Are AI Data Centers Really Driving Up Electricity Prices?
The rapid expansion of artificial intelligence has intensified concerns that energy-hungry data centers will raise electricity bills for households. Michael Underhill, chief investment officer at Capital Innovations, argues that the evidence points elsewhere. Research cited by Underhill suggests large, consistent electricity users can lower average rates by spreading fixed grid costs across a broader customer base.
Underhill discusses the drivers of electricity prices, the risks created by regulatory bottlenecks and the investment needed to modernize the U.S. power system.
“The debate should not be framed as AI versus ratepayers. The real question is how we design regulatory frameworks so that growth pays for growth. History suggests that when large new loads are integrated properly, they can strengthen utility economics, support investment in critical infrastructure, and contribute to lower average costs for consumers over time,” Underhill said.
CM: What are mainstream narratives around AI data centers and power prices getting most wrong right now?
MU: The most common misconception is that AI-driven data-center growth automatically translates into higher electricity bills for residential consumers. That narrative sounds intuitive, but electric utility economics are more complicated than a simple supply-and-demand story.
Electric systems have large fixed costs associated with generation, transmission, and distribution infrastructure. When new, long-duration demand enters the system, those fixed costs can often be spread across a larger base of electricity consumption, improving asset utilization and lowering average costs. Recent research suggests that, historically, load growth has frequently been associated with lower average electricity costs rather than higher ones.
Another misconception is that data centers are the sole driver of rising power prices. In reality, many of the largest contributors to recent electricity inflation have been transmission upgrades, wildfire mitigation, storm-hardening investments, natural gas price volatility, environmental mandates, and capacity-market design issues.
From an infrastructure-investment perspective, AI is not creating a power crisis. It is accelerating the need for investment in the systems that support economic growth.
CM: What evidence most strongly supports the conclusion that data-center growth has historically reduced, rather than increased, residential rates?
MU: The strongest evidence comes from the recent Electric Power Research Institute (EPRI) study titled“Have Data Centers Raised Your Electric Bill? Causal Evidence from the United States.” Using an econometric methodology designed to isolate causation rather than simple correlation, the study found that data-center growth modestly reduced average retail electricity prices between 2015 and 2024.
The study’s conclusion is economically intuitive. Utilities recover substantial fixed costs through customer bills. When significant new demand enters the system, those fixed costs can be shared across more kilowatt-hours sold. Larger system utilization can improve economies of scale in generation, transmission, and distribution.
Importantly, the research found that states experiencing higher load growth often saw lower relative price increases than states with stagnant demand. The findings suggest that data centers have historically acted more like an incremental contributor to system efficiency than a source of widespread consumer-rate inflation.
CM: Under what circumstances would a new data center raise electricity costs for existing customers?
MU: A data center can raise customer costs when the regulatory framework allows infrastructure expenses to be socialized across the rate base rather than directly assigned to the new customer causing the load growth.
For example, if a utility builds new transmission lines, substations, generation capacity, or grid upgrades specifically to serve a large hyperscale facility and those costs are recovered from all ratepayers, consumers may experience higher electricity bills. This is particularly true if projects are delayed, cancelled, or significantly downsized after infrastructure investments have already been made.
The risk is not the presence of data centers themselves. The risk is poor cost allocation.
That is why many regulators and utilities are increasingly adopting large-load tariffs, minimum-demand commitments, collateral requirements, and “beneficiary pays” principles that require hyperscalers to bear the incremental costs they create.
CM: Which policy choices are contributing most heavily to rising power bills in the Northeast and on the West Coast?
MU: In the Northeast, natural-gas infrastructure constraints, transmission congestion, capacity-market design issues, and delays in adding new generation have all contributed to higher electricity costs. PJM’s recent capacity-price volatility has also been a significant factor.
On the West Coast, particularly California, wildfire mitigation expenditures, grid-hardening programs, reliability investments, and aggressive infrastructure replacement programs have become major drivers of rate increases. Several studies have concluded that wildfire-related spending has been among the largest contributors to elevated retail power costs in the region.
The broader lesson is that consumer electricity prices are typically driven more by policy, regulatory design, capital-investment requirements, and fuel costs than by data-center demand alone.
CM: From an investor’s perspective, how does this emerging data on load growth and rate impacts change your view on digital infrastructure versus traditional utility investments?
MU: The data reinforces our conviction that the most attractive opportunity is not choosing between digital infrastructure and utilities but investing across the ecosystem that enables both.
AI demand is creating a durable multi-decade need for power generation, transmission, substations, grid modernization, energy storage, and digital infrastructure. The beneficiaries extend far beyond the data-center operators themselves.
For utilities, sustained load growth is particularly important because the industry spent much of the last fifteen years operating in a low-demand environment. A return to meaningful electricity demand growth improves opportunities for regulated asset-base expansion and infrastructure investment.
For digital infrastructure, the key investment question is no longer whether demand exists. Demand is evident. The differentiator will be access to reliable power, interconnection rights, favorable regulatory frameworks, and contracted cash flows. The winners will be those assets that can secure and deliver power efficiently rather than simply those located in established data-center markets.
CM: Would data-center moratoriums protect consumers, or could they ultimately raise average electricity costs?
MU: In our view, broad moratoriums are a blunt instrument that may create unintended consequences.
If historical evidence indicates that load growth can lower average costs through greater system utilization and improved economies of scale, then prohibiting new demand could actually eliminate those benefits. In addition, moratoriums may discourage investment in generation and grid infrastructure that would otherwise improve reliability and expand system capacity.
The better policy approach is not to stop development but to ensure that new large-load customers pay the incremental costs associated with serving them. Properly structured tariffs, transmission charges, long-term purchase commitments, and cost-allocation rules can protect residential customers while still attracting economic investment and supporting AI leadership.
Ultimately, consumers are best protected not by restricting growth, but by aligning costs with beneficiaries. When that principle is followed, data centers can become a source of additional utility-system efficiency rather than a burden on households.
