Nobody Put a Meter on It

In the first three months of 2026, local opposition blocked or delayed at least seventy-five data center projects worth roughly $130 billion. That is not a trend line. That is the entire preceding year compressed into a quarter. Active opposition groups more than doubled, to 833, across forty-nine states, and more than three hundred bills landed in statehouses in the first six weeks alone.

Something moved. It is worth being precise about what.

New Jersey passed a law requiring facilities over fifty megawatts to commit to paying at least eighty-five percent of their projected power costs for a decade. Virginia added a levy of $0.011 per kilowatt-hour. New York passed a one-year moratorium on permits for anything above twenty megawatts. Texas ordered an audit of every data center connecting to its grid, which functioned as a pause on roughly eighteen hundred projects. These are not gestures. These are enforceable numbers attached to specific balance sheets, written into statute in a single legislative season, in states that agree on almost nothing else.

Now notice what they measure.

Water has a gauge. Somebody installed it, somebody reads it, and the reading is admissible. When a Virginia groundwater study concluded in July that new facilities would be unlikely to secure reliable withdrawal permits under current conditions, that conclusion had a unit behind it. Power has a meter, and more to the point, the meter produces a bill that arrives in an envelope at a residential address every thirty days. You do not have to persuade anyone that their electricity bill went up. They are holding it.

That is the whole story of why this moved so fast. Not virtue. Not that communities finally woke up to Big Tech. The two harms that were metered got legislated across forty-nine states in one session. The third harm did not.

The harm with no gauge

Everyone in these hearings knows the jobs arithmetic. It is not hidden and it is not disputed.

Virginia’s own Joint Legislative Audit and Review Commission found that a typical 250,000-square-foot facility employs about fifty full-time workers once it opens, roughly half of them contractors, after employing as many as fifteen hundred during a twelve-to-eighteen-month build. Outside Reno, a 1.1-million-square-foot campus projected seventy-three permanent jobs over a decade against more than four thousand construction positions. In Illinois, a Microsoft facility took more than thirty-eight million dollars in sales tax exemptions and produced twenty permanent jobs.

Communities say this out loud. They say it in the same meetings where they say the water thing and the electricity thing.

And in the last week of August the building trades began saying it back. Unions and construction trade groups have entered the fight from the other side, warning that halted approvals endanger thousands of building jobs and threatening to withhold support from candidates who oppose the projects. They arrived with a number: one electricians’ local reports its members worked twenty-eight million hours last year against fourteen million a decade earlier. That is a gauge, and it is behaving exactly as gauges behave. It is moving votes.

So the jobs argument did not go home. It came back metered.

But look at what that meter counts. Hours are throughput. They measure work passing through bodies, say nothing about what those bodies became, and read zero the day the build ends. The apprentice who topped out and the hand who swept the same floor for the same hours appear in that number identically.

The formation still has no gauge. It has just acquired a very convincing decoy.

Why? Because a construction boom is itself a kind of meter — a bad one, but a real one. Fifteen hundred hard hats on a site is a countable, photographable, ribbon-cuttable fact. Fifty full-time employees three years later is also countable. What is not countable, in any instrument a county commission can enter into the record, is the thing actually being traded: the formation those fifteen hundred people were acquiring, and the formation the next fifteen hundred will not acquire, because the work moved on and nothing accumulated locally except a slab and a substation.

There is a second, larger version of the same invisibility, and it is the one people are most anxious about even though they cannot say it cleanly. The concern is not only that the building employs fifty people. It is that what the building computes may quietly close the first rung of the ladder everywhere else. Stanford’s Digital Economy Lab found a sixteen percent relative decline in employment for workers aged twenty-two to twenty-five in the most AI-exposed occupations. Among software developers in that age band the decline runs closer to twenty percent, while employment for older developers at the same firms grew.

And here is where I have to be honest, because the honesty is the argument.

That finding is contested. David Deming at Harvard points out that the junior hiring decline appears to start about six months before ChatGPT was released, which is an awkward fact for anyone who wants a clean causal story; he suspects remote work. The Economic Innovation Group attributes much of the pattern to the sharpest monetary tightening cycle in four decades. Reasonable economists are genuinely unsure, and anyone who tells you otherwise is selling something.

I used to think that uncertainty was the weakness of the case. It is not. It is the mechanism.

The unmetered harm is always contestable. That is what unmetered means. Water was contestable too, right up until somebody put a gauge on the aquifer and published the number, at which point the contest ended and the statute got written. The absence of a meter does not indicate that the harm is small. It indicates that no one has yet built the instr