For a household in northern Virginia, the connection between an AI data centre and next month's electricity bill is becoming harder to ignore. Dominion Energy, the dominant utility in the state, has asked regulators to approve increases of roughly $20 per month for the average residential customer over two years, citing data centre demand alongside inflation and higher fuel costs.
That request sits within a broader national pattern. Utilities filed more than $29 billion in rate increase requests in the first half of 2025 alone — double the pace of the same period in 2024 — with changes expected to affect around 40 million customers, according to the Energy and Environment Study Institute. Goldman Sachs reported electricity prices rose 6.9% year-on-year in 2025, more than twice the overall inflation rate of 2.9%, and projected a further 6% increase through 2027, attributing 40% of demand growth to data centres.
The United States is in the middle of an unprecedented data centre construction surge, driven largely by AI investment from major technology companies. Those facilities require enormous and continuous quantities of electricity, and connecting them to the grid means spending heavily on power lines, transformers, and generation capacity. Under current utility regulation in most states, such infrastructure costs are typically spread across all customers, not just the large commercial users whose needs prompted the spending.
The most specific figures come from PJM, the grid operator covering the mid-Atlantic and parts of the Midwest. Monitoring Analytics, PJM's independent market monitor, attributed $9.3 billion — around 70% — of a single year's capacity cost increase to data centre demand, and projected $23 billion in total customer price increases lasting at least through the end of 2028. A Bloomberg News analysis found wholesale electricity prices at grid nodes near major data centre clusters rose by as much as 267% between April 2020 and April 2025.
That last figure requires care. It describes wholesale prices at specific nodes on the transmission network — the prices utilities pay to buy power in bulk — not what consumers see on their bills. When Senator Elizabeth Warren cited the 267% figure as applying to residential bills near data centres, fact-checkers at WRAL rated the claim mostly false on exactly that basis. The distinction matters: wholesale node prices are volatile and localised; retail bills move more slowly and reflect a broader mix of costs.
The U.S. Energy Information Administration recorded an average residential electricity price increase of approximately 42% over the five years through mid-2025, taking the national average to around 19 cents per kilowatt-hour, up from about 13 cents before 2019. That is a real and sustained increase, but it reflects multiple pressures — ageing infrastructure, fuel costs, supply chain inflation — not data centres alone.
A March 2026 study by the Institute for Energy Research complicated the picture further. It found no statistically significant correlation between data centre concentration and state-level electricity prices or price growth rates, and noted that states with faster electricity sales growth had actually experienced lower average price increases than states where sales were shrinking. The IER is a market-oriented think tank, and its methodology and conclusions remain disputed, but the finding illustrates why isolating the data centre contribution from the overall cost trend is genuinely difficult.
Market structure adds another layer. SemiAnalysis, an industry research firm, has argued that the Base Residual Auction mechanism used by PJM to set capacity prices amplifies price signals in ways that can produce large swings independent of underlying demand changes. On that reading, the design of the market itself — not data centre growth alone — is partly responsible for the surges the independent monitor has documented.
What remains unknown is considerable. No reliable estimate exists for the precise share of household bill increases attributable solely to data centres rather than to fuel costs, inflation, or infrastructure ageing. The distributional impact on low-income households has not been quantified beyond general references to affordability pressure. Legislative and regulatory proposals to require data centre operators to bear a larger share of grid upgrade costs are under active discussion in multiple states, but their legal viability and political prospects are unclear; reporting suggests they may be difficult to implement in practice.
For now, the evidence supports a limited but real conclusion: data centre demand has materially increased wholesale electricity costs in parts of the country, particularly within the PJM region, and those costs will work their way into retail bills over time. The scale of that pass-through, and whether policy can redirect more of it toward the industries driving demand, is where the argument is now being fought.
Comments (0)