Who Pays When the Grid Blinks?

There is a boundary line running through every building you have ever lived in, and most people never have a reason to find it: the point where the electric utility’s responsibility stops and yours starts. On a house it’s usually at the meter. Everything past it — the panel, the wiring, the appliance that dies in a brownout — is yours to fix, whoever caused the interruption.

Chapter 14 of Everybody’s Science tells the version of this I know best, from my family’s farm in New Mexico, where the stakes on the wrong side of that line are well pumps, repair bills, and a man working close to 480-volt equipment to avoid paying for an electrician. I won’t retell it here — the book does it properly. What matters for this page is the rule the story exposes. As the chapter puts it, the cost of an unreliable system “doesn’t land on whoever controls the reliability. It lands on whoever’s holding the switch when it fails.”

That rule was written for yesterday’s grid, and today’s grid is being asked to do something it has never done: absorb the fastest sustained demand growth in a generation, most of it from data centers, on infrastructure and rules designed before anyone imagined a single customer requesting the power of a small city. Which means the who-pays question is about to be put to voters, county boards, and utility commissions all over the country — probably including yours — and it will usually arrive dressed as something else: a zoning application, a rate case, a moratorium, a reactor siting study.

Same evidence, opposite answers — and why that’s not a failure

Northern Virginia spent decades courting data centers on purpose and got what it asked for: enormous tax revenue, real jobs, and a strained grid whose regulators are now deciding, docket by docket, how much of the cost belongs to the data centers and how much lands on everyone else’s monthly bill. Several Iowa farm counties looked at the same industry and froze it out entirely, at least for now, citing water, power, and the fear of what their residents’ bills would do. It is tempting to grade one answer as informed and the other as reactionary — in either direction. The more accurate reading is that both are careful rankings made under the same uncertainty, by communities that weighed jobs, wells, growth, and groundwater differently. When two careful rankings disagree, what settles the matter isn’t more data alone — it’s argument made in the open, by people who have to answer for it. That is the Reaching Standard doing county-level work, and the book builds it in full in Chapter 12.

The three questions, aimed at a grid vote

You will never hold the whole docket. You don’t need to. Before any vote or public comment that touches this:

1. What specifically does this build or permit? Megawatts, gallons, acres, years — a proposal that can’t answer in units isn’t ready for your yes.
2. Who is accountable if the promises don’t hold? Named parties, enforceable terms — not projections.
3. Who pays when it blinks? Find the meter line in the deal: when reliability fails or costs overrun, does the burden land on the party who controls the system, or on whoever happens to be holding the switch?

Run the check yourself — it’s all public

This is the part a printed page can’t do for you, and it takes an evening, not a degree.

Your utility’s rate case. Every state has a utility commission, and every rate increase runs through a public docket you can read — in Virginia it’s the State Corporation Commission, and the current Dominion case is where the new large-user rate class was created. Search the filing for who bears “demand” costs, and check who intervened.
Your region’s waiting line. New power plants queue for grid connection through your regional operator — PJM in the mid-Atlantic, MISO in the Midwest, ERCOT in Texas — and the queues are published. How many years is the wait where you live, and what’s stuck in it?
Your county’s posture. Zoning ordinances and moratoria are public record. Does your county have a data-center policy at all, or will it be improvising one the week an application lands?

The water half, briefly

Cooling a large data center can consume millions of gallons of water a day — and in dry country, that demand competes directly with wells and irrigation drawing on the same aquifers, a collision Chapter 9 documents in my family’s own basin. The reason this belongs on a checklist rather than in a lament is that water-free, closed-loop cooling already exists and is already deployed on some of the driest ground in the country. So when an application reaches your county, the question isn’t whether the technology exists. It’s whether anyone in the room makes the applicant say, in units, which cooling design they’re bringing and what it draws. That is a one-sentence public comment, and anyone can file it.

The figures, as of this update

Updated August 2026, matching the book’s mid-2026 snapshot (“Notes — Time-Bound Figures,” Ch. 14 note). These numbers move; this page is where they stay current after the book is printed.

U.S. data-center power demand: roughly 31 gigawatts in 2025, projected to 41 GW in 2026 and 66 GW by 2027 (Goldman Sachs Research). The EIA’s January 2026 outlook called for the strongest four-year electricity demand growth since 2000. The national interconnection queue stood near 2,600 gigawatts — more than the country currently uses — with median waits approaching five years. Virginia’s new large-user rate class (GS-5, 25 MW and up, effective January 2027) requires 85 percent minimum payment of contracted demand on 14-year contracts, alongside a residential increase of about $13.60 a month for the average Dominion customer. Eight Iowa counties held data-center moratoria by midsummer 2026. The NRC’s Part 53 licensing rule was finalized March 26, 2026, more than a year ahead of its statutory deadline — the standing proof that these bottlenecks are choices, not physics. Domestic HALEU reactor fuel: roughly 900 kilograms a year from one demonstration cascade, against multi-ton annual demand; DOE named four build-out contractors in January 2026, targeting adequate supply by the mid-2030s. The power-engineering workforce gap: 450,000 to 1.5 million additional engineers needed globally by 2030 (Kearney/IEEE).

The book’s side of this story: Chapter 8 (a farm’s morning, run on water and electricity), Chapter 9 (the same water, somewhere drier), and Chapter 14’s “Who Pays When the Grid Blinks” — the story and the argument this page hands you the tools for.