Two laptops of similar size and similar specification, two published environmental reports, two numbers that ought to be comparable. One says the machine is responsible for about 300 kilograms of carbon dioxide equivalent across its life. The other says 60. Neither maker is lying, neither report is sloppy, and the gap is not a measurement error. It is four decisions, taken before either measurement began, that nobody prints on the summary page.
Where the number comes from
A product carbon footprint is the sum of four phases: making it, moving it, using it, and disposing of it. For computing hardware the first phase dominates — published assessments put manufacturing at 75 to 85 per cent of a laptop's lifetime total, and one maker's own report attributes 79 per cent of a phone's footprint to production.
Those shares are what make the headline figures look authoritative. They are also the first thing that moves when an assumption changes, because the manufacturing number is nearly fixed while the use number is entirely a projection — a guess about how long you will keep the thing and where you will plug it in.
The first lever: how many years the report assumes
Every report picks a service life. Three years, four, five — the choice is disclosed in the methodology and almost never in the summary, and it multiplies the entire use phase.
Take a laptop drawing about 30 kWh a year, which is an ordinary figure for a machine used daily for work. Over four years that is 120 kWh. Over five, 150. Nothing about the hardware changed; the report's assumption did, and the use phase grew by a quarter.
This is why a longer assumed life can make a machine look worse in absolute terms while making it look better per year of service — and why a maker choosing a short assumed life quietly shrinks the only part of the footprint its engineering does not control.
The second lever: which country's electricity
Here is where the numbers stop being comparable at all, because the same kilowatt-hour carries wildly different carbon depending on the grid that supplied it.
Published grid intensities: the global average sat near 445 g CO₂/kWh in 2024. The European Union averaged 213. Germany, inside that same union, came in at 363. China was 560. The United Kingdom's official conversion factor has fallen from 207 to 177 to 131 g/kWh across three successive annual releases, as the last coal station closed and wind output rose.
Now run our laptop through the two extremes. Four years in the United Kingdom at 131 g: 120 kWh × 0.131 — 15.7 kg of carbon dioxide equivalent for the entire use phase. Five years in China at 560 g: 150 kWh × 0.560 — 84 kg.
The same machine, drawing the same watts, doing the same work, has a use-phase footprint that varies by a factor of five depending on two lines in a methodology appendix. Eighty-four kilograms is not a rounding difference — it is more than the entire published footprint of several current phones.
What that does to the famous percentage
Put those two use-phase numbers next to a manufacturing figure of 300 kg and watch the headline share move:
- UK grid, four years: 15.7 of 315.7 — manufacturing is 95 per cent of the total.
- Chinese grid, five years: 84 of 384 — manufacturing is 78 per cent.
Both of those are defensible calculations of the same laptop, and both have been published about comparable machines. It also explains a discrepancy that looks like incompetence when you first meet it: one peer-reviewed assessment attributes 38 per cent of a laptop's footprint to the use phase, against the 15 to 25 per cent implied by manufacturers' own reports. That study assumed a longer life on a dirtier grid. It is not wrong. It answered a different question.
The third lever: where the boundary is drawn
The accounting framework that nearly all of these reports follow splits emissions into three scopes, and the third — everything in the supply chain that the company does not own — is where a computer's footprint actually lives. It has fifteen categories, and a report can be fully compliant while including a different subset than its competitor.
Common divergences: whether employee commuting and business travel attributable to the product line are counted; whether the semiconductor fab's emissions come from supplier-specific data or an industry average; whether packaging, the charger and the cable are inside the boundary; whether end-of-life is modelled as recycling credit, which can subtract from the total, or excluded.
None of this is hidden — it is all in the methodology section. It is simply not in the number, and the number is what gets compared.
The fourth lever: which configuration was measured
The last one is the simplest and the most consistently overlooked. Storage and memory carry real embodied carbon, and reports are published per configuration. One published family of phones runs from 66 kg at 128 GB to 107 kg at 1 TB — a 62 per cent increase in footprint from a storage tier, on a device otherwise identical.
So a maker quoting the footprint of its base configuration against a competitor's mid-tier is not comparing products. It is comparing order codes.
How to compare two reports honestly
The exercise takes ten minutes and is the only way these figures become useful.
- Find the assumed service life in both. If they differ, recompute the shorter one's use phase pro rata before comparing anything.
- Find the grid factor in both. If one used a global average and the other a national one, substitute your own country's published factor into both — for a UK buyer, that is the figure that describes your actual electricity.
- Compare manufacturing only. It is the largest term, it is the one the maker controls, and it is the least sensitive to assumptions. If you only take one number from a report, take that one.
- Check the configuration. Match storage and memory tiers or the comparison is meaningless.
- Note what the total does not include. A report that subtracts an end-of-life recycling credit is reporting a different quantity from one that does not.
What the spread actually tells a buyer
Published footprints for laptops range from about 275 to 585 kg across makers, and for phones from 29 to 107 kg. Some of that range is genuine engineering difference — a smaller battery, a simpler board, a supply chain running on cleaner power. Much of it is the four levers above.
Which leads to an unwelcome conclusion for anyone hoping to shop on this number. Two footprints from different manufacturers are not comparable as printed, and treating them as a league table rewards whichever company made the most flattering assumptions. The figure is real, and it is a description of a model, not a property of an object.
What the reports are good for is comparing a maker with itself. When the same company publishes the same model across generations, using its own unchanged methodology, the difference between those numbers is a genuine engineering signal — and it is the only comparison in this field that requires no correction.
The one figure that survives every correction
There is a way to use these reports that none of the four levers can distort, and it is the number nobody prints: the footprint divided by the years of service you actually get.
Manufacturing is fixed at the moment the machine is built. If it took 300 kg to make, that 300 kg is spent whether the laptop lasts three years or eight. Amortise it:
- Replaced after 3 years: 100 kg a year.
- Kept for 4: 75 kg a year.
- Kept for 6: 50 kg a year — a third less than the four-year case.
- Kept for 8: 37.5 kg a year.
Now compare that with the decision the reports are meant to inform. The entire published spread between laptop models is 275 to 585 kg, and most of it is methodology rather than engineering. The spread available from keeping one machine two years longer is a third of its whole manufacturing footprint, and it depends on nothing but a repair and a decision.
Which is the uncomfortable conclusion for a shopping comparison: the choice between two similar machines moves the number far less than what you do with the machine afterwards. That is also why a report's assumed service life is worth reading twice — a maker assuming three years is describing a replacement cycle, not a lifespan, and it is the assumption that flatters the total.
One more caution about comparing a maker with itself, since we recommended it above. Methodologies get revised between generations, and a company that changes its assumed life or its grid factor produces a discontinuity in its own series. If two of its reports differ by more than the engineering plausibly did, check whether the methodology page changed before crediting the improvement.
The check that takes a minute
Open any product environmental report and go straight past the pie chart to the methodology page. Two lines decide almost everything: the assumed years of use and the electricity grid factor. Write both down before you read the total.
If a report does not disclose those two lines, its total cannot be compared with anything — including the same company's number from two years ago. That absence is itself the finding, and it takes less time to establish than reading the summary you were meant to read.
How this was put together
Six independent sources sit under the figures above: published assessments putting manufacturing at 75 to 85 per cent of a laptop's lifetime footprint, and a manufacturer's own report attributing 79 per cent of a phone's footprint to production; a peer-reviewed assessment attributing 38 per cent to the use phase under different assumptions; national and global grid carbon intensities — 445 g/kWh globally, 213 for the European Union, 363 for Germany, 560 for China; the United Kingdom's official conversion factors falling from 207 to 177 to 131 g/kWh across three annual releases; the greenhouse-gas accounting framework's three scopes and the fifteen categories of the third; and published product footprints spanning 275 to 585 kg for laptops and 29 to 107 kg for phones, including one phone family varying from 66 to 107 kg by storage tier alone.
The derived figures are ours: the 15.7 kg and 84 kg use-phase totals for one 30 kWh-a-year laptop under British and Chinese assumptions, the factor of five between them, the resulting swing in the manufacturing share from 95 to 78 per cent, and the 62 per cent footprint increase attributable to a storage tier.
