The full life of a kilowatt-hour

From the deposit and the factory to the socket, decommissioning and recovery.

A kilowatt-hour is not born at the meter

Before the first watt: the invisible phase

The number that shows up on an electricity bill is the end of a much longer chain than it appears. Before a single power plant, turbine or panel exists, a resource has to be located — coal, gas, uranium, sun, wind, moving water or underground heat — and the materials for the machine that will harness it also have to be extracted: iron and coal to make steel, bauxite for aluminum, copper for every meter of cable, silica sand for a solar panel's silicon, limestone for the cement of a foundation. Mines, quarries, wells, smelting furnaces and component factories have already consumed energy — usually from the existing electricity grid or from fossil fuels — before the new installation produces its first kilowatt-hour. This phase is known as "embodied energy," and it is completely invisible on the electricity bill of whoever uses that electricity afterward, even though it is real and measurable.

A concrete example helps make this less abstract: manufacturing the steel, concrete and copper of a single medium-sized onshore wind turbine requires on the order of several hundred tonnes of materials, each of which went through its own chain of mining, transport and industrial processing before reaching the turbine factory. None of that appears once that turbine is standing and spinning without burning any fuel.

The operation phase: the one everyone sees

Once built and connected, an installation enters its operation phase: it converts an energy source into electricity for years, needs periodic maintenance, and delivers in practice an amount of energy different — almost always less — than its nameplate "rated power" suggests (the so-called capacity factor, which varies enormously between technologies: a thermal or nuclear plant can run close to maximum almost all year, while a solar farm only generates at full power for a few hours a day). During this phase, the electrical grid raises voltage, transports electricity through high-voltage lines, steps it back down, and distributes it, constantly balancing generation with consumption. If the primary source varies — as happens with sun or wind — extra tools are needed: weather forecasting, interconnection with other regions, storage or flexible demand, each with its own material cost that is rarely attributed to the source that made it necessary.

It is in this phase that almost all media and advertising attention on energy concentrates: it is the phase people can see, the one with a smoking chimney or its absence, the one that fits into short phrases like "zero emissions." Precisely for that reason it is the phase that least helps distinguish between technologies, because almost all low-carbon sources look very similar to each other during operation — the real difference lies in the other two phases.

The final service: where part of all that earlier effort is lost

When the kilowatt-hour finally reaches a home or a factory, its journey still is not over: another machine — a light bulb, a heating element, a motor, a processor — converts it into the service actually wanted: light, heat, cold, motion or information. Every one of those conversions has unavoidable physical losses (the second law of thermodynamics does not negotiate), so the final "useful service" a person receives is always less than the electrical energy that entered the appliance. This matters for life-cycle comparisons because two energy sources that produce the same amount of electricity do not necessarily produce the same amount of useful service, if the appliances consuming that electricity have different efficiencies.

The ending that almost never gets counted: decommissioning and waste

After twenty, thirty or sixty years — depending on the technology — an installation stops operating. What remains is parts, oils, reinforced concrete, metals, in some cases used nuclear fuel, in others solar panels or wind turbine blades that are hard to recycle, and often land that has to be restored before it can be returned to another use. This phase — the one furthest in time from the moment the decision to build the installation was made — is also the one most easily forgotten when calculating a cost or a footprint, simply because it happens decades later and is often paid for by a different generation than the one that enjoyed the electricity.

A worked example: why the boundary changes the answer

Imagine someone asks "how many grams of CO₂ equivalent does generating a kilowatt-hour with coal emit, compared with nuclear or solar energy?" If the boundary of the analysis is drawn only around combustion or fission at the plant itself (the operation phase), the answer is dramatic for coal — burning it directly releases a large amount of CO₂ per kilowatt-hour — and practically nil for both nuclear and solar, because neither burns anything during operation.

If instead the boundary is widened to include the whole life cycle — mining coal or uranium, building the plant or the panel, transport and decommissioning — coal's figure barely changes (combustion already dominated the result, so adding the rest of the life cycle does not shift the total much), but nuclear's figure, and especially solar's, stop being zero: they start to reflect the energy and materials it took to manufacture the reactor or the panel. Coal remains, by a wide margin, the option with the most emissions per kilowatt-hour under either boundary — that does not change — but the relative difference between nuclear and solar can change depending on the study, and the exact figure for each depends crucially on which electricity mix powered the panel's factory or the reactor's components. The data tab of this leaf develops this comparison with concrete, sourced figures.

This is not an isolated case or a trick: it is the structural reason why a good analysis always states its boundary explicitly, and why any figure without that statement — "this technology emits X grams of CO₂ per kWh," without saying which phase it refers to — should be read with caution.

The complete question that replaces the short question

It is not enough to ask "how much does producing this energy cost?" The complete question — the one that applies this section's method to any technology, without exception — includes: who finances the installation and at what cost of capital?, what fuel and what materials does it need, and where do they come from?, how much does it actually generate, not just how much it could generate in theory?, what grid infrastructure accompanies it and what does that infrastructure cost?, what emissions and what waste does it leave, and at which stage of the life cycle?, how long does the installation last?, and finally, who pays for its decommissioning once it stops operating? Life cycle assessment does not automatically decide which option is best — that remains a decision that depends on values and local context — but it does guarantee that no clean-looking stage hides, unnoticed, a dirty one somewhere else or at some other time.

How to read — and how not to read — an energy figure

Five questions before accepting any comparison

1. What boundary was used? Only operation, or the whole life cycle — extraction, manufacturing, transport, operation and decommissioning? A serious study states this explicitly on the first page, not in fine print at the end.
2. Is the same service being compared? "Per kilowatt-hour generated" is not the same as "per kilowatt-hour of useful energy delivered to the end user," nor "per vehicle manufactured," nor "per kilometer driven." Changing the unit of comparison can change the winner without any individual data point being false.
3. What electricity mix powered the manufacturing? Two physically identical solar panels can have very different carbon footprints if one was manufactured with electricity mostly from coal and the other with electricity mostly from hydro or nuclear. The final technology is the same; the manufacturing process is not.
4. Is it a range or a single number? Any life-cycle magnitude — emissions, water, land — varies by study, country and year. A single number with no range and no source is a red flag, not a more reliable data point.
5. Who pays for closure? A comparison that only covers construction and operation, without mentioning decommissioning, waste or land restoration, is incomplete by definition — not because those phases are necessarily expensive, but because omitting them is a boundary choice that has to be made explicit.

Common mistakes, in both directions

In favor of a technology: citing only its "operational" emissions figure (often zero or near zero) as if it were the full life-cycle figure, ignoring manufacturing and decommissioning.

Against a technology: citing the worst documented case across the entire life cycle — the most severe accident, the study with the widest and least favorable boundary — and comparing it with the ideal, incident-free operation of the alternative technology.

Both errors use real, verifiable data; the problem is not the truthfulness of each individual figure, but the asymmetry in how they are chosen and combined to reach a conclusion that was already decided in advance.

What to do when two reliable sources disagree

It is normal, not a sign that one of them is lying. Before dismissing either study, check: the year of publication (technology improves over time and old figures can be outdated), the country or region of reference (the local electricity mix changes the manufacturing result), and whether both state the same boundary and the same functional unit. Most discrepancies that look like contradictions resolve once those three conditions are compared — and the ones that do not resolve that way are, precisely, the real uncertainty that any honest comparison has to acknowledge.

The same data, two different boundaries

The table that demonstrates the method

CO₂-equivalent emissions per kilowatt-hour generated, comparing operation only (what happens at the plant itself, counting nothing before or after) against the full life cycle (extraction, manufacturing, transport, operation and decommissioning):

Sources: NREL — Life Cycle Assessment Harmonization; IPCC AR6, WG3, chapter 6.

What this table teaches about the method, not just about these six sources

Notice the pattern: for coal and gas, widening the boundary from "operation only" to "full life cycle" barely moves the number, because combustion already concentrates the vast majority of that technology's climate impact. For hydropower, solar, wind and nuclear, on the other hand, widening the boundary is the difference between saying "zero emissions" and saying "low but non-zero emissions" — a qualitative change, not just a quantitative one, even though the final figure remains very small compared with any fossil fuel.

This explains why advertising and public debate about energy tend to use the "operation only" boundary: it is the one that makes low-carbon technologies look flawless, and the one that least changes the already-bad result of fossil fuels. Neither boundary is "the dishonest one": both are mathematically correct for what they measure. The error lies in not stating which one is being used, or in using a different boundary for each technology being compared in the same conversation.

A second illustration: water during operation

The same pattern shows up with other magnitudes. Solar PV and wind generation consume practically no water during operation (they boil nothing to turn a turbine), while a thermal or nuclear plant can withdraw more than 100 liters of water per kilowatt-hour for cooling, even though much of it is returned to its source. If the question is "how much water does this technology consume during its daily operation?", the answer clearly favors solar and wind. If the question is "how much water was needed in total, including manufacturing the materials?", the picture is more nuanced, though the difference between wet-cooled technologies and those without cooling remains large. The lesson repeats: the exact question asked partly determines the answer obtained.

What this table is missing

None of the figures on this page include land footprint, biodiversity impact, or the risk of a low-probability, high-consequence accident — a life-cycle analysis focused on CO₂ emissions is only one of several dimensions that need comparing, not the only one. The data pages for each specific source, in the "Energy Sources" section, develop those other dimensions one by one.

Sources

• The EU's Joint Research Centre explains the life cycle assessment methodology, including how a system boundary and a functional unit are defined.
• The IPCC analyzes whole energy systems, not just generation, in AR6, WG3, chapter 6.
• NREL harmonizes hundreds of life-cycle studies of different electricity technologies into a single comparable framework: Life Cycle Assessment Harmonization.
• On water use by technology, with and without cooling: Water use of electricity technologies: A global meta-analysis (ScienceDirect).