How long does a UPS last?

The short answer, because this question is really two questions:

• Minutes on battery during a cut: typically 3–11 minutes at a realistic load. A 1500 VA unit — VA is the size printed on the box, and it works out to roughly 900 real watts — running a small rack is nearer ten than sixty.
• Years before the unit needs attention: the batteries last 3–5 years; the chassis often lasts 10–15.

If you came here for one of those and got the other, that’s not your fault. Search results for this question are split roughly down the middle between them and almost nobody says which they’re answering.

Question one: how long will it run when the power goes out?

Shorter than people expect, and the reason is worth understanding.

I pulled the 851 certified UPS models in the US EPA’s ENERGY STAR database, each of which publishes the manufacturer’s own declared runtime at two load points. For units rated 1500 VA or under — the size most homelabs and small offices run:

Load	Median declared runtime
100% of rated watts	3 minutes
50% of rated watts	11 minutes

Three minutes at full load. That is the honest number, from manufacturers’ own published figures, and it is nothing like the half-hour people tend to assume.

The bit that surprises everyone

Look at those two rows again. Halving the load didn’t double the runtime — it more than tripled it.

Across those 399 small-office units the median ratio is 3.57×, not the 2× a linear model predicts. That’s the Peukert effect: lead-acid batteries give up proportionally less of their capacity the harder you drain them. Drain gently and you get closer to the rated capacity; drain hard and a chunk of it is simply unavailable to you.

The practical consequence is the useful part: taking load off a UPS buys you more time than adding batteries does. Unplugging two monitors is free and can add more minutes than the next unit up the range would. If you remember one thing from this post, that’s the one.

I built a UPS sizing and runtime calculator that works this out for your actual kit — it uses each model’s own declared figures rather than a generic formula, and shows the arithmetic.

Question two: how many years before it dies?

Different question, different answer, and the honest version has two parts because a UPS is two things.

The batteries: 3–5 years. This is the part that fails, and it fails gradually rather than suddenly. A sealed lead-acid battery loses capacity from the day it’s made, and heat accelerates it — a UPS in a warm cupboard ages faster than one in open air. By year five a unit commonly holds around half the runtime it had new, which is why a five-year-old UPS that reports “100%” on its display can still drop your load in under a minute.

The unit itself: 10–15 years, often more. The electronics outlive several sets of batteries. Replacing batteries is usually far cheaper than replacing the UPS, and on most rack units it’s a drawer.

How to know rather than guess

• Run the self-test. Most UPSs have one, and many report the real figure to their management software. This is the only number that describes your unit.
• Watch the runtime estimate over time. A steady decline is the battery ageing normally. A sudden drop usually means a cell has failed.
• Note the install date somewhere you’ll find it. Batteries don’t announce themselves; the date does.

Why the two answers get tangled

It’s a genuinely ambiguous question, and it stays ambiguous because both answers are useful to different people on different days. Someone mid-outage wants minutes. Someone doing a budget wants years.

Worth knowing: the runtime number is also the one most often inflated. Vendor figures are quoted at convenient loads, the Peukert effect makes light-load numbers look far better than heavy-load ones, and a three-year-old battery quietly invalidates whatever the box said. A tool that gives you one confident number without telling you which load and which battery age it assumed is not being straight with you.

Where these numbers come from

The runtime figures are derived from the ENERGY STAR certified product database, published by the US EPA and in the public domain — dataset ifxy-2uty on data.energystar.gov, 851 models across 32 brands, each carrying the manufacturer’s declared runtime at 100% and 50% load. You can download it yourself.

The service-life figures are the well-established ranges for sealed lead-acid batteries — the type in nearly every UPS — and they’re a guide rather than a measurement — which is exactly why the self-test matters more than any number in this post.

⚠️ One caveat on the runtime table: it’s a certified list, so it skews toward efficient models, and declared runtimes are manufacturer figures rather than my measurements. They’re the best public numbers available, and they are still someone’s marketing department’s best case.