The argument over AI data centers is becoming too simple. One side treats them as the inevitable engine room of the next economy. The other sees them as a power-hungry threat to household bills, water supplies and local control. Both instincts are understandable. Neither is enough.
The better answer is not to freeze data centers out or wave them through. It is to make the electric grid rules catch up with the speed of AI investment.
That is why the Federal Energy Regulatory Commission's June 18, 2026 action matters. FERC ordered the six regional grid operators under its jurisdiction to justify or reform the tariffs that govern how data centers, manufacturers and other large energy users connect to the transmission system. In plain English: the country's power markets are being told to explain whether their rules can handle a new class of very large customers without weakening reliability or shifting unfair costs onto everyone else.
That is the right question. AI infrastructure has real value. Faster computing can support medical research, logistics, cybersecurity, scientific modeling and business productivity. A country that wants to compete in advanced technology cannot pretend the servers will run on slogans.
But communities are also right to ask harder questions. If a data center needs new substations, transmission upgrades, backup generation or water infrastructure, who pays? If a utility raises rates after committing to serve a cluster of massive new loads, how much of that cost belongs to ordinary customers? If a developer promises jobs and tax revenue, how durable are those benefits compared with the long-term grid obligations?
The scale is no longer a niche planning issue. The U.S. Department of Energy has said data centers consumed about 4.4% of U.S. electricity in 2023 and could reach roughly 6.7% to 12% by 2028. The Energy Information Administration reported in June 2026 that servers alone represented an estimated 7% of commercial-sector electricity consumption in 2025, with growth expected in every long-term case it modeled. The International Energy Agency's Energy and AI work points in the same direction globally: data-center demand is rising fast, and AI-focused facilities can grow faster still.
Those numbers do not prove that AI data centers are bad. They prove that the old permitting and rate-making posture is too casual for the moment. A few principles should become standard.
First, large-load customers should pay their fair share of the upgrades needed to serve them. That does not mean punishing growth. It means avoiding hidden subsidies that show up later in residential and small-business bills.
Second, grid operators should reward flexibility. Some data-center work can shift by minutes or hours. Some facilities can use on-site generation, batteries or contracts that reduce strain during peak demand. A data center that can ease off when the grid is tight is more valuable than one that demands constant priority service and leaves everyone else to absorb the risk.
Third, the public should get clearer disclosure. Communities do not need every proprietary detail of a company's computing plans. They do need understandable estimates of power demand, infrastructure costs, water use where relevant, backup-power plans and local benefits before approvals become inevitable.
Finally, speed should be earned. FERC's phrase, speed-to-power, captures a real economic need. But speed without accountability would invite backlash, and backlash would slow the very infrastructure the AI sector says it needs.
A measured path is available. Let data centers connect. Make them transparent. Price their grid impact honestly. Demand flexibility where technology allows it. If AI is going to reshape the economy, it should also help prove that growth and public accountability can move on the same wire.
