AI doesn't only consume electricity. Eventually all those computers have to go somewhere
Keep up with IT Club
Add IT Club as a preferred source in Google Search.
Basel Action Network estimates that AI-related data-centre equipment and possible early device upgrades could produce 395–617 million metric tonnes of retirements between 2025 and 2050. The estimate depends on long-range assumptions and does not say how much equipment will be reused, recycled or disposed of. For small businesses, the practical question is simpler: replace devices when there is a real need, and give useful equipment a secure second life.
Most conversations about AI's environmental impact focus on electricity, water and the data centres that need both. There is another, less glamorous question: what happens to all the physical equipment when it is replaced?
A September 2026 report from the Basel Action Network (BAN) estimates that AI-attributable equipment retirements could total 395–617 million metric tonnes between 2025 and 2050. That is a striking range. It is also a model, not a count of equipment already thrown away or a prediction of how much will end up in landfill.
The headline number is a scenario
BAN combines two kinds of equipment. The first is the infrastructure built for data centres: servers and accelerators, networking, power distribution, backup systems and cooling. The second is what the report calls an “AI waste contagion”: the possibility that the AI boom prompts earlier replacement of equipment outside data centres, including PCs and phones.
That second part is not a tally of businesses already replacing working laptops. It is a scenario about what might happen if technical requirements, product design or enthusiasm for local AI encourage people to upgrade sooner than they otherwise would. The report's total therefore depends on how quickly data-centre capacity grows, how long each category of equipment stays in service and how much early replacement is attributed to AI.
For example, BAN extends a projected growth rate for data-centre capacity over a long period and uses different assumed replacement lives for servers, network equipment and power systems. Those are modelling choices, not a universal timetable that every operator follows. Change the assumptions and the cumulative result changes too.
Retired equipment is not automatically landfill
The report's 395–617 million tonnes describes estimated equipment retirements. It does not calculate how much will be repaired, sold for reuse, refurbished, recycled, exported, stored or disposed of. Retired from one data centre does not necessarily mean useless everywhere else.
BAN also translates its modelled mass into an equivalent 15–23 million 40-foot shipping containers, which it says would stretch around the Earth about six times. That is an illustrative mass-to-container comparison. It is not a measured volume of waste, a known pile of containers or evidence that all the equipment will be dumped.
The wider e-waste picture is already worth taking seriously without turning a scenario into a certainty. The UN's Global E-waste Monitor reported 62 million tonnes of electronic waste generated worldwide in 2022, of which 22.3% was documented as properly collected and recycled. That global figure covers all kinds of e-waste; it is not an AI-specific recycling rate.
What this means for a small business
Most SMEs will not be buying racks of AI accelerators. But they do replace laptops, desktops, phones and network equipment, and the AI sales pitch may add a new reason to do it. Before replacing a perfectly usable computer because it lacks a fashionable “AI PC” feature, check what you actually need to run. Many hosted AI services do their heavy processing in the provider's cloud; a local AI workload may have different hardware requirements.
- Keep a useful, supported computer in service when it still meets the business's security and performance needs.
- Check whether a repair, memory or storage upgrade would solve the real problem before replacing the whole device.
- Reassign working equipment to a role with lighter demands, or use a reputable refurbishment route when it no longer suits its first user.
- Before equipment leaves your control, remove business data and accounts using a process appropriate to the device and the information it held.
- When equipment really has reached end of life, use an appropriate reuse or WEEE route for your location rather than putting it into ordinary waste.
The Environment Agency's guidance explains when electrical equipment becomes waste and what to consider when equipment is intended for reuse. It applies to England; separate guidance applies elsewhere in the UK.
Track useful life before it becomes waste
A small business does not need a complex sustainability programme to make better refresh decisions. Keep a basic record of what each device is used for, whether it still receives security updates, its repair history and who has it. When it is replaced, decide before collection whether it can be reassigned, refurbished or recycled. Make data removal part of the handover, so a useful second life does not also inherit the previous user's files or access.
There is a cost argument too. A machine that is too unreliable for important work may be a false economy, while buying a replacement that sits unused is wasteful in a very literal sense. Extend a device's life when it remains secure and suitable; replace it when there is a clear operational or security reason, and record that reason.
If you are weighing a new laptop against a refurbished one, these IT Club guides cover value and the practical checks to make on a used Windows computer.
Do I Really Need to Spend £1,000+ on a Business Laptop? →
Buying a Used Windows Laptop? Check the Battery and SSD Health First →
A longer life is often the sensible business choice
Keeping a suitable machine useful for another year can save money and avoid creating demand for a replacement. That does not mean hanging on to devices that are insecure, unreliable or slowing important work. It means making replacement decisions for a real business reason, then planning what happens to the old equipment.
BAN's scenarios are a reason to ask better questions about the hardware behind AI, not proof that a particular amount of future equipment will become landfill. For an SME, the immediate opportunity is more ordinary: buy carefully, repair where it makes sense, reuse securely and dispose of equipment responsibly when its useful life is genuinely over.
Sources and further reading
Basel Action Network: The Coming AI Waste Wave, Part I (September 2026) →
UNITAR: Global E-waste Monitor 2024 →
Environment Agency: When electrical equipment becomes waste (England) →
Found this useful? Forward it to someone who might too.
Plain-English Takeaway
The AI e-waste total is a modelled scenario, not known landfill; small businesses can still control how long their own equipment remains useful and where it goes next.
Related Articles
Nuclear fuel from seawater sounds impossible. It isn't — but there's a catch.
The ocean contains an extraordinary amount of dissolved uranium. Turning that theoretical resource into practical nuclear fuel is a much harder engineering, economic and environmental question.
Read articleCan an App Really Tell You What Is Hidden in Your Clothes?
Clothing-transparency apps can analyse product links, screenshots, label photographs and product descriptions to explain fabric composition and flag possible concerns involving synthetic fibres, microplastics or PFAS. They do not chemically test garments and cannot identify treatments that manufacturers have not disclosed. This article explains how these apps work, where their estimates have limits and what consumers and businesses should consider before acting on an app score.
Read articleCan We Really Generate Electricity from Earth's Rotation?
A research team has reported producing a tiny continuous electrical voltage using a specially designed magnetic shell interacting with Earth's own magnetic field. The researchers argue the energy comes from Earth's rotation. The measured output was only tens of microvolts, the interpretation remains disputed and no practical generator has been demonstrated. This is a physics experiment — not a new electricity tariff.
Read article