The outage is the small half of the bill. The large half is the replacement lead time — and in 2026 that runs from 60 weeks for medium-voltage switchgear to 150 weeks or more for a large step-up transformer.
Two numbers decide it. The first is what an hour of downtime costs you — for most large operators that is six figures an hour, and for one operator in five the last major outage cost more than a million dollars. The second is how long you wait for the replacement part. In 2026 that second number has grown far faster than the first. A failed unit you cannot replace for two to four years is not a maintenance event. It is a capacity event, and it should be priced like one.
These are the published figures. They are worth knowing precisely, because they are the half most operators already have a number for.
Share of operators whose most recent major outage cost more than $100,000. One in five said it exceeded $1 million — the second consecutive year at that level. Power remains the leading cause, with UPS systems, transfer switches and generators the dominant failure points.
Share of enterprises with 1,000+ employees reporting average hourly downtime cost above $100,000. 41% put it between $1 million and $5 million per hour. The figures exclude litigation and regulatory penalties.
Share of unplanned data center downtime incidents attributed to batteries, with UPS system failure averaging roughly $710,000 per incident — the third most expensive root cause measured.
This is the number that changed. Every figure below is a 2026 quoted lead time, not a forecast.
Generator step-up transformer, 50–150 MVA — 100 to 150+ weeks
2–3 yearsSubstation transformer, 25–50 MVA — 85 to 110 weeks
~2 years38 kV switchgear — 78 to 104 weeks
~2 yearsDiesel generators, 3,500–4,500 kW — 90 to 110 weeks
~2 yearsModular UPS, 2,000+ kVA — 40 to 60 weeks
~1 yearLead times compiled by Terrapin Construction Group, June 2026. At the tier-1 OEM level the picture is longer still: PwC analysis reported by pv magazine USA in May 2026 put high-capacity transformer lead times at four years, with component prices up roughly 80% over five years. Bloomberg reported in April 2026 that transformer waits now stretch to as long as five years against 24 to 30 months before 2020.
Run-to-failure was a defensible strategy when the replacement arrived in twelve weeks.
When the replacement arrives in three years, the same failure stops being a repair cost and becomes a stranded-capacity cost for as long as the queue lasts. A megawatt you cannot energise earns nothing, and it earns nothing for every week of that lead time — while the debt service, the land, and in many contracts the take-or-pay obligation continue.
That is the whole case for prediction, and it is an arithmetic case rather than a technology one. The value of knowing eighteen months early is not that the repair is cheaper. It is that eighteen months is long enough to get in the queue before the asset fails, so the replacement lands near the failure instead of years after it.
Scheduled beats unplanned by the length of the lead time. In 2026 that is the largest single term in the equation, and it is the one most maintenance budgets still do not carry.
A read-only assessment runs on data you already have, before any hardware conversation. You see the record; you decide what it is worth.
Choir™ Corp