In today's era dominated by hyperscalers, a growing apprehension about data centers is closely tied to concerns over escalating utility costs. A YouGov survey conducted last year with a sample of 1,000 Americans revealed that over 66% anticipated an increase in electricity costs if a data center were established nearby. Furthermore, Goldman Sachs forecasted earlier this year that the expansion of AI infrastructure could lead to a 6% rise in electricity prices from 2026 to 2027, with an additional 3% increase expected by 2028.
However, a recent working paper from the Electric Power Research Institute has introduced a more nuanced perspective on the connection between the AI surge and its impact on electric bills for consumers. The study indicates that through at least 2024, the operational activity of data centers may not only allay consumer fears but could actively drive down retail electricity prices. Analyzing data from the Federal Energy Regulatory Commission (FERC) alongside retail revenue from the U.S. Energy Information Administration from 2015 to 2024, the findings demonstrated a causal link: for each doubling of data center capacity, average retail electricity prices dropped by 3.5%. On a state level, this reduction reached about 6%.
This phenomenon can largely be explained through the concept of economies of scale.
“Electricity markets operate differently compared to many other industries,” remarked Asa Watten, coauthor of the study and researcher at EPRI, speaking to Fortune.
In contrast to commodities like soybeans or gasoline, where prices fluctuate based on production costs, electricity prices hinge on cost recovery, which depends on the volume of consumption. As fixed costs are spread over a larger consumer base and increased kilowatt-hour usage, the distribution of these fixed expenses leads to lower prices. Additionally, heightened data center activity encourages more energy-efficient generators to activate due to increasing loads.
Yet, there is a caveat to this trend: its sustainability is uncertain. A reversal could herald larger issues for the future of AI. Recent reports from PJM, the nation’s largest power grid operator, anticipate a $6.3 billion rise in consumer electricity expenses over the next three years, primarily driven by growing data center energy demands. The accelerated construction of data centers, which is projected to reach $7 trillion in investments by 2030, is already linked to rising power costs. For instance, Virginia, home to the highest number of data centers, has seen residential electricity prices surge by over 13% in the past year, based on data from the U.S. Energy Information Administration.
The future dynamics between data centers and electricity pricing hinge on several factors.
Watten notes that the critical factor will be whether the rapid expansion of AI fulfills expectations.
“If the grid enhances capacity in anticipation of substantial demand from data centers that fails to materialize, it could significantly alter the outlook, potentially leading to higher prices in the future,” he warned.
Data centers are likely to encounter substantial fixed overheads, and if demand from clients does not match projections, “your denominator will be smaller than anticipated,” Watten explained. “This means those fixed costs would be divided among fewer consumers, contrary to the desired outcome, potentially driving prices up.”
As discussions about an anticipated AI bubble circulate, skepticism among investors is starting to surface. Notably, shares of Tesla and Alphabet fell following announcements regarding increased capital spending related to AI.
Mark Cuban, a billionaire investor, also weighed in on the matter during an episode of the All-In podcast, suggesting that many data centers could ultimately be repurposed for far less high-tech uses, such as pickleball courts. He pointed out that while hyperscalers may rightly predict continued AI adoption, the technology is expected to become less expensive to use due to improved energy efficiency, indicating that the extensive capacity being built might not be necessary.
Nonetheless, there is a glimmer of hope, according to Watten. While he refrains from making predictions about the future of AI and its implications for data center development, he remains optimistic about ongoing advancements in energy efficiency. The growth of electrification, spurred by factors like electric vehicles and electric heat pumps, coupled with data center expansion, could potentially lead to reduced household energy costs independent of an AI surge.
“Efficiencies and overall cost reductions could benefit not just individuals but their communities as well,” Watten stated. “A rise in electric vehicle usage, done effectively, might ensure that prices decrease or at least stabilize.”
