Data centers shift grid costs onto households, Cornell study finds

Households are helping finance the power infrastructure behind the AI boom, even when the data centers driving that spending are in another state. A Cornell study published November 9 in Nature Energy finds that data-center expansion raises electricity prices and shifts infrastructure costs disproportionately onto lower-income and rural residents.

Researchers Cheng Feng and Fengqi You examined four major U.S. power grids from 2020 through 2025. Their analysis combined records for 4,000 existing and proposed data centers with electricity prices, utility rate filings, transmission plans and federal household and employment data. The paper uses causal analysis of price records alongside regulatory documents to trace both price effects and how infrastructure bills are divided.

New demand, backward-looking bills

The study identifies two routes from data-center growth to household costs. The first runs through wholesale electricity markets, where utilities buy power before delivering it to homes. The researchers estimate that data centers increased those prices by about 20% in California’s southern Bay Area and 15% in Virginia by 2025. New York and New England showed no clear increases attributable to data-center expansion.

The second route is the bill for expanding the grid itself. Cost-sharing formulas often rely on previous electricity consumption, rather than the new demand prompting construction. A proposed data center needs infrastructure before it can operate, but has little or no consumption history. Existing households therefore take on much of the cost under those formulas, You explained in Cornell’s account of the research.

Across the four grids, transmission investment rose from $2 billion in 2020 to about $34 billion in 2025. The researchers attributed $8.25 billion of the 2025 infrastructure bill to data-center growth. Data centers paid $3.12 billion of that amount; the remaining $5.13 billion fell largely on residents. Separately, the paper reports that residential customers received a median 47% allocation of regulated infrastructure costs despite contributing much less to demand growth.

38%Paid by data centers

Data centers covered $3.12 billion of the $8.25 billion in infrastructure costs attributed to their growth.

62%Remaining share

The remaining $5.13 billion fell largely on residents across the four grids studied.

The burden travels beyond the server campus

The household effect was uneven. Cornell reports that the share of rural families facing a high energy burden—spending more than 6% of income on energy—rose 76% more than among urban residents. That is a comparison of increases in the share of heavily burdened families, not a claim that every rural electricity bill rose 76%. The lowest-income households faced extra energy burdens of up to 28 percentage points.

Geography offers no simple shield. Residents in New Jersey, Pennsylvania and Ohio helped fund grid upgrades linked to data-center growth in Virginia, the researchers found. Shared electricity networks connect the communities physically, while cost-allocation rules connect their finances. A household can therefore face costs from development well beyond its own local planning jurisdiction.

The study also estimates economic benefits: each gigawatt of data-center capacity can create 15,000 jobs across the economy, including an average 2,700 directly associated with the facility, though not necessarily on-site. Returns varied, however. The researchers found that an existing qualified workforce was a stronger indicator of high returns than the amount states spent on incentives.

Less peak demand, different payment rules

Feng and You propose changing both grid use and who pays. Their analysis found that reducing data-center power use during just 22 regional peak-demand hours each year could relieve strain and minimize expansion needs across all four regions.

• Shift computing between regions: companies with multiple facilities could move operations elsewhere during local peaks, although You said implementation requires coordination.
• Charge for added demand: states could bill data centers directly in line with the new demand they bring, or direct funds to low-income households.
• Tie incentives to results: tax benefits could depend on job creation and higher wages, rather than leaving those economic benefits less assured.

For readers seeking to examine the estimates, the authors have released their analysis code, aggregated data and regulatory records. One reproducibility limit remains: the facility-level data-center inventory is commercially licensed and cannot be redistributed. Access to that underlying inventory requires obtaining it from S&P Global under its licensing terms.