Walk into the basement of almost any hospital, school district office, or distribution centre, and you will find a rack of recorders humming in a room that nobody visits. It has been running continuously for six years. It will run continuously for four more until someone replaces it with a nearly identical box.
Multiply that by the tens of thousands of buildings across a corporate estate, a school district, or a municipal portfolio, and you have a category of energy consumption and electronic waste that rarely appears in a sustainability report. Security infrastructure is treated as a safety line item, not an environmental one. That framing is increasingly hard to justify.
The always-on baseline nobody counts
Physical security is unusual among building systems in that it is designed never to stop. Lighting dims at night. HVAC responds to occupancy. Recording infrastructure does neither. It captures and stores continuously, on the reasonable assumption that the footage you need is the footage you did not know you would need.
The energy cost of that assumption sits in two places. The first is the equipment itself: cameras, network switches, and on-premise recorders drawing power around the clock. The second, and larger, is the environment those recorders live in. On-premise server rooms are among the least efficient computing environments in existence. The Uptime Institute’s 2025 Global Data Center Survey put the weighted global average Power Usage Effectiveness at roughly 1.54. That means for every watt reaching the hardware, another half-watt is spent on cooling and power conversion. The figure has barely moved since 2018, and small server closets and edge installations typically perform worse still. Google, by comparison, reported a fleet-wide PUE of around 1.09.
That gap is not a rounding error. It is the difference between overhead consuming roughly nine per cent of your power and roughly fifty-four per cent of it.
The refresh cycle is the bigger number
Energy is the visible half of the problem. The material half is larger and less discussed.
Recording hardware has a service life. Storage arrays fail, processors fall behind the demands of newer camera resolutions, and, critically, proprietary systems force replacement on the vendor’s schedule rather than the building’s. When a manufacturer discontinues a product line or a software platform stops supporting older recorders, functioning hardware becomes waste for commercial reasons rather than technical ones.
The UN’s fourth Global E-waste Monitor, published by ITU and UNITAR, found that the world generated 62 million tonnes of electronic waste in 2022, with only 22.3 per cent formally collected and recycled. Annual generation is on track to reach 82 million tonnes by 2030. Security hardware is a small slice of that total, but it is a slice with an unusually forced replacement cadence and unusually little pressure toward reuse.
This is where the standard rip-and-replace modernisation project does real damage. An organisation modernising a hundred-site estate under a proprietary model does not just buy new software. It discards every camera, every recorder, and every reader that does not speak the new vendor’s protocol. The embodied carbon in that discarded equipment was already spent. Replacing it spends the same carbon again.
Moving the compute does not automatically move the problem
Here is where the sustainability case for cloud-native security has to be made carefully, because the easy version of it is wrong.
Shifting video processing from a server closet to a data centre does not eliminate emissions. It relocates them into facilities whose growth is now one of the most contested topics in energy policy. Global data centre electricity consumption reached an estimated 415 TWh in 2024, and the AI workloads driving new construction are precisely the workloads that modern security systems depend on. Google’s decision to commission small modular reactors for its data centres is not the behaviour of a sector with a shrinking footprint.
An honest accounting has to hold three things at once. Centralised compute runs at far better efficiency and far higher utilisation than distributed on-premise hardware, since enterprise server rooms commonly run at single-digit utilisation while purpose-built facilities run several times higher. That is a genuine gain. But AI inference across continuous video streams is computationally expensive, and it adds load that a passive recorder never carried. And the net result depends heavily on choices the buyer makes, not the architecture alone.
The environmental case for modernising security infrastructure is real. It is just conditional.
What actually moves the number
Four decisions do most of the work, and none of them are marketing claims.
Keep the hardware you have. The single largest environmental variable in a security modernisation is whether existing cameras stay in service. A platform that works with installed infrastructure avoids the embodied-carbon cost of a full hardware replacement entirely. A platform that requires proprietary cameras guarantees it. For estates with hundreds of functioning devices, this dwarfs every operational efficiency gain combined.
Consolidate systems rather than adding them. Most organisations run video, access control, visitor management, and emergency workflows as separate systems with separate servers, separate maintenance cycles, and separate hardware footprints. Replacing a standalone video management software deployment with one platform that also handles access and emergency response removes duplicated infrastructure, which is a straightforward reduction in both energy draw and eventual waste.
Interrogate retention policy. Storage is the quiet driver of cost and consumption. Retention periods are frequently set once, by default, and never revisited against actual legal or insurance requirements. Reducing over-retention is unglamorous and immediately effective.
Make the AI earn its load. If systems are running continuous analysis, that analysis should be replacing human review hours and reducing incident response times, not generating alerts nobody acts on. Compute spent on detections that change outcomes is defensible. Compute spent producing noise is not.
Why this belongs in the ESG conversation
Under emerging European sustainability reporting requirements, purchased electricity and value-chain emissions both have to be accounted for. Building systems that run continuously across a property portfolio are not a trivial contributor, and hardware refresh cycles sit squarely in the upstream emissions that companies are now expected to disclose.
Security teams have never been asked to defend their infrastructure on environmental grounds. Facilities and sustainability teams, meanwhile, have rarely had visibility into what the security estate consumes or discards. The two conversations happen in different rooms.
They should not. The next generation of physical security platforms is being evaluated on operational grounds: faster detection, faster investigations, coordinated response across sites. Those same architectures happen to change the energy and materials profile of the estate substantially. Not automatically, and not by default, but meaningfully, when the buyer asks the right questions.
The most useful of those questions is also the simplest. What happens to the equipment we already own?
Editor’s Note: The opinions expressed here by the authors are their own, not those of impakter.com —
