According to the report on data center interruptions published by the Uptime Institute in May 2026, power remains the most common cause of impactful outages: failures in backup power units, transfer switches, and generators outrank every other technical cause. More than half of the companies surveyed — 57% — said their most recent major outage cost above US$ 100,000, and for the second year running, one in five reported losses above US$ 1 million.
This is not a problem exclusive to giant data centers. It's an electrical installation problem — and every company with a server, a router, or a point-of-sale system has one. The difference is scale, not nature: the same backup power unit — the equipment that keeps power running for a few minutes when the lights go out — that gets scheduled maintenance in a large operation is, in a small business, usually bought once, plugged in under a desk, and forgotten until it starts beeping.
For whoever decides the IT budget, the question that matters isn't "does the backup unit exist" — almost every company has one. It's "does anyone know, today, whether it still holds the load plugged into it." In most cases, the answer is no.
The Battery Nobody Replaced
A backup power unit doesn't last forever because the battery inside it doesn't last. According to support documentation from Schneider Electric — the manufacturer behind the APC brand — the battery in a Smart-UPS unit needs replacing every three to five years, even if the equipment keeps turning on and appears to work normally.
The problem is that between one power cut and the next, the unit spends most of its time silent. The panel light stays green, the equipment doesn't complain, and there's no natural warning — no screen, no email — saying the battery has lost the capacity to hold a load. The company only finds out at the exact moment the information stops being useful: in the middle of the outage.
It's the same pattern that shows up in the Uptime Institute numbers: power failure remains the most cited cause of severely impactful outages, and the reason is rarely the absence of a backup unit — it's a backup unit that's present but without a battery capable of holding the time the operation needs.
Capacity is another thing nobody recalculates after the initial purchase. The unit comes sized for whatever was plugged into it on installation day. Over time, the company adds another server, another switch, another monitor to the same power strip — and nobody revisits the math. The result is a unit that would hold five minutes "on paper" and shuts off in ninety seconds in practice, because it's now carrying twice the original load.
There's also the environment around the equipment: backup unit and server crammed into a room with no ventilation, near a window with direct sun, or stacked next to cleaning supplies. Heat shortens battery life faster than usage does — a silent factor that stays off the list of suspects until the day the battery fails well before its expected deadline.
Why Buying a Backup Unit Isn't Enough
The common way of handling power is to solve it once: a backup unit is bought the day the server arrives, it gets forgotten in a corner, and the company moves on until the next outage — which is exactly the moment when it's discovered whether that years-old decision still holds up.
Another version of the same habit is blindly trusting the grounding and electrical wiring that were already there when the company moved into its current address. Old wiring, an outlet without proper grounding, or an overloaded electrical panel don't show up on any screen — they only show up when they burn out a piece of equipment.
Shutdown is usually the other blind spot. When the battery runs out, the server doesn't get a "shut down safely" command: it simply loses power, the same as if someone had pulled the plug. A database mid-write, a shared spreadsheet being saved, a point-of-sale system closing a sale — all of it can get corrupted when the cut is abrupt, even if the hardware itself takes no physical damage.
The side effect of this habit is a false sense of security: the company sees the backup unit's green light, hears the beep when power drops, and assumes it's covered — without knowing how much time that equipment actually has left, or whether the server's shutdown will be clean or abrupt.
What Has to Be in Place
An inventory of everything that depends on continuous power. Server, router, phone system, point of sale, cameras — every critical piece of equipment needs to be listed, with its installation date and power draw, so the capacity math is done with real data, not memory.
Periodic load testing of the backup unit, deliberately simulating a power cut at a controlled time, to confirm how long the equipment actually holds — instead of finding out during the first real outage of the year.
Battery replacement scheduled by time, not by symptom. Replacing it before the manufacturer's deadline costs a fraction of what it costs to rebuild a corrupted database or recover a lost day of billing.
Automatic, orderly server shutdown when the battery nears its end, instead of letting the equipment go dark on its own in the middle of an operation.
Grounding and electrical wiring reviewed by someone who understands the subject, especially in older buildings or after any renovation.
A suitable environment for the equipment: a ventilated room, away from direct heat, with no flammable material nearby — the battery lasts longer when the environment helps instead of working against it.
This is how Skills IT works: with an inventory of what depends on power, periodic testing of the backup unit, and battery replacement scheduled by deadline, before the equipment decides the date on its own.
What Changes When Power Fails and the Business Keeps Running
The payoff of handling this beforehand doesn't show up day to day — it shows up exactly on the day the lights go out. A company that has already tested its backup unit knows, without surprise, how many minutes it has to act. One that has already set up automatic shutdown doesn't lose its database mid-write. The difference between the two scenarios isn't luck: it's what got reviewed beforehand.
Financially, the math is simple: replacing a battery costs a fraction of what it costs to have a technician rebuild a corrupted database, or to lose a day of sales because the point-of-sale system won't come back. The Uptime Institute numbers show the same proportion at a larger scale — more than half of severe outages cost six figures or more, in a problem that, at its root, is the same one: power that should have been covered and wasn't.
There's also a gain that doesn't show up on a spreadsheet: the team stops working in the dark, literally. Without knowing how long the backup unit lasts, every power cut turns into a race against the clock, with everyone trying to save what's open before the equipment goes dark on its own. With calculated capacity and automatic shutdown, a power outage becomes an annoyance, not an emergency.
A Checklist to Avoid Finding Out Too Late
- When was the backup unit's battery last replaced? If nobody can answer, it has probably already passed its deadline.
- Has anyone ever tested the backup unit by cutting power on purpose? Without that test, its real capacity is a guess.
- Is what's plugged into the backup unit today the same as on the day it was bought? If the answer is "more stuff," the capacity math changed and nobody redid it.
- Does the server shut down on its own when the battery runs out, or does it go dark as if unplugged? That difference decides whether the database survives the cut.
- Does the room where the backup unit sits have ventilation, or does the equipment share space with heat and direct sun? The environment also decides how long the battery lasts.
