Artificial intelligence (AI) companies may be booming, but they’re still struggling to get around a fundamental constraint: an insatiable demand for electricity. The federal government isn’t likely to help them. “I never want Americans to pay higher electricity bills because of data centers,” President Donald Trump has said, insisting that big-tech companies “pay their own way.”
This is the right approach. Ensuring that AI companies pay their full construction and operating costs will not only make electricity affordable again but will also encourage these companies to innovate. To get there, we need regulatory reforms at both the state and federal levels aimed at ensuring that price signals, not government intervention, determine the trajectory for data-center growth.
Finally, a reason to check your email.
Sign up for our free newsletter today.
These reforms should include rolling back any remaining federal-level Inflation Reduction Act policies and state-level Renewable Portfolio Standards (RPS), coupled with congressionally mandated permitting reform. These steps will lay the foundation for AI to flourish long into the future.
Last month, President Trump’s National Energy Dominance Council announced a consensus view on AI data centers: they must pay the full cost of their power needs, not pass these costs onto consumers. This view brought together a bipartisan group of federal- and state-level politicians, including all 13 governors (eight Democrats, five Republicans) representing the states within the PJM Interconnection regional electricity grid. PJM is the largest electricity grid operator in the U.S., serving more than 65 million people and powering more data centers than any other region. The PJM region includes Virginia, home to 35 percent of all hyperscale data centers in the world.
Affordability is central to this bipartisan consensus. Residential electricity bills have risen 10 percent since 2022, while electricity charges for the data centers driving the greater electricity demand have increased by just 4 percent. More worrisome still, data centers’ demand for electricity is expected to grow: the Department of Energy projects it will double or triple by 2028, and the Energy Information Administration recently forecasted the “strongest four-year growth in U.S. electricity demand since 2000, fueled by data centers.”
How do we meet our growing electricity needs? The answer is simple: get government out of the way.
The states and the federal government should insist that data centers secure their expected electricity needs under long-term procurements, without putting taxpayers on the hook for unneeded investment. No public underwriting of data-center costs.
At the same time, data centers must be free to source their power from the wide range of energy sources available in the U.S.—especially cheap, abundant natural gas. And that means deregulation.
First, state legislatures and Congress need to roll back regulations found in the Inflation Reduction Act and in the RPS, which mandate that utilities must procure a minimum share of electricity from renewable sources. The majority of states have implemented RPS. Repeal would allow lower-cost electricity for data centers and residential consumers.
The federal government has subsidized the RPS through tax credits, including ones created by the Biden administration’s Inflation Reduction Act. President Biden’s policies muzzled natural gas production and subsidized other forms of expensive energy (especially solar and wind), despite his own administration’s findings that natural gas was 3.3 times cheaper than electricity for residential power needs in 2023 and 2024. These regulations should be eliminated—or, at a minimum, exceptions should be carved out for data centers.
Congress must also enact meaningful permitting reform. At a recent United States Energy Association forum, a collection of associations representing all energy sources—including natural gas, nuclear, hydropower, solar, and wind—agreed: lengthy, antiquated permitting processes were preventing them from bringing critical energy infrastructure projects online.
Forcing AI companies to pay their fair share of electricity costs will also encourage innovation. To remain profitable, firms will need to develop new ideas and products that take the cost of energy into consideration.
Some big tech companies are already trying to do this, using direct-purchase agreements and onsite power generation to bypass the grid. Other forward-thinking companies have looked for ways to boost computing power and efficiency. For example, quantum computing could reduce the energy needs for some AI tasks by up to 90 percent. Other businesses may explore decentralized data centers as a way to reduce costs, improve reliability, and mitigate security risks.
At present, AI companies are like Icarus, flying too close to the sun with wax wings. Smart policy choices—in this case, forcing AI companies into a truly free market, without subsidies—will keep those wings from melting.
Photo by Tayfun Coskun/Anadolu via Getty Images