Electricity bills at some U.S. businesses now carry a real AI line item. PJM grid-region capacity prices have jumped roughly tenfold since 2024 on data-center demand; one Ohio manufacturer's monthly capacity charge rose from $1,600 to $12,000. **Cutting your own AI software usage won't lower that bill: the cost comes from regional grid economics, not your usage.**
Why Is My Electric Bill Going Up Because of AI Data Centers?
AI data centers are large, round-the-clock electricity customers, and the grid has to build generation and transmission capacity to serve them. The Federal Reserve Bank of Dallas modeled the effect directly: existing data centers have already pushed U.S. wholesale electricity prices up 3 to 5% on average nationwide, and under a high-utilization build-out scenario that could reach roughly 50% by 2028 (Federal Reserve Bank of Dallas, 2026). That's a national average — the real exposure is concentrated in specific grid regions, not spread evenly across every state.
How Much Have AI Data Centers Actually Raised Electricity Prices?
The sharpest numbers are in PJM, the grid operator covering 13 states and Washington, D.C. — including Pennsylvania, Ohio, New Jersey, Maryland, and Virginia — a corridor with heavy data-center buildout. PJM's capacity-market price, what utilities pay to guarantee future power supply, rose from $28.92 per megawatt-day in 2024 to $329.17 per megawatt-day for the 2026/27 delivery year — more than a tenfold increase in two years, which PJM itself attributes to peak-load growth from "intensifying data center buildout" (PowerMag, 2026).
That number isn't abstract. Reuters reported industrial electricity rates up 31% year-over-year in Pennsylvania and 26% in Ohio, against a 7% national average — and named two real companies absorbing it: Belden Brick Company in Sugarcreek, Ohio, saw its monthly capacity charge jump from $1,600 to $12,000; Plaskolite's annual capacity charges rose from $200,000 to $1.2 million (Reuters, 2026).
| Region / Customer Class | YoY Electricity Rate Increase | Why |
|---|---|---|
| Pennsylvania, industrial | 31% | Inside PJM; data-center buildout drove peak-load growth |
| Ohio, industrial | 26% | Same PJM capacity-market pass-through |
| National average, all customers | 7% | Blends PJM states with regions outside the buildout |
| Pennsylvania, residential | 14% | Smaller exposure than industrial/commercial rate classes |
| Ohio, residential | 9% | Smaller exposure than industrial/commercial rate classes |
Does Cutting AI Usage at My Business Lower This Bill?
No — and that's worth saying plainly, because it's the first thing a business owner reaches for. A single hyperscale data center can draw as much power as a small city; the chatbot, scheduling assistant, or AI phone system a small or mid-size business runs is a rounding error next to that load. The price pressure comes from regional capacity-market economics — how much generation a grid operator has to guarantee years in advance — not from how many AI queries your business runs this month. Switching to a different AI vendor doesn't fix it either: every vendor's compute still draws from the same regional grid, so vendor-hopping doesn't move the capacity-market price your utility bill is pricing in.
Who Pays for AI Data Centers' Electricity — Tech Companies or Ratepayers?
Historically, a lot of that cost has been socialized across all ratepayers in a grid region, which is exactly what's showing up in Pennsylvania and Ohio bills. That's starting to change: the U.S. House passed the Ratepayer Protection Act 417-3 on September 16, 2026, which would require large-load data centers over 100 megawatts to pay the full incremental cost of serving them instead of shifting it onto residential and small-business ratepayers — a practice 24 states already require in some form. The bill is now in the Senate (House Energy and Commerce Committee, 2026).
Does This Apply to Montana and the Pacific Northwest?
Not directly, and that distinction matters. Montana businesses run on NorthWestern Energy, not PJM — the sharp, data-center-driven capacity-price spikes hitting Pennsylvania and Ohio bills aren't showing up the same way on a Kalispell or Missoula utility statement. What can still reach a Montana or Pacific Northwest business indirectly is vendor pricing: an AI software company whose own compute runs partly on PJM-footprint data centers absorbs this cost somewhere, and a metered, per-seat or per-minute AI contract can pass it along at renewal. A flat-fee system a business owns outright doesn't get re-priced that way.
What Can a Business Actually Do About Rising AI-Driven Power Costs?
The useful levers depend on whether you're inside an affected grid region or just watching vendor pricing from the outside:
- Check whether your utility sits inside PJM (or another grid region seeing the same capacity-price pressure) before assuming this applies to your bill directly.
- If you are in an affected region, ask your utility about rate class, demand-response enrollment, or a fixed-rate contract — the actual levers large industrial customers use.
- Read the fine print on any AI tool billed per-minute, per-seat, or by usage; ask what happens to that price if the vendor's own compute costs rise.
- Favor a system you own outright over a metered subscription if long-term cost predictability matters more than month-to-month flexibility.