Doberman

How long do CCTV cameras lastfour parts, four clocks, and no published answer

Everyone quotes five to ten years and nobody sources it. Manufacturers publish a warranty, not a service life, and the camera body is rarely what runs out first. Here is what the figures actually say.

The quick answer

Nobody publishes a service life for a CCTV camera, and the confident numbers you will find elsewhere are not sourced to anything. What manufacturers do publish is a warranty, commonly five years on a professional range. The more useful framing is that a system is four things ageing on four different clocks, and the camera body is rarely the one that runs out first. Most cameras stop being good enough long before they stop working, which makes replacement a judgement about what you can see rather than about whether anything has broken.

What this guide covers

  • What manufacturers publish about lifespan, and the figure they pointedly do not.
  • Why a million-hour MTBF is not a lifespan, and what number to use instead.
  • Why the night picture goes first, and the reason it is so easy to miss.
  • Why the hard drive is measured in terabytes a year rather than in years.
  • The replace-or-repair test, taken from government guidance rather than opinion.

Four parts, four different clocks

Asking how long CCTV lasts is really four questions. Only some of them have published answers, and the last column is the one worth reading.

The camera itself

What it is measured in
Years, but rarely the first thing to go
What is actually published
A warranty. One major manufacturer gives 5 years. Service life: not published

Infrared illumination

What it is measured in
Fades gradually from day one
What is actually published
LED output convention is L70, the hours until 70% of initial output remains

The recorder hard drive

What it is measured in
Terabytes written per year, not years
What is actually published
Workload rating. Surveillance drive 180 TB/yr and 24/7; desktop drive 55 TB/yr and 2,400 hours

Seals, glands and connections

What it is measured in
Set by the install, not the product
What is actually published
An IP rating for the housing. Nothing for the joint you made on site

By the Doberman install team

CCTV system designers & installers, Leicester

Last reviewed September 2026

Why nobody will give you a number

Search this question and you will be told five to ten years, repeatedly and confidently, with nothing behind it. We are not going to add an eleventh version of that, because the interesting thing is why the number does not exist.

Manufacturers publish two things about durability. One is a warranty: a major professional camera manufacturer covers most of its range for five years against defects in design, workmanship and material under normal use. The other, on some products, is a mean time between failures. What you will struggle to find anywhere is a stated expected service life, and that is not an oversight. Service life depends on where the thing is mounted, how hot it runs, how wet it gets and how well the install was done, none of which the manufacturer controls.

So the warranty is the useful signal. It is the period a manufacturer will put its own money behind, which makes it a floor rather than an estimate. Read a five-year warranty as confidence that the hardware should comfortably clear five years, not as a prediction that it fails in year six.

A million hours is not a lifespan

This is the number that gets quoted at people during a sale, and it is almost always misunderstood. A surveillance hard drive advertised with a mean time between failures of one million hours sounds indestructible.

Do the arithmetic. A million hours is about 114 years. Nobody is claiming your drive will see out the next century, and the manufacturer is not trying to. MTBF is a statement about the failure rate of a large population of units during their useful life, not a prediction about one unit on your wall. Treating it as a lifespan is a category error.

The number that actually tells you something is an annualised failure rate, measured across a real fleet. Backblaze runs hundreds of thousands of drives and publishes its own statistics: in the first quarter of 2026 it reported a 1.24 per cent annualised failure rate across 341,263 drives. That is a figure you can plan around. Roughly one drive in eighty, per year, will fail, and if your recorder has one drive in it, that is your exposure. Worth knowing where that number comes from, though: those are enterprise drives in a data centre, and no surveillance-rated models appear in the fleet. Take it as the right order of magnitude rather than as a figure for the drive in your NVR.

It is also the argument for noticing before it matters. A drive that dies quietly takes your retention with it, and nothing on the app tells you.

Not sure how much life is left in your system?

Doberman can assess an existing installation and tell you what is worth keeping, what is worth replacing, and what is only worth replacing when it finally goes.

The night picture goes first

If one thing on a camera ages visibly, it is the infrared illumination, and it ages in a way that is almost designed to be missed.

LEDs do not burn out the way a bulb does. They fade. Output drops gradually over thousands of hours, which is why the lighting industry does not quote a lifetime in the ordinary sense at all: it quotes the number of operating hours until light output has fallen to a stated fraction of where it started. The common convention is L70, the point at which 70 per cent of the original output remains. The standards behind it measure lumen maintenance over a test period and then project it forwards.

Apply that to a camera on the front of a house. The infrared array runs every night for years. It loses output steadily, so the illuminated range creeps inwards, a little each season. Nothing fails. No alert appears. The daytime image is unchanged, so a quick look at the app reassures you. Then one night somebody walks up the drive and the face that would have been lit three years ago is a grey shape.

Two practical consequences. Judge night performance deliberately, at night, against how it used to look rather than against how it looks now. And when a camera is described as having a given infrared range, treat that as a figure for a new camera, because it will not be the figure in year five.

The drive is not measured in years

The recorder drive ages on a completely different clock from everything else, because what wears it is not time but writing.

Surveillance-rated drives are specified accordingly. One is rated for workloads up to 180 TB a year, handling simultaneous streams from up to 64 cameras, and is sold explicitly for systems running 24 hours a day. That annual workload figure, rather than any age, is the specification that matters, because a CCTV recorder writes continuously in a way a computer never does.

The contrast with an ordinary computer drive is stark once you look at the same specification row on both. One manufacturer rates its standard desktop drive at 2,400 power-on hours per year, a 55 TB annual workload and a two-year warranty. Its surveillance drive, from the same catalogue, is rated for 180 TB a year and sold for round-the-clock operation. Two thousand four hundred hours is about six and a half hours a day. A recorder asks for 8,760.

So a desktop drive in an NVR is not merely a cheaper choice, it is being run at more than three times the duty it was specified for, writing three times the data. That is the most common reason a system quietly stops holding as much as it used to, and the question worth asking about any recorder is simply what its drive is rated to write per year and whether it is specified for continuous operation at all.

We have written separately about how a failing drive actually behaves and what a service visit checks, in CCTV maintenance and servicing, and about what sets your retention in the first place in how long CCTV footage lasts.

The part with no published lifespan at all

A camera housing carries an IP rating. The cable joint somebody made behind it on a wet Tuesday carries nothing at all, and that joint is where outdoor systems usually fail.

It is worth being clear about why this belongs in a discussion of lifespan. A camera rated for a decade of weather, mounted well, will outlast a poorly sealed connection by years. The lifespan of the installation is not the lifespan of the products in it. Nobody publishes a figure for workmanship.

We have covered what IP ratings actually guarantee, why the joint rather than the camera lets water in, and how condensation gets into a housing, in weatherproofing outdoor CCTV, so we will not repeat it here.

Replace when you cannot see, not when it breaks

Here is the reframe that makes this question answerable. Cameras very rarely die. They become inadequate, which is a different event and one nobody schedules.

Government guidance on police requirements for CCTV supplies the test, and it is refreshingly short: if you cannot see it, it is not fit for purpose. The same guidance adds the detail that catches people out, which is to assess the system by viewing the recorded video rather than the live screen, because the quality of the two can differ. A live feed on a phone can look perfectly sharp while the recording it is making is not good enough to identify anyone.

So the annual question is not how old the cameras are. It is three checks:

  • Pull up recorded footage, not the live view, from the cameras that matter most, at the time of day that matters most.
  • Ask whether you could identify a stranger from it, at the distance you actually need, rather than recognise someone you already know.
  • Check the night version separately, because that is where the degradation described above shows up first and where most incidents happen.

If it passes, the age of the hardware is irrelevant and you should leave it alone. If it fails, it fails today, whether the kit is two years old or nine, and no warranty figure changes that.

About this guide

Who wrote this

Written by Doberman, a CCTV system designer and installer working with homes and small businesses across Leicester and Leicestershire. Specifying a system means reading the durability figures manufacturers publish, and noticing which ones they leave out.

How this guide was produced

The warranty position comes from Axis Communications' published warranty terms. The drive figures come from Seagate's SkyHawk product specification and the measured failure rate from Backblaze's published drive statistics for the first quarter of 2026. The LED degradation convention is from the US Department of Energy's guidance on LED luminaire lifetime. The replacement test is from GOV.UK police requirements for CCTV systems. Naming those companies is a citation of published figures, not a brand recommendation.

What we have deliberately not said

We have not given a service life in years, because no manufacturer publishes one and every figure in circulation traces back to other articles rather than to a source. We have not attached an L70 hours figure to camera infrared arrays, because we could not find one. We have also kept the failure-rate figure in its proper context: it is measured on enterprise drives in a data centre, which is not the same population as the drive in a domestic recorder.

Disclosure

We sell and install CCTV, so we benefit when someone decides to replace a system. That is exactly why this page argues for testing the recording before spending anything, and says plainly that a system passing the test should be left alone regardless of its age.

Common questions

How long does a CCTV camera last?

Nobody publishes a figure, and that is the honest answer. Manufacturers publish a warranty, and sometimes a mean time between failures, but almost none states an expected service life. One major camera manufacturer warrants most of its range for five years against defects in design, workmanship and material. Five years is therefore what a manufacturer will stand behind, not the point at which a camera stops working. In practice cameras usually stop being good enough before they stop working at all.

What does a 1 million hour MTBF actually mean?

Not what it looks like. A million hours is about 114 years, which nobody is claiming. MTBF describes the failure rate across a large population of units during their useful life, not how long one unit survives. The useful real-world version is an annualised failure rate. Backblaze, which publishes drive statistics from its own fleet, reported a 1.24 per cent annualised failure rate across 341,263 drives in the first quarter of 2026.

Why does my CCTV look worse at night than it used to?

Because the infrared illumination is probably fading, and fading is how LEDs age. They do not burn out like a bulb, they gradually put out less light, which is why LED lifetime is conventionally quoted as the hours until output has fallen to 70 per cent of where it started. Nothing appears broken, the daytime picture still looks fine, and the night range quietly shrinks year on year until somebody needs a face at the end of a driveway and it is not there.

Does the hard drive last as long as the cameras?

No, and it is not even measured the same way. A drive is rated by how much you write to it per year rather than by age. One manufacturer rates its surveillance drive for 180 TB a year, 64 HD streams and continuous 24/7 running. Its ordinary desktop drive, same catalogue, is rated for 2,400 power-on hours a year, a 55 TB workload and a two-year warranty. That is roughly six and a half hours a day, and a recorder needs 8,760 hours. Put a desktop drive in an NVR and it runs at more than three times the duty it was specified for.

When should I replace rather than repair?

Judge it on what the recording shows, not on whether the camera is working. Government guidance on police requirements puts the test simply: if you cannot see it, it is not fit for purpose. The same guidance says to assess performance by viewing the recorded video rather than the live screen, because the two can differ. A camera producing a perfectly stable picture in which nobody is identifiable has failed at its job, whatever its age.

Is it worth replacing cameras one at a time?

Often yes, if the recorder and cabling are sound and the cameras use a common standard. The awkward case is a mixed system, where you end up with one sharp camera and five that make it obvious how dated the rest are. The thing that usually forces a whole-system decision is not the cameras at all, it is the recorder or the cabling reaching its limit.

Want to know what your system has left in it?

We visit, map the blind spots, and quote one fixed price on the spot.