■ HIGH RISK ■ Construction & Trades
Not soon — but the job is being hollowed out from both ends. Diagnostics are going algorithmic and utilities increasingly swap failed units instead of repairing them, leaving fewer, more specialized hands-on roles.
“Smart grid monitoring predicts failures. But rewinding coils is still manual.”
Our AI replacement risk score — how we score jobs
Transformer repairers keep the grid's most expensive iron alive: testing insulation resistance, sampling and analyzing oil, replacing bushings and tap changers, and — the heart of the craft — untanking a unit to rewind burned windings by hand. It's a mix of electrical testing, heavy rigging, and coil work that looks more like industrial weaving than electronics. A large power transformer can cost millions and take a year to replace, which is exactly why someone gets paid to fix them.
The diagnostic half is rapidly becoming software. Dissolved gas analysis used to mean pulling an oil sample and mailing it to a lab; now online DGA monitors stream data continuously, and predictive models flag a failing unit weeks before a human would have caught it. Condition-based maintenance platforms schedule the work, prioritize the fleet, and increasingly tell the repairer what's wrong before the tank is open. Meanwhile, at the distribution level, economics have shifted toward swap-and-scrap: a failed pole-top transformer gets replaced, not repaired, and the old one goes to a regional rebuild shop or the recycler. That consolidation is what our 65 score is really measuring — fewer shops, fewer generalist repair jobs.
The physical work resists automation stubbornly. Rewinding a coil, wrestling a bushing, brazing leads inside a tank — nobody has built a robot that does this economically, and the production volumes are too low and too varied to justify one. Add an aging workforce, a global transformer shortage, and grid expansion driven by electrification, and the specialists who remain are in genuine demand. The trade shrinks in headcount but not in value.
Automatability: our editorial assessment of current and near-term AI capability
Serious pressure lands by around 2030, but unevenly. Online monitoring and predictive analytics are already standard on transmission fleets, shrinking routine test-and-inspect work this decade. Distribution-level repair keeps consolidating into fewer rebuild shops. The countervailing force is the transformer supply crunch — with new units on multi-year backorder, utilities are repairing things they'd rather replace, which props up demand for skilled rewinders through the decade even as diagnostic roles thin out.
Shrinking, not dying. Routine diagnostics are going to sensors and software, and small units get swapped rather than fixed. But large power transformers are scarce, expensive, and backordered for years, so skilled rebuilders are in demand — arguably more than a decade ago. The trade is consolidating into fewer, more specialized roles rather than disappearing.
New transformer manufacturing uses winding machines, but repair is a different animal — every failed unit is a one-off with unknown damage, and untanking, cutting out, and rebuilding windings requires adaptive handwork in awkward spaces. The volumes are too low and too varied to justify robotic tooling. This part of the job is safe well past 2040.
Mostly by moving the diagnosis upstream. Online gas monitors and predictive analytics now flag failing units and often name the fault type before anyone opens a tank. That kills some troubleshooting mystique but means fewer catastrophic surprises and better-planned outages. Repairers who can read and validate the monitoring data become more valuable, not less.
Two directions pay off: go deep on hands-on rebuild skills that few people still have, especially large power transformers; or go up the stack into condition assessment, test engineering, and fleet management. Either beats staying a generalist. Also watch the renewables build-out — every solar and wind interconnection needs transformers, and someone to keep them alive.