Who Pays if the Data Center Boom Falls Short

The risk beyond today's electricity bill

The financial risk in America’s data center expansion extends beyond the electricity those facilities actually consume. It also includes infrastructure built for demand that may arrive late—or never arrive at all. A June federal laboratory report identifies precisely that exposure: large customers can delay opening, use less capacity than requested or close after utilities have invested to serve them.

A comparison of national forecasts, state oversight findings and industry commitments shows why “paying their share” requires two separate tests: whether data centers cover allocated costs today, and whether other customers are protected if tomorrow’s expansion changes course.

A 322-terawatt-hour spread in projected demand

The scale of the uncertainty is measurable. Lawrence Berkeley National Laboratory’s 2025 Update, published in June 2026, projects U.S. data center electricity consumption of 521 to 843 terawatt-hours in 2030 across its combined uncertainty scenarios. Its central reference case is 649 terawatt-hours. The endpoints represent roughly 9.5% to 15.3% of projected national electricity use.

Our calculation puts the distance between those endpoints at 322 terawatt-hours, with the upper scenario approximately 62% above the lower one: (843 − 521) ÷ 521. That is a scenario spread, not a forecast error or an estimate of wasted power. The model varies assumptions about equipment shipments, chip lifetimes and server operation. It does not establish how much infrastructure any particular utility should build.

Virginia separates current bills from future costs

Virginia’s legislative watchdog documented why even properly allocated existing bills do not settle the larger question. Its 2024 investigation found that rates then in place appropriately assigned costs to the customers responsible, including data centers. Yet it also warned that expansion could raise costs for other customers through new infrastructure requirements and higher energy prices.

The same study projected that a typical Dominion residential customer could see generation and transmission costs increase by $14 to $37 a month, in inflation-adjusted dollars, by 2040. It also warned about infrastructure costs becoming stranded if demand failed to materialize or facilities closed. Those are dated scenario findings, not a measured nationwide surcharge or a current forecast incorporating every subsequent policy change. The audit also recognized economic benefits, particularly construction spending and local tax revenue.

What contracts and state safeguards can cover

There are mechanisms to address the risk. Berkeley Lab’s June 2026 review describes minimum contract terms, exit fees, required minimum payments and financial guarantees in large-customer tariffs. Such provisions can require a company to pay even when consumption falls short. The report also documents regulatory movement, including Dominion’s separate large-load customer class. Its examples show that the issue is already being addressed, although protections vary by utility and jurisdiction.

New York’s July 14 executive order makes the concern explicit. It warns about investments made for demand that does not fully materialize and directs officials to consider a grid fund that could include upfront developer contributions and protection against cancellations. The order’s language matters: it directs consideration of a mechanism; it does not itself establish that a fully funded protection system is operating.

Company commitments and the remaining question

Industry has also acknowledged the responsibility. Microsoft’s January 13 national infrastructure plan committed to seeking utility rates that cover the cost of serving its data centers, including infrastructure, and to coordinating earlier with utilities about projected demand. It described advance electricity contracting and payment for required transmission and substation improvements. Those are published company commitments, not independently verified results for every site.

Read together, these records identify the practical accountability question: what payment obligation survives when a project’s business plan changes? A forecast describes possible demand. A company pledge describes intent. The contract and approved tariff determine which costs remain payable when actual demand disappoints.

Reporting method

This article compares five publicly available government, laboratory and company sources. The percentage calculation is original arithmetic using published scenario endpoints. No interviews were conducted, and this review does not quantify nationwide household costs attributable to data centers.