AI data centers consume enormous amounts of power, but their predictable, constant demand could actually make electrical grids more efficient and help lower electricity costs for everyone, experts say. While communities across the U.S. are fighting proposed data center projects over concerns about rising utility bills, the real driver of electricity prices isn't the data centers themselves, but how utilities respond to new demand.

Why Are Communities Worried About Data Centers and Electricity Costs?

The anxiety is understandable. In Joliet, Illinois, hundreds of residents packed a planning commission meeting in March to oppose a proposed $20 billion AI data center on 795 acres of farmland. Many voiced concerns that the facility would drive up electricity rates for households in the area. Similar opposition has emerged across the country as tech companies race to build massive complexes filled with thousands of graphics processing units, or GPUs, to power artificial intelligence systems.

The scale of these facilities is genuinely staggering. A modern AI data center can cover hundreds of thousands of square feet, house tens or even hundreds of thousands of GPUs, and consume hundreds of megawatts of power. Collectively, data centers in the U.S. already consume 5% of the nation's electricity, and nearly 3,000 new ones were under construction or planned as of 2025.

How Do Data Centers Actually Affect Electricity Prices?

Here's where the story gets more nuanced. Electricity prices depend less on data centers themselves and more on how utilities expand capacity and manage supply in response to new demand. With the right policies and technologies in place, data centers can actually improve grid efficiency and help lower electricity costs rather than raise them.

One critical factor is the difference between traditional cloud data centers and new AI facilities. Traditional data centers, which house CPUs (central processing units) for general computing, typically employ only 30 to 50 permanent staff members. AI data centers, by contrast, require significantly more infrastructure and maintenance because GPUs compute at much higher rates. At the Berry Hill megasite in Pittsylvania County, Virginia, officials expect an AI data center campus to create 2,050 permanent jobs over a 30-year period, with salaries averaging $85,000. The facility would span nearly 3,000 acres and represent a $73 billion investment.

Another crucial distinction involves how these facilities are powered. At the Pittsylvania megasite, Stack Infrastructure, the Colorado-based developer, will enter a service agreement with Appalachian Power and will pay for all costs associated with providing power to the campus. This means there will be no impact on neighboring ratepayers served by Danville Utilities, which operates on a separate grid. The megasite itself has been part of the Appalachian Power service territory since the mid-2010s, isolating it from the residential utility system.

What Technologies Help Data Centers Use Power More Efficiently?

Modern AI data centers are adopting innovations that reduce their environmental footprint and operational costs. One significant advancement is the shift from water-intensive cooling systems to closed-loop systems that recirculate water and eliminate evaporation. The majority of new facilities being implemented today use closed-loop systems, which are typically filled with about 10,000 gallons per building as a one-time fill and require recharging only every 8 to 10 years.

Beyond cooling, the future of data center power infrastructure is diversifying. Pete Sacco, founder and CEO of PTS Data Center Solutions, argues that the era of massive, centralized data centers is ending. Instead, clusters of small data centers requiring only 5 to 20 megawatts of power will become the norm, interconnected to create distributed computing systems. This shift is driven partly by the transition from AI training to AI inference, or answering user prompts, which requires data centers to be physically located near where queries originate to minimize response delays.

Sacco's company, Gray Wolf Data Centers, is building its first facility in Connecticut based on this new model. The facility will convert carbon-based waste, including tires, food, and medical waste, into energy. This approach allows Sacco to produce electricity at under 10 cents per kilowatt-hour, compared to Connecticut's typical residential rate of 30 cents per kilowatt-hour. By generating power locally and selling excess capacity to manufacturers at 20 cents per kilowatt-hour, the model undercuts standard utility rates while remaining profitable.

How to Evaluate Data Center Projects in Your Community

• Check the Power Grid Arrangement: Determine whether the proposed data center will be served by the same utility that serves residential customers. If it's on a separate grid or has its own power agreement, it won't directly impact household electricity rates.
• Review the Service Agreement: Confirm whether the developer will pay all costs associated with power delivery. Stack Infrastructure's agreement to cover all power-related expenses at the Pittsylvania site means no costs are passed to existing ratepayers.
• Assess Water Usage Plans: Ask whether the facility will use closed-loop cooling systems that recirculate water, or older evaporative systems. Closed-loop systems require minimal water replenishment and have far less environmental impact.
• Evaluate Local Job Creation: Distinguish between temporary construction jobs and permanent positions. AI data centers typically create 150 full-time positions per building due to the intensive maintenance and testing required by GPU infrastructure.
• Understand the Broader Grid Impact: Consider whether the data center's steady, predictable power demand could actually stabilize the grid and improve efficiency for all users, potentially lowering costs over time.

The Pittsylvania project illustrates how transparent communication can address community concerns. Matt Rowe, economic development director for Pittsylvania County, emphasized that the region is working on a performance agreement designed to be "the strongest, most lucrative, most community beneficial local performance agreement that has ever been done for an economic development project in the country". The project involves no discretionary monetary incentives beyond a state-mandated $1,000 per job created grant, and officials are exploring a separate tax on data center equipment to generate additional revenue for schools and services.

The broader picture suggests that AI data centers, far from inevitably raising electricity costs, could become tools for grid modernization. Their constant, predictable power demand allows utilities to optimize generation and distribution in ways that benefit all consumers. The key is ensuring that projects are built with transparent agreements, efficient technologies, and community input, rather than allowing fear and misinformation to drive blanket opposition.