Is AI raising your electricity bill?
The evidence, both directions, with dates
| Finding | Direction | Source and date |
|---|---|---|
| PJM capacity auction, delivery year 2026/2027, cleared at $329.17/MW-day — about 22% above the previous record | Upward pressure | PJM auction results, released 22 July 2025 |
| Independent market monitor attributed a majority share of the prior auction's increase to data centre demand, in the billions of dollars recovered from customers | Upward pressure | Monitoring Analytics analysis of the 2025/2026 auction |
| US average residential price about 18.02 ¢/kWh in 2026 vs about 17.29 ¢ in 2025 | Context, not attribution | EIA Short-Term Energy Outlook |
| From 2015 to 2024, average retail prices decreased about 3.5% for every doubling of data centre capacity | Downward pressure | Electric Power Research Institute |
| Connection queues effectively closed around several European hubs; multi-year waits for large connections reported by network operators | Cost shifts into network tariffs | European grid operators and regulators, 2025–2026 |
Every source above is named and linked on the data sources page. We publish no figure we cannot attribute to the body that produced it, and we do not convert these into "$X on your bill" — that number depends on your utility and your tariff, and anyone quoting it for a national audience is guessing.
Why both things can be true
A large, steady customer can lower everyone's average cost, because the fixed cost of poles, wires and substations gets spread across more kilowatt hours. That is the EPRI finding, and it describes a decade when the grid had room.
The same customer raises everyone's cost when it arrives faster than the system can absorb: capacity has to be procured at a higher clearing price, new transmission has to be built, and the question of who pays for the connection gets settled by tariff design and regulatory proceedings. That is the PJM picture.
So the real variable is the rulebook in your region — whether large loads pay their own way, and whether the capacity cost lands on residential customers. That is a decision made in dockets, and it is the subject of the second judgement on this site.
Europe: a grid access story
In Europe the binding constraint has been connection rather than commodity price. New data centre connections around Dublin and in parts of the Netherlands and Frankfurt have been restricted for years, and operators report multi-year queues for large connections. Where reinforcement is funded through regulated network tariffs, households carry a share. Ireland is the most studied case and the one with the sharpest political argument attached.
For a country's actual level, use Eurostat's household electricity price series rather than a campaign figure. We link the series instead of reprinting numbers that change every half-year.
The AGI-2027 Thesis Tracker — our parent site scores the "AGI by 2027" predictions against evidence, including the capital expenditure build-out that puts these loads on the grid in the first place.
Reading as of 2026-09-06; the parent site is authoritative. agiscorecard.com/progress-index · the capex question
What would change this verdict
| Watch | Flip condition | Status (2026-09-15) |
|---|---|---|
| Capacity clearing prices | PJM and neighbouring markets clear materially lower for two consecutive auctions → "upward pressure" is downgraded to a one-off | 2026/2027 cleared at a record $329.17/MW-day |
| Cost allocation rules | Large-load tariffs that make data centres carry their own capacity and connection costs become the norm → the household link is cut at the source | Being litigated and legislated region by region |
| Retail price data | EIA and Eurostat series show residential prices flat or falling while data centre load keeps growing → the EPRI mechanism is winning again | US residential average projected up about 4% for 2026 |
| Demand itself | The build-out slows materially → this whole page becomes historical | Parent tracker at 62.5/100; capex still expanding |
Frequently asked
Is AI raising my electricity bill?
In some regions yes, and the route is indirect. Data centre demand shows up first in capacity and grid costs rather than in the price of electricity itself. The clearest single data point is the PJM capacity auction for the 2026/2027 delivery year, which cleared at a record $329.17 per megawatt-day, about 22% above the previous record. Those capacity costs are recovered from customers across the PJM footprint. Outside such regions the effect on a household bill can be small or hard to separate from fuel prices, weather and utility investment that would have happened anyway.
Is there evidence pointing the other way?
Yes, and pages that hide it are not being straight with you. Research from the Electric Power Research Institute found that from 2015 to 2024 average retail electricity prices decreased by about 3.5% for every doubling of data centre capacity, because large steady loads can spread fixed network costs over more kilowatt hours. The honest reading is that the sign of the effect depends on whether new load arrives with the network investment already paid for, and on who is made to pay for the new capacity. That is a rules question, not a technology question.
How much are US electricity prices actually going up?
The US Energy Information Administration's Short-Term Energy Outlook put the average US residential price at about 18.02 cents per kilowatt hour for 2026, up from about 17.29 cents in 2025 — roughly a 4% rise. That is an average across a country where regional differences are large, so your own bill can move very differently. This site's bill decomposer exists precisely because a national average tells you nothing about your own two bills.
What is happening in Europe?
The European picture is about grid access more than wholesale price. Several hubs have effectively closed the queue: new data centre connections around Dublin and in parts of the Netherlands and Frankfurt have been restricted for years, and network operators report multi-year waits for large connections. Where network reinforcement is paid for through regulated tariffs, households carry a share of it. Ireland is the most studied case, and Eurostat's household electricity price series is the neutral place to check any country's level rather than a campaign figure.
What can I actually do about my own bill?
Three things, in order. First, separate price from usage: take your last two bills and work out how much of the change came from the rate and how much from kilowatt hours used, because people routinely blame the wrong one. Second, find out which mechanism governs your area — a capacity market, a large load tariff, or a pending rate case — because that determines whether new data centre load can be charged to you at all. Third, put your objection where it counts: rate cases and tariff dockets at your state or national regulator are the decision point, and they take public comment.
Split my own bill: price vs usageWhich rules apply where I liveThe data sources
Written by the AGI Scorecard editorial desk (about us). Not energy or financial advice. Last updated 2026-09-15.