The Plant Doesn’t Remember

The claim, and the one part of it that is wrong

The version of this that goes around online runs roughly like this. A startling share of the people who run water plants, grid substations, commercial boilers, and municipal utilities are in their late fifties and sixties. The work depends on unwritten knowledge about legacy systems that have been patched together for decades. Young people were steered away from these paths, public pay scales never kept up, and in the next ten years the people who keep clean water flowing are going to walk out the door with no pipeline behind them. And almost nobody is preparing for it.

Three quarters of that is correct, and federal documents say so in about the same words. The last quarter is wrong, and the way it is wrong turns out to be more interesting than if it had been right.

Here is the paper trail. In January 2018 the Government Accountability Office published a report with a title that reads like a shrug and a warning at the same time: Recruiting Approaches Helped Industry Hire Operators, but Additional EPA Guidance Could Help Identify Future Needs. The utilities GAO interviewed were, on the whole, still filling their jobs. But some were struggling to hire certified operators, larger ones were struggling to hire electricians and mechanics, and nobody at the federal level had a good way to see what was coming. GAO’s recommendation was almost administrative in its modesty: add strategic workforce planning questions to the inspection guidance. Ask utilities not only whether they are staffed today, but what they will need later.

That same year, Congress created what became EPA’s Innovative Water Infrastructure Workforce Development Grant Program. In 2020 EPA published America’s Water Sector Workforce Initiative: A Call to Action, built on the finding that about one third of the drinking water and wastewater workforce could become eligible to retire within the decade. The first grants went out in October 2021: ten organizations, $3.8 million. In July 2024, EPA sent more than $20 million to thirteen more. In August 2026, another $10.8 million was put on the table, with applications due in October.

So: not unnoticed. Noticed continuously, by multiple administrations, for the better part of twenty years. Which poses a harder question than the one the internet asks.

How one generation ended up holding the whole plant

Start with how the buildings got there. Between roughly the 1930s and the 1970s the United States built an enormous amount of physical infrastructure at once: generation and distribution, municipal water, wastewater treatment, highways, dams, district heating, telecommunications. New plants need staff on day one. So utilities hired large cohorts of young workers inside a fairly narrow window, and those workers stayed, because these were careers rather than jobs.

For forty years that looked like excellent management. Turnover was low. Positions were filled. Recruiting was not urgent, because nothing was vacant. A municipal water department could go a decade barely hiring anyone and never once feel understaffed.

The trouble is that the metric everyone was watching cannot detect this problem. Headcount only moves on the day somebody leaves. Watch the figure below and you can see what was actually happening underneath a perfectly flat staffing line.

The machines aged on the same schedule

Now the nastier half. The people and the equipment got old together, and they got old in a way that made each one harder to replace.

Very few utilities tear out a working plant and build a new one. They patch. A treatment works commissioned in the 1970s does not stay a 1970s plant; it becomes a stack of decisions, each one made by different people for different reasons under different budgets.

• BuiltOriginal basins, pumps, piping, and the drawings that are still the official record.
• +10 yrsA capacity expansion, because the town grew in a direction nobody planned for.
• +20 yrsNew controls bolted onto original mechanical equipment.
• +30 yrsA pump replaced with the closest available model, not the same model.
• +40 yrsA SCADA retrofit, so the plant can now be watched from a screen.
• +45 yrsTreatment changes to meet a requirement that did not exist when it was built.
• +50 yrsCybersecurity hardening on top of all of the above.

An operator arriving today is not learning how water treatment works. That part is in books, and it is teachable in a classroom. They are learning the archaeology of one specific site: which layer a given behavior comes from, which drawing is still true, and which one has been quietly wrong since a modification in the 1990s.

The number everyone quotes, and the number nobody does

The famous figure is the one third. It comes from EPA’s own analysis, it is repeated in the 2024 report, and it sits alongside two others: a median age in the water workforce of about 48, and only about 10 percent of that workforce under the age of 24. Those three numbers are the whole demographic argument, and they are solid.

The number almost nobody puts next to them comes from the Bureau of Labor Statistics, and it is strange in a way that changes the shape of the problem.

Read those three together. The occupation is projected to shrink. And it still needs to hire on the order of ten thousand people a year, every year, for a decade, because of who is leaving. BLS is explicit that those openings come from workers transferring to other occupations or exiting the labor force, including retirement.

This is where most intuitions about labor shortages break. We are trained to think a hiring problem means a growing field that cannot keep up. This is the opposite: a contracting occupation with an enormous replacement bill. The two require completely different responses, and they get sorted differently by the systems that allocate training money. A workforce board scanning for “high-growth occupations” will look at a projected 7 percent decline and skip right past a field that needs to hire roughly its entire workforce over the course of a decade.

Federal programs have noticed the same trap. One of the four things EPA’s current funding round pays for is plain career exposure: telling people these jobs exist at all. That sounds almost too basic to fund, until you consider that a utility trying to replace a 58-year-old operator is competing for a 22-year-old against several hundred occupations they have actually heard of.

Automation subtracts operators, not expertise

Part of that projected decline is automation, and this is where the problem is most often misread. Automated valves, continuous sensors, and SCADA systems genuinely do mean a plant can run with fewer people in it.

But look at what happens to the work that is left. The old task was mechanical and legible: open valve A. The new task is diagnostic. Why does the controller believe valve A is open when the physical limit switch disagrees, and which of those two is lying?

Answering that requires chemistry, mechanics, electrical work, instrumentation, software, regulatory knowledge, and the specific history of this plant, held by the same person at the same time. Fewer operators, each carrying more.

Which means the two trends do not cancel. The headcount curve bends down while the per-person depth requirement bends up, and the gap between them is exactly the thing that takes years to build and cannot be bought in a hurry.

Why one hundred replacements is not one hundred replacements

Here is the crux, and it is the reason a headcount-based response can look successful while failing.

Suppose a hundred operators retire this year and a hundred newly certified operators are hired. On paper, resolved. In practice it depends entirely on one variable that no national statistic tracks: whether those two groups were ever in the building at the same time.

And overlap is not just about hours of instruction. Some of what an operator knows can only be taught by the calendar. You cannot explain what the basin does during spring turnover to somebody who has never watched it happen. You have to be standing there in the spring.

So is the response big enough?

This is the question the 2024 report cannot answer, and I do not think that is an oversight so much as an absence in the underlying data. The report is genuinely good at diagnosis. It lays out the challenges, organizes the response around four pillars, recruitment, training, retention, and partnerships, and fills the rest with examples of utilities, colleges, and communities already doing the work.

What I could not find anywhere in it is a target. No line saying the country needs X new certified operators a year, that it currently produces Y, and that it is therefore at Z percent of where it needs to be. Without that number, scaling a program nationally has no units. You can tell whether it got bigger. You cannot tell whether it got big enough.

The denominator, at least, we already have. BLS says roughly 10,700 operator openings a year. So the national system needs to produce and retain about that many qualified replacements annually, just to hold even.

You cannot get there by enrolling 10,700 people, because the pipeline leaks at every stage. Some do not finish. Some finish and take a job somewhere else. Some take the job and leave within a year. Turn those three dials and watch what the required intake does.

The money, and the honest way to compare it

It is tempting to put the federal program next to the problem and declare a verdict. The dedicated grant program has run roughly $3.8 million in 2021, more than $20 million in 2024, and $10.8 million offered in 2026. Call it something on the order of $35 million across three rounds, against an occupation that needs to hire around 107,000 people over a decade.

That comparison is genuinely unfair, and I want to be precise about why. Almost nobody becomes an operator through a federal grant. They get hired by a utility, trained by that utility or a community college, certified by a state agency, and paid out of a municipal budget. The federal program is not the pipeline. It is a small amount of money aimed at the places where the pipeline is missing entirely.

But notice what happened when we tried to make the comparison. We could not, because there is no national accounting of what the country spends on replacing these workers, and no national count of how many qualified entrants it produces. We can measure the intervention. We cannot measure the system it is intervening in.

And there is one leading indicator better than all of them, because it degrades years before anything breaks: qualified applicants per vacancy. A plant with eight good applicants per opening and a plant with one are in completely different states of health, and both of them are fully staffed. By the time the number reaches zero you are not measuring risk anymore, you are measuring an incident.

National averages hide where this actually hurts

One more reason the aggregate numbers mislead. A large metropolitan water department can lose five operators and still have dozens. A small town might run its entire system with three operators, a superintendent, and a maintenance technician. Two retirements there is not a staffing gap, it is an operational emergency, and the arithmetic of it never shows up in a national average.

This is also why the problem resists a centralized fix. The federal government can supply money and standards. But the actual transaction, finding a 22-year-old who has never considered this work, training them, paying them enough that the next town over does not take them, getting them certified, and putting them next to the 63-year-old who retires in three years, happens in a specific building in a specific town, thousands of times over.

Which is worth saying clearly, because it cuts against the reflex to grade this as a federal failure. The reports are good. The diagnosis is right. The programs do real things. What is missing is not will. It is a thermometer.

What a country keeps inside people

The original complaint was that nobody is preparing. The accurate version is narrower and harder: we prepared for a headcount problem, and what we have is a memory problem.

Those are not the same thing and they do not respond to the same medicine. You can solve a headcount problem with recruiting. A memory problem can only be solved by two people standing in the same room long enough for one of them to say, without being asked, that pump number three shakes when the river runs low.

A retirement is not the subtraction of one worker. It is the deletion of an index: the only thing in the building that knew which of fifty years of documents to trust.

What makes this more than a story about water is that the same structure is everywhere. An enormous quantity of what a modern country knows is not written down. It is not in the drawings, or the databases, or the regulations. It sits inside people who have been somewhere for a long time, and it is invisible to every instrument we use for counting labor, because our instruments count bodies and the scarce resource was never bodies.

The encouraging half of this is that the fix is unglamorous and completely known. Hire the successor early. Pay for the overlap. Make the retiring operator’s last two years a teaching assignment and treat that as the deliverable, not a courtesy. None of that requires a breakthrough. It requires deciding that the months two people spend in a building together are a capital asset, and putting them on a ledger somewhere.

The plant will not remember on its own. Concrete does not remember, and neither do drawings, and neither, it turns out, does a spreadsheet of full-time equivalents. Somebody has to, and right now we are not counting whether anybody still does.