Will a data center raise or lower electric bills here?

A data center’s total draw includes cooling overhead, measured by PUE (power usage effectiveness). The industry average is 1.56; efficient new builds run 1.2–1.3.⁴⁸ Nationally, data centers used 4.4% of U.S. electricity in 2023, heading to a projected 6.7–12% by 2028.¹¹

The national headlines about data centers raising rates come from booming regions where utilities must build expensive new plants and lines to serve them. Kentucky Power’s problem is the mirror image: as coal employment collapsed, mines and businesses closed and people left — so the utility sells less power while spreading the same fixed costs (poles, wires, plants) over fewer customers. That dynamic, documented for years, is a core reason Eastern Kentucky bills are so high.⁶¹

The numbers are stark. In its 2025–26 rate case, Kentucky Power sought a 14.6% residential increase; the PSC approved 6.63% over two years, taking the average residential bill from $183.37 to $194.13 — among the highest in Kentucky even though the state’s average power price is well below the national average.⁵⁹Customers will also pay about $2/month more to keep the aging Mitchell coal plant running, in a decision the PSC approved only “to prevent a worse outcome.”⁶²

Against that backdrop, Kentucky Power has said publicly that attracting large new loads is “the best vehicle for eventually balancing or lowering electric rates” — and it is already building a 345 kV substation to serve TeraWulf’s 1,000+ MW campus near Ashland.⁶³²¹

• Spreading fixed costs.A big, steady customer adds revenue against infrastructure that already exists. In Georgia, regulators credited the large-load pipeline with enabling a three-year base-rate freeze, and the settlement requires at least $556 million/year of large-load revenue to put “downward pressure” on everyone else’s rates (consumer groups note that’s pressure, not a guaranteed decrease).⁶⁷⁶⁸
• Data centers can pay their full freight. Virginia’s legislative auditor (JLARC), with an independent cost-of-service study, found data centers there currently pay the full cost of their service — rates were allocating costs to the customers causing them.¹⁴
• Pikeville’s specific advantage: the city says the initial 25–30 MW phase can be served with existing power capacity in the industrial park, without new transmission²— meaning the cost-causing buildout that drives rate increases elsewhere isn’t needed for phase one.

• The same JLARC study’s warning:future data center growth in Virginia “will likely increase system costs for all customers” as new plants and lines get built — both findings are true at once.¹⁴
• Harvard’s Electricity Law Initiative reviewed ~50 regulatory proceedings and concluded utilities are structurally incentivized to build for Big Tech and shift costs to captive ratepayers, often behind confidential contracts.⁵⁰
• The PJM capacity-market spike (this affects us): Kentucky Power buys capacity through PJM, the 13-state grid market. PJM capacity prices exploded from $28.92/MW-day to $269.92 and then to the price cap in successive auctions,⁵²and PJM’s independent market monitor attributes 40% of the December 2025 auction’s $16.4 billion cost — and $21.3 billion across three auctions — to data center load forecasts.⁵¹ Estimated bill impacts ran ~$16–18/month in AEP Ohio and western Maryland territory.⁵³ A Bloomberg analysis of 25,000 grid nodes found wholesale prices rose up to 267% since 2020 near major data center clusters.⁵⁴
• The honest counterpoint: national electricity prices rose ~40% since 2021 for many reasons — fuel costs, grid hardening, inflation — and analysts dispute how much is attributable to data centers outside hotspot regions.⁵⁵

Regulators nationwide now use special “large-load tariffs” to ring-fence data center costs. The model is AEP Ohio (Kentucky Power’s sister company): data centers over 25 MW must pay for at least 85% of their subscribed capacity every month whether they use it or not, for up to 12 years, plus exit fees and collateral — so if the facility shrinks or leaves, ratepayers don’t hold the bag.⁵⁶

Kentucky has started down the same road, but with higher thresholds. The PSC approved LG&E/KU’s “Extremely High Load Factor” tariff — minimum 80% of forecast energy for 15 years, with collateral up to $100 million — but it only applies to customers of 100 MW or more.⁵⁷Kentucky Power’s own large-load tariff filing (20-year terms, heavy collateral) applies at 150 MW or more.⁵⁸A 2026 bill to require all data centers over 100 MW to “pay their own way” died without a hearing.⁷²

Two failure modes are well documented. First, stranded costs: if infrastructure is built for a data center that never materializes or leaves early, remaining customers can inherit the bill — the central objection when Kentucky regulators let LG&E/KU build $3 billion of gas plants before any data center contracts were signed.⁶⁵ Standard protections: minimum monthly payments, long contract terms, exit fees, and collateral.⁵⁶ Second, phantom load: developers shop the same project to multiple utilities, producing 5–10× more interconnection requests than will ever be built — and inflated forecasts are already costing PJM customers real money.⁶⁶⁵¹Kentucky’s own energy commission (EPIC) warned in June 2026 that the state must set cost-allocation rules before approving major investments, because once assets are built, regulators have limited tools to reassign their costs.⁶⁴

A 25–30 MW load sits at the boundary where existing distribution and sub-transmission may suffice — which matches the city’s statement that the park’s existing capacity can serve phase one.² Growth to 75–100 MW is a different matter: loads that size typically require a dedicated transmission-level substation and new lines, with multi-year timelines and interconnection costs commonly in the tens of millions.⁷³The developer’s own framing — expansion only “if additional power capacity becomes available” — reflects that.² Who pays for any new infrastructure is precisely what tariff and contract terms decide.

Worth knowing about the broader landscape: some operators now sign demand-response deals to pause heavy computing during grid peaks — Google does this with Indiana Michigan Power, another AEP company⁷¹ — and a Duke University study found modest flexibility (curtailing 0.25% of hours) could free up ~76 GW of existing U.S. grid capacity.⁷⁰Flexibility commitments are another negotiable protection. Meanwhile some projects bypass the grid entirely with on-site generation (xAI’s Memphis turbines) or dedicated nuclear deals (Microsoft’s Three Mile Island restart)⁶⁹ — nothing public suggests that model here.

Honest summary of the evidence: a steady industrial load served from existing capacity, paying full local taxes²and bound by minimum-payment and exit-fee terms, is the scenario in which a data center plausibly helps stabilize a shrinking utility’s rates — the outcome Kentucky Power itself is betting on.⁶³A facility that triggers new generation or transmission whose costs are socialized, or that leaves early, is the scenario in which everyone else pays — the outcome documented in PJM’s capacity market and Virginia’s forecasts.⁵¹¹⁴Which one Pikeville gets depends on terms that haven’t been negotiated yet — which is why the contract details matter more than the slogans on either side.