Large data-center load shifts grid and capacity costs onto households and other ratepayers

Observable evidence

PJM's independent market monitor attributed $6.3 billion of the most recent capacity auction's $16.4 billion cost (about 38%) to data-centers, and $29.4 billion of $63.6 billion (about 46%) across the last four auctions (Utility Dive). Capacity prices rose from $28.92 per MW-day for 2024/25 to $269.92, then $329.17, $333.44 and $325 for 2028/29—more than an eleven-fold increase in two years. The 2027/28 auction cleared at a capped $333.44 and would have reached roughly $530 uncapped, while procurement fell 6,625 MW short of the reserve margin (Utility Dive).

Capacity is roughly 20% of a ComEd retail bill, which has risen about 50% in two years (Citizens Utility Board).

The queue is substantially speculative

ERCOT's large-load interconnection queue grew from 63 GW at end-2024 to 226 GW by November 2025, roughly 77% data-centers, against only about 23 GW of new generation added across 2024 and 2025. Much of this is duplicate load: the same project registered in multiple places while shopping for terms (Latitude Media). Planning against a phantom forecast results in ratepayers funding capacity that was never needed.

The mechanism, where documented

Confidential special contracts are the least visible element. Harvard's Electricity Law Initiative cites Duke Energy litigation records showing a plan to shift $325 million of discount costs to other ratepayers, and an Exelon estimate of a $58–$140 million cost shift from a single co-location arrangement (Harvard ELI). This is a law-school analysis citing underlying filings, not peer-reviewed work.

Generation planning is also affected. Georgia Power's July 2025 IRP projects about 8,500 MW of load growth in six years and keeps the Scherer and Bowen coal plants (4,000 MW combined) running to 2038 rather than 2028 and 2035 (Utility Dive). In December 2025 the state PSC approved 9,985 MW of new generation, about 80% for data-centers (Georgia PSC).

What this issue does not claim

PolitiFact rated "mostly false" the claim that bills near data-centers rose 267%, because that figure is wholesale and wholesale is only 30–50% of a retail bill. US residential electricity rose about 42% over five years nationally, and attribution specifically to data-centers is confounded by weather, fuel prices and ageing infrastructure (PolitiFact). The defensible claim is narrower: capacity and transmission cost attribution is documented by independent monitors, and the rules determining who pays were written before loads of this scale existed.

Who is affected

Every other customer on the same system—households, small businesses, and industrial users who did not choose the load and cannot negotiate special contracts.

Impact if unaddressed

Costs caused by a small number of very large, creditworthy customers are socialised across a customer base with no ability to respond, and utilities build generation against forecasts inflated by duplicate requests. Both effects are structural consequences of tariff design.