Considering the impact of AI on electricity bills, data centers in the Northeast could be part of the problem.
Across the country, from Ohio to New Jersey, electricity prices are spiking [2, 3]. It’s a tangible financial pain, costing $122 more per month for the average household’s total utility costs compared to 2020 [1], adding serious strain to household budgets.
The AI revolution, from ChatGPT to generative video, runs on an astonishing amount of electricity [2, 6]. These data centers are “energy-sucking operations” clustered in “hot spots” like Virginia’s “Data Center Alley,” which shares the same PJM power grid that serves 13 states, including New Jersey, Pennsylvania, and Ohio [3,4].
This sudden, massive demand is overwhelming the grid. In the PJM market, the price for “capacity” (a fee paid to guarantee power is available on peak demand days) rose a staggering 833% in a recent auction [2]. An independent monitor’s report found that data centers were responsible for about three-quarters of that jump.
As a result, households are effectively subsidizing the AI boom. Utility companies fund massive new infrastructure projects by raising rates for all customers. Everyday households, whether they use AI or not, are paying for the power-hungry digital economy of the world’s most profitable companies [2, 4, 6].
A deeper analysis reveals a far more complex, and more expensive problem than simply big tech.
A recent academic study [8] analyzed state-level electricity prices from 2019 to 2024 and found that, counterintuitively, states with the greatest spikes in electricity demand (load growth) often saw lower prices overall. North Dakota, for example, saw a nearly 40% jump in demand but its inflation-adjusted prices fell.
This is because electricity isn’t a normal product. The largest part of electricity bills isn’t the cost of making the electricity (generation); it’s the cost of delivering it [3]. The source of the problem is a combination of the new data centers and the century-old grid they are plugging into.
Notably, the primary drivers of rising electricity costs are the “fixed costs” of the grid:
- Transmission & Distribution: The cost of laying the “poles and wires” is skyrocketing. In the last 20 years, while generation costs have fallen by 35%, transmission costs have nearly tripled, and distribution costs have more than doubled.
- Grid Hardening: It costs billions to make the aging infrastructure resilient to extreme weather, from hurricanes in Houston to wildfires in California. In California, wildfire-related costs alone accounted for about 40% of recent price hikes [3, 8].
In this context, adding a new, large, and predictable customer like a data center can lower rates for everyone. That new customer pays into the same fixed-cost system, spreading the cost of all those poles and wires over far more kilowatt-hours sold. But how can data centers be both the cause of price spikes (like in the PJM auction) and, historically, a source of price suppression? The answer is the speed of deployment and demand from the data centers.
The academic study which found prices could go down reflects a world of steady, predictable load growth. But the PJM auction spike reflects a new reality: the growth of AI data centers is so fast and so massive that it has broken that old model. This frantic pace creates acute congestion [5]. When the grid is congested, the system operator has to pay for the most expensive, last-resort power. That is the source of the cost spike.
In summary, the problem isn’t just the AI boom. It’s trying to run a 21st-century technology race on a 20th-century power grid, all while using an outdated regulatory playbook. Building massive new transmission lines is the right long-term solution, but it’s excruciatingly slow. That process takes at least 5 to 10 years for planning, permitting, and construction; however, the grid congestion is happening now, and electricity bills are rising now.
The real problem is knowing the bills are rising, without knowing by how much or exactly where. There is evidence of high-level auction prices [2] and broad, state-level trends [3, 8], but there’s a massive gap in the data. Without measuring the direct, household-level economic harm from this AI “arms race,” policymakers are essentially “flying blind,” trying to balance the economic promise of data centers with the risk to consumers.
This New Jersey State Policy Lab-funded research project is designed to fill that gap, shifting from top-down modeling to a bottom-up, empirical analysis of the actual impact on families. This study will start with a pilot study in Virginia and New Jersey, targeting zip codes next to these massive new buildouts. Using a data-donation model, this project will compensate 200 households for sharing their electricity bills. The goal is to secure zip-code level data (rather than relying on city-wide averages) and pair that quantitative analysis with qualitative research to understand the implications of this financial pressure on individuals.
There are concrete and practical technical solutions that can be deployed. It is critical to ask what is being paid for. The “brute-force” solution of only building new lines is too slow and expensive to solve the immediate problem [5]. A smarter, faster path is to use technology that moves beyond conservative “static ratings” by implementing dynamic line ratings based on real-time conditions and better power-rerouting optimizations to send electricity around bottlenecks. This approach gets far more out of the current grid.
Studies show these solutions are proven, can be deployed in just 1-2 years (not 5-10), and have exceptionally high benefit-to-cost ratios. They are not a permanent replacement for new transmission, but they are the critical bridge necessary to build the next generation grid intelligently [5].
References:
[1] J.D. Power. (2025). Average Household Utility Costs Rise 41% in Last Five Years.
[2] Whoriskey, P. (2025). Electricity rates in Ohio and elsewhere rise due to AI and cloud computing. The Washington Post.
[3] Osaka, S. (2025). The real reason electricity prices are rising, and it’s not data centers. The Washington Post.
[4] Saul, J., et al. (2025). How AI Data Centers Are Sending Your Power Bill Soaring. Bloomberg.
[5] The Brattle Group. (2025). Incorporating GETs and HPCs into Transmission Planning Under FERC Order 1920.
[6] Luccioni, S., & Jernite, Y. (2025). How Your Utility Bills Are Subsidizing Power-Hungry AI. TechPolicy.Press.
[7] Beckler, H., et al. (2025). How Business Insider investigated the true cost of data centers. Business Insider.
[8] Wiser, R., et al. (2025). Factors influencing recent trends in retail electricity prices in the United States. The Electricity Journal.