By Robin Millican, Douglas J. Arent, and David Sandalow | Center on Global Energy Policy, Columbia SIPA, June 23, 2026
Drawing on new LBNL/Brattle data and a CGEP roundtable series, the authors argue rising US electricity prices stem from structural failures — utility incentives, cost allocation, permitting — not data center demand growth itself, which can push prices either way depending on how it’s handled.
The load-growth verdict isn’t guilty on its own
LBNL’s 2026 analysis found residential electricity prices rose 6% nominally in 2025 — more than double inflation — driven by fuel costs, distribution and generation capex, transmission buildout, and storm recovery. Equipment inflation is quietly brutal here: transformer prices are up 89% since 2019, wire and cable 152%. But the paper’s sharper point is that load growth’s price effect isn’t fixed. States with the biggest demand growth from 2019–2025 (Nebraska, New Mexico, North Dakota) actually saw inflation-adjusted prices fall more than a cent/kWh, because cheap wind and solar served new load below system-average cost while spreading fixed infrastructure over more consumption. PJM is the counterexample: pushing capacity payments up an estimated 1.5 cents/kWh through 2025–2026. Problem being solved by the Feds demanding flexibility. db
Three horizons, three different fixes
Near-term, the authors push grid-enhancing technologies (dynamic line rating, advanced conductors) and demand response as underused levers — PPL Electric cut congestion by up to 65% with dynamic line rating, and DCFlex’s Arizona pilot got data centers to shift 10–40% of workload during a simulated peak event. Medium-term, they want performance-based utility regulation (returns tied to cost outcomes, not capital deployment) and cost allocation that makes large loads pay for the upgrades they trigger. The paper frames this as pending in FERC’s RM26-4 docket, with Commission action expected by June 2026 — that action has since landed: on June 18, 2026, FERC answered with six tailored §206 show-cause orders, one per RTO/ISO, rather than a single rule, putting each region on a 30/60-day clock to defend or fix how it allocates large-load costs. Early utility projections (NIPSCO, DTE) claim residential savings from data center contracts, though the authors flag these as unproven forecasts, not results — a 2025 Harvard Law review of nearly 50 data center rate proceedings raised real doubts about isolating those costs from consumer bills. Longer-term, storm and wildfire cost recovery (about 4 cents/kWh in California from wildfire mitigation alone) and cybersecurity get treated as structural, recurring costs rather than one-off shocks — the authors want proactive hardening and liability reform, arguing ex-ante investment beats ex-post recovery.
The throughline: none of this is inevitable. Price outcomes are a function of regulatory choices, not the arithmetic of new demand.
edited and authored by Dave with close collaboration by Claude