TL;DR
A device carbon footprint is not one number. It is a number plus a boundary, and the boundary does most of the work. Published figures put manufacturing anywhere from a quarter to almost all of a laptop’s lifetime emissions. Both camps can be right, because they are counting different things. Once you know which boundary your figure uses, the reporting gets easier and the strategy gets obvious.
- A footprint figure is meaningless without its boundary. Ask what the study counted before you quote it.
- HP reports the EliteBook 830 G10 at 172.00 kg CO2e cradle to grave, with manufacturing at 153.10 kg.
- On that split, manufacturing is 88.14 per cent and the use phase is 8.92 per cent.
- If manufacturing dominates, buying fewer machines beats running them efficiently.
- The single biggest lever you control is how long each device stays in service.
- Vendors publish their own numbers. Treat them as self-reported and say so in your report.
What a device carbon footprint actually measures
A device carbon footprint is an estimate of the greenhouse gas emissions attached to one machine across its life. It is expressed in kilograms of CO2 equivalent. The method comes from life cycle assessment, and ISO 14040, the life cycle assessment standard is the standard most vendors name when they publish.
The life is usually cut into four stages. Manufacturing covers the materials and the factory. Distribution covers getting the machine to you. Use covers the electricity it draws. End of life covers what happens when it is finished.
Best tools for Device Management
So far so simple. The trouble starts with where each boundary sits.
The boundary is the whole argument
Two assessments of the same laptop can report very different manufacturing shares. One folds raw material extraction into manufacturing. Another reports extraction as its own line. The machine has not changed. The headline percentage has moved by tens of points.
Use-phase assumptions move the number just as much. An assessment that assumes four years of daily use in a country with coal-heavy electricity will show a much larger use share than one assuming three years on a clean grid. Neither is wrong. They answer different questions.
This is why two suppliers can hand you contradictory figures and both be telling the truth. Before you quote any number, find out what it counted and over how long.
What the published numbers actually say
Manufacturers publish per-model data, and it is the most specific evidence available to most teams. HP reports the EliteBook 830 13 inch G10 at 172.00 kg CO2e cradle to grave. The breakdown is published alongside it.
| Stage | kg CO2e | Share | What sits in it |
|---|---|---|---|
| Manufacturing | 153.10 | 88.14 per cent | Materials, components, assembly |
| Use | 15.50 | 8.92 per cent | Electricity drawn over the assumed life |
| Distribution | 3.40 | 1.96 per cent | Moving the finished machine |
| End of life | 1.70 | 0.98 per cent | Collection, treatment, recovery |
Those figures come from HP’s self-reported footprint for the EliteBook 830 G10, which records them as self-reported by the manufacturer for 2025. That matters. They are not independently audited, and your report should say so.
Across the wider literature the manufacturing share is far less settled. Studies land anywhere from roughly a quarter to almost all of the total. The low end comes from work that separates raw material extraction out. The high end comes from work that folds it in and assumes a short, efficient use phase.
Apple takes a different reporting approach again and publishes per-model reports that lead with change rather than absolutes. Its January 2023 report for the 16 inch MacBook Pro states an overall 15 per cent decrease in carbon footprint compared with the previous generation. You can read the method behind that in Apple’s environment programme and its product environmental reports.
What to do with a range this wide
Pick one source and stay with it. A report that mixes a vendor figure for one model with an academic figure for another is not comparable, and a reader who checks will find that out. Consistency beats precision here.
Why the manufacturing share decides your strategy
Here is the part that changes decisions. If most of the carbon is spent before the machine arrives, then efficiency measures during use are working on a small slice.
Take the HP figures above. The entire use phase is 15.50 kg. Suppose you ran an energy campaign and cut that by a fifth. You would save about 3 kg per device. Replacing that laptop a year earlier than necessary costs 153.10 kg.
One decision is fifty times the size of the other. The arithmetic is not close.
So the lever is not power settings. It is how many machines you buy, and how long each one stays in service. Everything else is rounding.
The four things that actually move the number
- Service life. Every extra year spreads the manufacturing cost over more work. This is the biggest single lever and it is entirely yours.
- Redeployment rate. A returned machine that goes to the next joiner avoids a new manufacture. A returned machine in a cupboard avoids nothing.
- Specification discipline. A heavier machine carries more embodied carbon. Buying to the job rather than to the job title reduces it.
- Recovery rate. Devices you never get back cannot be redeployed. A leaver who keeps a laptop has forced a purchase.
Three of those four are logistics problems, not sustainability problems. That is an awkward finding for a lot of reporting, because the levers sit with whoever runs equipment rather than with whoever writes the report.
What you can actually change this year
Disclosure: RemoAsset, named below, is owned by the same people who publish HROpsLab. It is described with its limitations stated, on the same terms as every other supplier. This site takes no vendor money and sells no placements.
Start by measuring the thing you can move. Service life and redeployment rate are both countable from your own records, and neither needs a consultant.
Measure average service life honestly
Take every machine retired in the last twelve months. Record the months between issue and retirement. Take the median, not the mean, because a handful of very old machines will flatter the average.
Most teams find the median is shorter than their stated refresh policy. Machines leave early because they were damaged, because a leaver kept one, or because nobody checked whether a returned device was still good.
Measure the redeployment rate
Count the devices that came back last year. Count how many went out again to another person. Divide. That percentage is the single most useful sustainability number an equipment team has, and almost nobody tracks it.
If the rate is low, find out why before buying anything. The usual causes are storage with no condition check, no wipe process that anybody trusts, and no record of what is in the cupboard.
Close the recovery gap
A device that never comes back is a forced purchase and a forced manufacture. If recovery is the weak point, that is where the carbon work is, however little it sounds like sustainability. A laptop retrieval service for remote employees exists for exactly this gap, and the reason it shows up in an emissions conversation is that an uncollected machine cannot be reissued to anyone.
Wiping sits in the same place. Teams that cannot certify a wipe will not redeploy a returned device, so it sits idle while a new one is bought. Certified data wiping and disposal for returned devices removes that blocker, though you should ask any supplier which standard they wipe to and get it in writing.
Where the number gets misused
Four patterns come up often enough to be worth naming.
Quoting a figure without its boundary. A report that says a laptop is 172 kg without saying cradle to grave, which model, and whose data is not checkable. Anybody who tries to verify it will get a different answer.
Mixing sources in one table. A vendor figure for one model beside an academic figure for another, compared as though they used the same method. They did not.
Treating a self-reported number as audited. Manufacturer data is the best available for most models. It is still the manufacturer’s own work, and your report should label it that way.
Claiming a saving you did not make. Buying a lower-carbon model is not a reduction against last year unless you also replaced the same number of machines. Avoided emissions and reduced emissions are different claims.
The fourth is the one that causes real trouble, because it reads as progress while the fleet grows.
The decision table
| Situation | Scale | Setup | Primary pain | Recommended starting point |
|---|---|---|---|---|
| First emissions report due | Under 50 | No device records | No data at all | Count devices and ages before anything else |
| Report exists, numbers questioned | 50 to 200 | Mixed vendor sources | Figures not comparable | Pick one source and restate on one boundary |
| Fleet growing faster than headcount | Any | No redeployment | Buying new by default | Measure the redeployment rate first |
| Devices not coming back | Distributed | No recovery route | Forced replacement purchases | Fix recovery, then recount |
| Customer or investor questionnaire | Any | Ad hoc answers | Cannot show method | Write the boundary down and reuse it |
What to put in a report you have to defend
Someone will eventually ask where your number came from. Write the report so that question is easy to answer.
Name the boundary in the same sentence as the figure. Say cradle to grave or cradle to gate, and say which stages you counted. A reader who knows the field will stop reading if that is missing.
Attribute every figure to its source and say whether it is self-reported. Manufacturer data is acceptable and widely used. Presenting it as independent is not.
Separate the fleet number from the per-device number. Total emissions move when headcount moves, which can hide a real per-device improvement or invent one that did not happen.
Report service life and redeployment rate alongside the carbon figure. They are the two numbers you can actually change, and showing them makes the report useful rather than decorative.
State what you did not count. Peripherals, monitors and docks are usually excluded, and saying so costs nothing. Discovering the omission later costs credibility.
Where a managed service changes the arithmetic
If redeployment is the lever, the constraint is usually operational rather than analytical. Somebody has to collect the machine, check it, wipe it, store it and match it to the next joiner. Teams without that capability buy new instead, whatever the policy says.
That is the gap a laptop lifecycle management service is built to close, by running collection, wiping, storage and reissue as one chain. It is worth being plain about the limits. A service of that kind does not generate your emissions figure, it does not audit anybody’s data, and it will not make a growing fleet smaller. What it changes is how many of your returned machines are fit to issue again, which is the input the arithmetic above turns on.
For how this fits a wider reporting obligation rather than a single number, our guide to equipment and ESG reporting covers what tends to get asked for.
See how the collection and reissue chain works in practice with device retrieval and redeployment across countries, or read our guide to equipment and ESG reporting.
Frequently asked questions
What is a device carbon footprint?
It is an estimate of the greenhouse gas emissions attached to one machine across its life, expressed in kilograms of CO2 equivalent and produced using life cycle assessment. The figure always comes with a boundary that says which stages were counted and over what assumed life. Without that boundary the number cannot be compared with anything, because two assessments of the same laptop will disagree sharply depending on whether raw material extraction is folded into manufacturing or reported on its own line.
How much of a laptop’s carbon comes from manufacturing?
It depends entirely on the study. HP publishes 153.10 kg of a 172.00 kg total for the EliteBook 830 G10, which is 88.14 per cent. Across the wider literature the share runs from roughly a quarter to almost all of the total, because studies draw the boundary differently and assume different use phases. The practical takeaway survives the disagreement: manufacturing is the largest single stage in most published assessments, so the length of service life matters more than anything you can change about how the machine is run.
Is it better to replace an old laptop with an efficient new one?
Rarely, on carbon grounds alone. Using the HP breakdown, the entire use phase is 15.50 kg while manufacturing is 153.10 kg, so an early replacement spends about ten times the use-phase total to buy a fractional efficiency gain. Replacement makes sense when a machine genuinely cannot do the work, when repair costs approach replacement, or when it fails a security requirement. It does not make sense as an emissions measure.
Can I trust manufacturer carbon figures?
They are the most model-specific data most teams can get, and they are self-reported. The Climatiq record for the EliteBook 830 G10 labels the source as HP itself. Use the figures, cite them, and say in your report that they are manufacturer declarations rather than independently audited results. That distinction protects you if somebody checks, and leaving it out is the most common way these reports get picked apart.
What should we measure if we have no carbon data at all?
Count devices and their ages, then work out the median months between issue and retirement for everything retired in the last year. Add the redeployment rate, which is the share of returned devices that went out again to another person. Those two numbers are countable from your own records, they are the levers you control, and they will tell you more about your footprint than a borrowed per-device figure ever will.
Do monitors and accessories count towards the footprint?
They carry their own embodied carbon and most device-level reports exclude them. That is a defensible choice as long as you state it. The risk is quietly leaving them out and then being asked, because a fleet with a monitor and a dock per person has a materially larger total than the laptop figures alone suggest. Write the exclusion into the method note the first time and it stops being a problem.
Does a lower-carbon purchase count as a reduction?
Only if you replaced the same number of machines as the year before. Buying a lighter-footprint model while the fleet grows is an avoided emission against a hypothetical, not a reduction against last year. Reporting it as a reduction is the error most likely to be challenged, and separating the per-device figure from the fleet total makes both claims honest.
