue currently facing America’s power grids. Last year, the U.S. Department of Energy estimated that data centers account for about 5% of U.S. electricity use, potentially doubling by 2028.
What could that mean for Kentucky’s communities?
One single proposed “hyperscale” data center project in Louisville could use as much as 400 megawatts of power at full capacity. That’s enough energy to power roughly a tenth of the homes in the state, and it raises the obvious question: If the data center is using that power, where does the electricity come from to keep the lights on in those 200,000 homes?
Will Kentucky’s utilities need to build more power plants paid for by their electrical customers through their utility bills? If that’s the case, taxpayers will pay twice: first to subsidize costly imported computers for data centers, then through higher electric bills to fund new power plants to run them – all to host businesses that create few jobs or economic benefits for local communities.
Perhaps Kentucky should tell data centers: You’re as welcome here as any other industry – as long as you pay your own way.
Bluegrass Institute Scholar John C. Mozena is president of the Center for Economic Accountability. This commentary was first published in the Courier-Journal on June 16, 2025.
