Data centers' share of U.S. electricity consumption rose from 1.9 percent in 2018 to 4.4 percent in 2023. Grid power directed to data centers surged 22 percent last year alone. Jeremiah Johnson, associate professor of civil and environmental engineering at North Carolina State University, published a study in Environmental Research Letters last week projecting that national average wholesale electricity costs rise 6 to 29 percent by 2030 under the current buildout trajectory. Virginia, the most data center-dense state in the country, faces a potential 57 percent spike in electricity generation costs by the same date.
These are not tail-risk projections. Data centers could account for up to 17 percent of all U.S. electricity usage by the end of the decade. U.S. utilities filed a record $31 billion in rate increase requests during 2025. The cost of AI infrastructure is moving from the tech industry's balance sheet to the residential utility bill, and the public is noticing.
Gallup polling finds seven in ten Americans oppose AI data center construction near their homes. Fifteen percent cite higher utility costs specifically as their concern. A YouGov and Economist poll finds more than half of Americans say AI development is moving too fast. More than $156 billion in planned data center construction was stalled or halted across 48 projects in 2025, not by regulatory action but by a combination of permitting friction, community opposition, and grid interconnection delays.
Johnson's framing in the study is direct: "The challenge here is the magnitude of this demand we're talking about is really big. It's at a scale that dwarfs some of the other changes" to the power sector. The comparison point is not other large industrial loads. It is the entire history of demand-side changes to the U.S. grid.
Johnson's model distinguishes between a clean energy scenario and a fossil-dominated scenario. Without clean energy incentives, natural gas would supply roughly 70 percent of the additional generation required to meet data center demand; coal, wind, and solar would split the remainder. With restored incentives, natural gas drops to 41 percent and wind rises to 29 percent. The emissions implication: data center expansion could increase CO2 from electricity generation by up to 28 percent by 2030 under the high-fossil scenario.
For cooling vendors, the rate increase trajectory matters because it directly affects the economic case for liquid cooling. A facility that reduces its PUE from 1.4 to 1.15 by moving from air cooling to direct-to-chip liquid cooling saves 17 percent of facility electricity consumption on the cooling load alone. In Virginia, where electricity prices are projected to rise up to 57 percent by 2030, that efficiency gain is worth substantially more in 2030 than it is in 2026. The capex decision made today is priced against 2026 energy costs. The opex savings will accrue at 2030 rates.
The EIA's Annual Energy Outlook 2026 and the Johnson study are describing the same structural reality from different time horizons. The EIA projects 2050. Johnson projects 2030. The near-term number is what drives the next legislative session in Richmond.