The electric bill has become another envelope many Americans hesitate to open. Families already paying more for groceries, rent, insurance, and medical care are now watching the cost of keeping the refrigerator cold and the air conditioner running climb higher.
In May 2026, the average U.S. residential electricity price reached 18.44 cents per kilowatt-hour, up 6.2% from a year earlier. In Virginia, the center of America’s data center industry, residential prices surged approximately 15.4% over the same period.
Now utilities are preparing for an even more expensive future. Artificial intelligence companies want enormous computing campuses that can consume as much electricity as small cities. Grid operators are planning new power plants, substations, and high-voltage transmission lines before anyone can know whether the promised AI demand will actually arrive.
That is the dangerous part of America’s data center boom. The country is not merely building server warehouses. We are making a multitrillion-dollar bet on electricity demand that may change before the infrastructure is finished.
If technology companies remain for decades and pay the full cost of serving them, customers could benefit. If projects are canceled, delayed, or made unnecessary by more efficient technology, families and small businesses could be left paying for the wreckage.
Americans are already struggling with electricity prices.
The latest national numbers leave little room for comfort. According to the U.S. Energy Information Administration, residential electricity prices increased from 17.37 cents per kilowatt-hour in May 2025 to 18.44 cents in May 2026. That 6.2% increase arrived while many households were already cutting spending elsewhere.
A $150 monthly bill rising by the same percentage becomes roughly $159. A family facing Virginia’s 15.4% increase would see that $150 bill climb to about $173, assuming electricity consumption remained unchanged. That additional money does not buy better food, a tank of gas, or medicine. It buys the exact same amount of electricity the household used before.
The burden becomes especially painful during summer. Parents cannot simply turn off the air conditioner when children are sleeping in dangerous heat. Older Americans cannot safely treat cooling as an optional expense. Renters often cannot replace inefficient appliances, improve insulation, or install solar panels.
When electricity rates rise, lower-income households face an ugly choice. They can reduce necessary power use, fall behind on the utility bill, or cut spending on other essentials. Meanwhile, the largest new users on the grid are technology companies pursuing one of the most expensive corporate expansions in history.
The EIA’s May 2026 electricity data show that the national residential rate increased 6.2% year over year. Virginia’s rate climbed from 15.26 cents to 17.61 cents per kilowatt-hour.
The cheaper data center electricity story looks backward.
Supporters of data center development have received unexpected help from a 2026 working paper by Asa Watten, John Bistline, and Geoffrey Blanford. The researchers examined electricity prices and data center expansion between 2015 and 2024. They estimated that every doubling of data center capacity caused average residential electricity prices to decline by approximately 3.5%.
At first, the finding sounds like proof that public fears are misplaced. It is not. The research describes what happened during a specific historical period. It does not guarantee what will happen during the far larger AI construction boom now unfolding.
From 2015 through 2024, the researchers identified approximately 21.3 gigawatts of new data center capacity across the continental United States. Much of that growth occurred while utilities still had room to use existing grid infrastructure more efficiently.
The economics were straightforward. Electricity systems contain enormous fixed costs. Power plants, transmission lines, substations, distribution networks, and control systems must be financed whether customers use them heavily or lightly.
When a data center is connected to a system with unused capacity, it purchases enormous quantities of electricity around the clock. Its payments helped spread fixed costs across more kilowatt-hours. That could reduce the average cost assigned to everyone else.
But that benefit depends on one critical condition: the data center must not require infrastructure spending that exceeds the value of its new electricity purchases. That condition is beginning to collapse. The researchers themselves warn that supply constraints and excessive demand forecasts could reverse the historical price benefit.
Their preferred estimate also carried a wide statistical range, and the price reduction appeared gradually rather than immediately. The paper is evidence that data centers once lowered average rates under favorable conditions. It is not a blank check for utilities to approve every multibillion-dollar expansion placed before them. The working paper explicitly warns that future supply constraints could reverse its finding.
The $7 trillion AI buildout is an enormous financial risk.
McKinsey estimates that data centers could require nearly $7 trillion in global capital spending by 2030. Approximately $5.2 trillion would support AI workloads, while another $1.5 trillion would support conventional computing. That figure is not a guaranteed investment. It is an estimate of what the industry may need if projected computing demand materializes.
The difference is crucial. Utilities do not build major infrastructure with imaginary money. They recover approved investments through rates, long-term contracts, and charges paid by customers. A data center may take two or three years to open. A major transmission line or power plant can require far longer. By the time the electricity infrastructure is completed, the technology that justified it may have changed dramatically.
More efficient chips could reduce the energy required for each AI task. Improved software could produce similar results with less computing power. Companies could move workloads to regions with cheaper electricity. Proposed data centers could fail to secure tenants. A change in financing conditions could kill projects that currently exist only in development presentations.
Yet utilities may have already ordered equipment, acquired land, and started construction. The public would then confront a brutal question: Who pays for infrastructure built for a corporate customer that never arrived? Unless regulators establish binding protections, the answer could be everyone still connected to the grid.
McKinsey describes the expansion as one of the largest infrastructure buildouts in modern history. The potential reward belongs largely to the companies selling AI services. The infrastructure risk could be distributed much more widely.
The grid is being forced to move at reckless speed.
The International Energy Agency projects that worldwide data center electricity consumption could reach approximately 945 terawatt-hours by 2030, more than double the 2024 level. U.S. data center consumption alone could increase by roughly 240 terawatt-hours, or 130%, from 2024 through 2030. The United States and China are expected to account for nearly 80% of global data center electricity demand growth.
These national figures hide the local danger. Data centers are not distributed evenly across the country. They cluster where developers can obtain land, fiber connections, tax incentives, and access to large amounts of electricity. This concentration can overwhelm a regional grid even when data centers remain a relatively small share of global power demand.
A utility may suddenly receive several requests for facilities requiring hundreds of megawatts each. Some developers may submit similar proposals in multiple locations while deciding where to build. If planners treat every request as a certain demand, the forecast can become dangerously inflated.
Electricity infrastructure cannot be built or canceled as quickly as a software product. Transformers, turbines, and transmission equipment already face long delivery times. Permitting can take years. Communities may resist new power lines or generating plants.
We are trying to operate a century-old utility system at the speed of the AI investment cycle. That mismatch is a recipe for higher costs. The IEA warns that future data center demand remains highly uncertain, particularly because technology can develop within a few years while major energy infrastructure requires much longer.
PJM is already showing Americans the bill
The clearest warning is coming from PJM Interconnection, the nation’s largest regional grid operator. PJM coordinates electricity across 13 states and the District of Columbia, serving approximately 67 million people. Its independent market monitor has concluded that data center demand is the primary reason for recent tight conditions and high prices in the region’s capacity market.
Capacity payments compensate power suppliers for remaining available when the grid needs them. These costs eventually become part of the wider electricity system, paid for by households and businesses. The market monitor estimated that existing and forecast data center demand added approximately $23.1 billion to capacity-market revenues across three auctions covering delivery years from 2025 through 2028.
For the 2026-to-2027 delivery year, data center demand was associated with an estimated $7.27 billion increase in capacity revenue. For 2027 to 2028, the estimated increase was nearly $6.5 billion. The situation became even more disturbing during the first five months of 2026. Total PJM wholesale power costs rose 62.7% compared with the same period in 2025.
The market monitor estimated that data center demand increased wholesale power prices by $11.26 per megawatt-hour, equivalent to a 24.4% increase. It attributed approximately $3.79 billion of the year-over-year wholesale cost increase to data center load.
Not every dollar appears immediately on a residential bill. State regulations, utility rate cases, and local contracts determine how costs reach customers. But those billions do not disappear. Someone pays them. The June 2026 PJM market monitor report provides a grim picture of a grid where demand is rising faster than dependable supply.
Virginia is becoming the warning Americans cannot ignore
Virginia has benefited economically from becoming the country’s largest data center market. It has collected tax revenue, attracted investment, and positioned Northern Virginia as a global center for digital infrastructure.
It has also become the place where the risks are hardest to dismiss. In May 2025, Virginia’s average residential electricity price was 15.26 cents per kilowatt-hour. By May 2026, it had reached 17.61 cents. That was a 15.4% increase in one year.
Data centers cannot be blamed for every cent. Fuel expenses, generation investments, base rates, and other grid costs also matter. However, Virginia demonstrates how quickly the promise of inexpensive electricity can sour once growth begins pushing against infrastructure limits.
The state’s earlier data center expansion may have used existing grid capacity efficiently. The next expansion requires major construction. That distinction changes everything. Filling unused space on the grid can spread costs. Building an entirely new grid for AI campuses can create costs faster than new electricity sales can absorb them.
Residents may hear about construction jobs, investment announcements, and economic development. Their monthly bill delivers a different message. The benefits of a data center are often presented in billions of dollars. The household cost arrives quietly, several dollars at a time, every month, for years.
AI efficiency could leave Americans paying for stranded infrastructure.
Technology companies regularly celebrate improvements in AI efficiency. Faster chips, better cooling, and optimized software can reduce the power needed to perform individual tasks. For consumers, those improvements create a troubling possibility. Utilities could build infrastructure around today’s aggressive energy forecasts only to discover that tomorrow’s AI systems need less electricity than expected.
The data center may operate below its promised capacity, open years late, or never open at all. The fixed costs would remain. A substation cannot be returned like an unwanted purchase. A transmission line cannot be erased from a utility’s balance sheet. A new generating plant may require customer payments for decades.
This is how an AI investment disappointment could become a household electricity crisis. The speculative project disappears, but its infrastructure bill survives. The worst outcome is not simply that AI fails to meet investors’ expectations. It is that utilities prepare for spectacular demand, technology companies retreat, and the public receives the final bill.
Small businesses will be squeezed, as will families.
Households will not suffer alone. Restaurants already pay to refrigerate food, operate ventilation systems, and run cooking equipment. Laundromats depend almost entirely on electricity and water. Grocery stores must keep freezers and display coolers operating every hour. Barbershops, repair businesses, and local retailers cannot negotiate private utility contracts like global technology companies.
When electricity prices rise, small businesses must increase prices, reduce staff, shorten operating hours, or accept smaller margins. That creates a second blow for consumers. Families pay more at home, then pay more again when local businesses pass along higher operating costs. A hyperscale data center may receive customized rates because of its enormous consumption.
A neighborhood store receives the bill available to everyone else. If regulators allow special contracts to shield data centers from the full cost of new infrastructure, the remaining customers may be forced to cover a larger share. Confidential agreements make it difficult for the public to determine whether the largest electricity users are truly paying their way
