AI data center power demand is now a primary driver of US electricity price increases, with EIA data showing a 7.4% rise year over year and prices projected to outpace inflation through at least 2026.
DOE expects US data centers, heavily AI-weighted, to jump from 4.4% to as much as 12% of national electricity consumption by 2028, meaning GPUs and racks are effectively competing with households for scarce grid capacity.
LBNL and watchdog analyses point to real grid and cost consequences: $16.6 billion in near-term grid upgrades on the largest US power system and a 23% real increase in retail electricity prices since 2019, contributing to higher arrears and household utility debt.
This demand spike risks slowing electrification and renewables buildout, as higher retail prices and grid constraints make it harder to justify new clean projects while AI operators still secure power at scale.
The political response is forming around curbing the pass-through from AI data center expansion to residential bills, with state leaders in New Jersey and Virginia explicitly targeting data center-driven cost increases.
For AI infra planners, the signal is clear: power is no longer a background assumption but a contested, politicized input, and future GPU deployments will be gated as much by regulated grid economics as by chip supply.
The article is worth a full read if you care how AI buildout is starting to reshape utility regulation, cost allocation, and siting strategies across US data center markets.
Source: AI Data Centre Electricity Demand Increasing US Utility Bills, Putting Households in Debt | IBTimes
