The Real Embodied Energy of Energy Technology

Every so often someone tells us, with great confidence, that a solar panel never generates as much energy as it took to manufacture. It's a fair question — and in 1975 it was a genuinely open one. It has since been answered, measured, and re-measured by people whose entire professional purpose is to be sceptical about exactly this. This article is the honest version of that answer, with the working shown.

We're an engineering firm, and we're suspicious of the word "green" full stop — it's a marketing colour, not an engineering unit. Energy will be in demand for as long as we're around, so efficiency and efficacy come first; environmental footprint is a consequence of doing the engineering properly, not a slogan to lead with. We specify kit when the numbers work. So when we say a solar panel pays back its manufacturing energy in about a year, we'd like you to be able to check that — which is why every figure below has a numbered reference to a primary source at the bottom of the page. If you have better figures, or a better argument, on any of this: tell us, and we'll update the article.

What "embodied energy" actually means

Embodied energy is all the energy consumed in making a thing: digging the raw materials out of the ground, refining them, manufacturing the product, shipping it, and installing it. Its sibling, embodied carbon, is the greenhouse gas emitted along the way. Neither is visible on the product, which is precisely why it attracts confident claims in both directions — the sales brochure's and the saloon bar's. The kit can't defend itself either way; only measurement can.

The measuring discipline is called life-cycle assessment (LCA). It is not done by manufacturers' marketing departments. The reference numbers for solar come from the International Energy Agency's Photovoltaic Power Systems Programme (IEA PVPS Task 12), run jointly by the US National Renewable Energy Laboratory and the University of New South Wales, which publishes consensus figures and — importantly — the methodology, so rival researchers can pick holes in it.[1]

The headline metric for anything that generates energy is energy payback time (EPBT): how long the device must operate before it has produced as much energy as it took to make it.[a] Everything after that point is, energetically speaking, profit.

A word on carbon before we start quoting it. Carbon has become the accounting currency of energy debates, and we'll use it here because you can't compare technologies without a common unit. But it is one metric, not the whole assessment — the same life-cycle studies also track particulate matter, acidification, water use and mineral resource depletion,[1] and a single-minded crusade for "low carbon" can quietly shift harm into those columns instead. Energy efficiency is primary; footprint — all of it, not just the fashionable molecule — follows from that.

Solar PV: about a year

The IEA PVPS consensus figures for a rooftop solar system in Europe — and this includes the panels, mounting, cabling, inverter and the installation itself, not just the shiny rectangle — put the energy payback at 1.2 years for mono-crystalline silicon (the type on most UK roofs), and 0.9 years for cadmium-telluride thin film.[1] The assumed yield is 975 kWh per kWp per year, which is a decent match for the UK; a well-oriented system in the south of England will do slightly better, a north-facing one in Scotland worse.

Fraunhofer ISE — Europe's largest solar research institute — reaches the same neighbourhood independently: around 1.1 years for a system made in Europe and installed in northern Europe, 0.9 in the sunnier south. Over a conservative 20-year life, the system returns roughly twenty times the energy invested in it.[2] The panels we specify are warranted for 25–30 years, so twenty times is the pessimist's number. (For the record: we focus on European manufacturers — for footprint, warranty and traceability. As you'll see in the caveats, where a panel is made changes its rucksack considerably.)

One more thing, clearly labelled as our own back-of-envelope rather than a published figure: the consensus numbers above assume 20%-efficient modules, because the LCA databases lag the market. The modules we currently specify run around 23–24%. More output from roughly the same slab of material means, all else being equal, today's panels should pay back in about a year or slightly under. We'll swap in the official number when the databases catch up; until then, treat 1.2 years as the documented ceiling rather than the current state of the art.

And the carbon? Life-cycle emissions for mono-crystalline solar work out at about 43 g of CO₂-equivalent per kWh generated on the IEA's European assumptions.[1] A UK-specific analysis, accounting for our weather, puts the range at roughly 20–34 g/kWh depending on orientation and location.[3] For comparison, the average kWh from the British grid in 2024 — its cleanest year ever — carried about 124 g,[4] and it ticked up slightly to around 126 g in 2025.[5] Fossil generation can run as high as 1,000 g per kWh.[1]

So a solar panel on a UK roof isn't carbon-free — nothing manufactured is, and we'd encourage you to distrust anyone who claims otherwise — but per unit of electricity it's roughly a third of today's grid average, and the grid average itself only got this low because of the renewables on it.

Batteries: the honest framing

Here's where we depart from the sales script. A battery doesn't generate anything. "Energy payback time" doesn't apply, and anyone quoting you one for a battery is either confused or hoping you are.

What a battery has is an embodied cost, and what it does is move energy — from the middle of a sunny day to the evening, or from a cheap overnight tariff to peak time. And it charges a toll for the service: a lithium battery returns roughly 85–92% of the energy you put into it, every cycle.[7] Far from leveraging energy, a battery loses some on every pass — which is why the economics have to stack up on their own terms (they generally do, but that's a calculation, not a given). Whether the embodied cost is worth it depends entirely on what the moved energy displaces.

The numbers: a large 2024 analysis in Nature Communications puts the median manufacturing footprint of lithium iron phosphate (LFP) cells — the chemistry in most current home storage — at about 62 kg of CO₂-equivalent per kWh of capacity (range 54–69). Nickel-based NMC811 cells, common in EVs, come out around 74 kg/kWh (range 59–115).[6] The dominant driver is the electricity used in manufacturing, which means these figures fall as grids decarbonise — and vary a lot with where the cells are made.

Do the arithmetic on a typical 10 kWh home battery: roughly 620 kg of embodied CO₂e. At the 2025 grid average of ~126 g/kWh, that's the emissions from about 4,900 kWh of grid electricity. If your battery mostly shifts your own solar generation that would otherwise have been exported to displace someone else's grid consumption anyway, the carbon case for it is thin. We'll say that plainly, because nobody selling batteries will: on today's British grid, you buy a home battery for bill savings, energy independence and backup — legitimate reasons all — not primarily to save the planet.

There is a carbon argument for batteries with more legs, though it depends entirely on how the battery is driven. The average gram-per-kWh figure hides a daily swing: the electricity you draw at the evening peak is disproportionately made by gas plant fired up to cover it, while overnight and midday the mix is dominated by baseload and renewables — sometimes to the point of surplus. A battery that charges from the cheap, clean end of the day and discharges into the dirty, expensive end is displacing the grid's worst kWh with its best. Add to that a quirk of the way arrays are growing faster than grid connections: export limits mean an increasing amount of rooftop generation gets clipped and thrown away, and a battery is the only place for it to go. Run like that, the carbon sums improve considerably. But that's an operating pattern, not a property of the box — a battery is only as clean as its charging schedule. As the grid gets cleaner still, the carbon case gets thinner either way, while the financial case rises and falls with tariff design. Engineering is about knowing which argument you're actually making.

Heat pumps: a metal box that leverages everything upstream

A heat pump is, materially, an unglamorous object: steel, copper, aluminium, a compressor, some electronics and a kilogram or two of refrigerant. Whole-life studies of heating and cooling systems consistently find the embodied share — manufacture, installation, and refrigerant leakage over its life — comes to at most around a fifth of the system's total climate impact, and usually much less; the rest is the electricity it runs on.[8]

The refrigerant deserves an honest mention, because it's the one place a heat pump can quietly misbehave: older refrigerants like R410A are potent greenhouse gases if they leak, which is why leak-tight installation and proper end-of-life recovery matter, and why the industry is moving to low-impact refrigerants like propane (R290). We'll cover what happens at end of life — refrigerant recovery included — in a separate article.

But the reason heat pump embodied energy barely registers is leverage: every kWh of electricity in becomes roughly three to four kWh of heat out, harvested from outside air. And the fair comparison isn't with nothing — it's with the heat source you'd otherwise buy. A gas boiler also has to be manufactured (steel, copper, electronics — it has an embodied bill of its own), and then spends its entire working life burning fuel at slightly under one unit of heat per unit of gas. The heat pump carries a similar manufacturing rucksack but delivers three to four units per unit of electricity for fifteen to twenty years. Spread over the energy each machine handles in its lifetime, the embodied cost of either box is small — the difference between them is made almost entirely in operation, and that contest isn't close. That leverage — and why it isn't witchcraft — gets an article of its own: Heat Pumps Are Not Witchcraft.

Electric cars: bigger battery, same story

An EV carries the biggest battery most households will ever own, so it has the biggest embodied bill: producing a battery-electric car emits roughly 40% more than producing a petrol equivalent. The International Council on Clean Transportation's 2025 life-cycle analysis — the most comprehensive for Europe — finds that this head start is erased after about 17,000 km of driving, typically within the first two years, after which the EV pulls ahead every mile. Over the vehicle's life, on the European electricity mix, total emissions come out 73% lower than a petrol car (63 vs 235 g CO₂e/km); charge it on renewable electricity and the gap widens to 78%.[9]

Note what that number is not: it is not zero. "Zero-emission vehicle" is a tailpipe claim, not a life-cycle one, and conflating the two is exactly the kind of sloppiness that hands ammunition to people who'd rather nothing changed at all.

The caveats we're not going to hide

• Where it's made matters. Most figures above are averages. A panel or cell made on a coal-heavy grid carries a heavier rucksack than one made on hydro power. The IEA PVPS notes that the shift of wafer and cell production to Asia has pushed some impacts up even as efficiency pushed them down.[1]
• The data lags the technology. The consensus PV figures are for cell architectures that are already a generation behind what's currently shipping; newer types aren't yet in the inventory databases.[1] Given the direction every previous revision has moved, today's real numbers are probably somewhat better than the published ones — but "probably better" is a hypothesis, not a figure, so we quote the published ones.
• Degradation is real — and varies. The EPBT figures assume panels lose 0.7% of output per year.[1] That's already in the maths, and it's conservative for good modern kit: degradation depends heavily on cell technology and manufacturer, and the panels we typically specify carry a warranted degradation of 0.38% a year — roughly half the studies' assumption. On the other hand, a badly installed, shaded or never-cleaned system will underperform any warranty figure, which is an argument for competent design rather than against solar.
• Ranges, not points. Any single figure we've quoted has a spread around it. Where a range exists we've given it; where we've quoted a median, the reference has the distribution.

The verdict

The "solar panels never pay themselves back" line was a reasonable suspicion fifty years ago and a measurable falsehood today. Equally, "zero-carbon technology" is a marketing phrase, not an engineering one. The truth — pays back its energy in about a year, then spends decades in profit — doesn't need embellishment from either direction. It just needs references.