Back to news
Daily Market Briefing · 16 November 2026

Data-Centre Grid Expansion Is Shifting Costs Toward Rural and Lower-Income Households

A peer-reviewed US study estimates that data centres paid 38% of infrastructure costs attributed to their growth in 2025, leaving most of the balance elsewhere in the rate base.

Who paid for expansion

Cornell researchers analysed demand growth and infrastructure projects across four major US power grids from 2020 through 2025. Their peer-reviewed Nature Energy study attributes $8.25 billion of 2025 infrastructure costs to data-centre growth. Data-centre operators paid an estimated $3.12 billion, or 38%; the remaining $5.13 billion was largely allocated to other customers. Total transmission investment across the studied regions rose from about $2 billion in 2020 to $34 billion in 2025.[1][2]

Wholesale prices mattered alongside wires

The research finds two channels: additional data-centre demand raises wholesale electricity prices, while grid-cost rules spread part of the required transmission investment across households. That means proximity is not necessary for exposure. Residents in New Jersey, Pennsylvania and Ohio can help fund infrastructure connected to data-centre expansion in Virginia because regional grids and tariffs share costs across state lines. The study does not claim that data centres explain every electricity-price increase or that all utility territories allocate costs identically.[1][2]

A regressive burden

Lower-income households experience a larger loss because energy already consumes more of their income. In the study, the lowest-income groups faced additional energy costs reaching as much as 28% of income in the most exposed cases. The share of rural households above a 6% energy-burden threshold increased 76% more than the comparable urban share. These are modelled and regional results, not a national household census, but they show why equal dollar increases can produce unequal economic harm.[1][2]

Peak management may avoid some capital spending

The authors estimate that reducing data-centre load during 22 peak-demand hours a year could materially relieve grid strain in all four regions. Large technology companies may be able to shift flexible computing among sites rather than run every facility at maximum load during the same peak. The policy implication is broader than a special tariff: planners need transparent connection costs, credible demand forecasts and incentives for flexibility. The remaining question is whether utilities and regulators can distinguish firm, near-term projects from speculative interconnection requests.[1][2]

Sources & evidence

  1. Data centres asymmetrically exacerbate energy burdens through wholesale price increases and regressive grid cost shifting (opens in a new tab)
    Nature EnergySource language en

    Peer-reviewed study identity, scope and central conclusion.

    Source note

    Article page was indexed, but full access redirected to an inaccessible identity service during verification.

  2. Data centers need power – lower income, rural residents pay the price (opens in a new tab)
    Cornell ChronicleSource language en

    Methods summary, cost estimates, household distribution and proposed peak-hour response.

    Source note

    Full university report accessible; it describes research by Cornell authors and is not independent journalism.