There is no single carbon footprint for a solar panel. Model the silicon supply chain behind a standard module and you get a range instead, and it is wide enough to decide whether you win a tender or lose it.
In our own cradle-to-gate modelling of a monocrystalline (mono-Si) module, the product carbon footprint sits anywhere between 145 and 420 kg CO₂e per square metre. Same technology, same 19.5%-efficient, 170-micron wafer. The total swings by roughly 60% from the highest- to the lowest-impact supply chain we tested. Nothing about the finished panel tells you where in that range it lands. The answer is buried two and three steps upstream, in how the silicon was purified and which grid powered the furnaces.
That gap, between “a solar panel’s carbon footprint” and your solar panel’s carbon footprint, stopped being academic in December 2025. It is now written into European procurement law.
This article is for the people who have to produce that number and stand behind it: module makers, polysilicon and wafer producers, and the developers and EPCs who now have to submit a defensible footprint to win public capacity. It covers what the footprint actually is, where it comes from, why two identical-looking modules differ so much, the specific 2026 pressures making this commercial rather than reputational, and what a credible answer looks like.
What is the carbon footprint of solar panel manufacturing?
The carbon footprint of solar panel manufacturing is the total greenhouse gas emissions, expressed in CO₂ equivalent, released across every stage of making a module: from quartz extraction through silicon purification, crystal growth, wafering, cell fabrication and final assembly. It is the embodied or cradle-to-gate carbon, everything before the panel leaves the factory, before it generates a single kilowatt-hour.
This is a different question from the one most of the internet answers. Most published figures address the life-cycle footprint per unit of electricity generated, usually quoted somewhere between 40 and 100 gCO₂/kWh once you spread manufacturing emissions across 25–30 years of generation, which is how the familiar line “solar is roughly 20 times lower-carbon than coal” is derived. That framing is correct, and it matters for anyone deciding whether to install solar. It is the wrong tool for anyone deciding which modules to make, buy, or bid.
For a manufacturer or a developer, the operative fact is the one the per-kWh average hides: manufacturing dominates the life-cycle footprint. Independent analyses put the manufacturing share at roughly two-thirds to the overwhelming majority of a panel’s lifetime emissions, because the use phase is, by design, almost emission-free. Decarbonising solar, in other words, is mostly a manufacturing problem. And manufacturing emissions are where the variation lives.
Where the emissions actually come from
Walk the mono-Si supply chain stage by stage and the picture is consistent and, for many people, counter-intuitive.
A mono-Si module is built in five major steps: metallurgical-grade (MG) silicon, then solar-grade (SoG) silicon, then the single crystal (the Czochralski ingot), then wafer, then cell, then module. When we disaggregate the climate-change contribution of each stage, two of them swamp the rest: solar-grade silicon production and single-crystal growth. During solar-grade silicon production, over half the impact comes from the electricity consumed in purification, with the remainder split between process heat and the embodied impact of the MG-silicon feedstock. During crystal formation, electricity and the SoG silicon feeding the furnace are again the dominant contributors.
For the three downstream stages (wafer, cell, module) the embodied impact of the silicon already produced upstream overwhelms everything added later. The aluminium frame, the solar glass, the silver paste, the encapsulant are all real, but secondary. The carbon was effectively baked in before the wafer was ever sliced.
The physics behind this is the energy ladder of purification. Bringing silicon to the 99.9999999% purity a solar cell needs is brutally energy-intensive. The literature puts the energy demand at roughly 10–20 kWh/kg for MG silicon, 40–200 kWh/kg for solar-grade silicon, and 120–300 kWh/kg for single-crystal silicon. Each step up in purity is a step up in energy, and therefore a step up in exposure to whatever generated that energy.
Which is why electricity is the master variable. The IEA estimates electricity provides around 80% of the total energy used in solar PV manufacturing, concentrated in exactly those polysilicon, ingot and wafer stages that need precise high-temperature heat. And today, coal generates over 60% of the electricity used for global PV manufacturing, well above coal’s ~36% share of global power generation, because production is concentrated in Chinese provinces such as Xinjiang and Jiangsu, where coal exceeds 75% of the grid. China’s share of polysilicon and wafer manufacturing is heading toward 95%.
So the single largest lever on a panel’s carbon footprint is not the panel. It is the carbon intensity of the grid that purified its silicon.
Why two identical modules can differ by 60%
Here is the finding that catches teams off guard, and the one worth dwelling on, because it changes where you should spend effort.
When we varied the supply chain in our model, we tested two things in sequence: the source of the metallurgical-grade silicon feedstock (higher-impact Chinese production versus lower-impact European), and the electricity grid mix across all five manufacturing stages (coal-heavy versus hydropower).
Swapping the MG-silicon source alone moved the final module footprint by only about ±5%. That surprises people who assume “where the raw material comes from” is the headline. It isn’t, because the per-kilo impact difference in MG silicon, real as it is, gets outweighed by the energy and material intensity of every stage that comes after it.
Swapping the electricity across the supply chain is what moves the number. Coal-powered purification versus hydro-powered purification is the difference between the top and bottom of that 145–420 kg CO₂e/m² range. The location decision, specifically the grid intensity at each energy-hungry stage, is the dominant design choice. France’s grid runs at roughly 56 gCO₂/kWh on the back of nuclear; a coal-heavy grid can be an order of magnitude higher. Same process, same equipment, radically different embodied carbon.
The practical implication: if you are trying to lower a module’s footprint, optimising the bill of materials or chasing a marginally cleaner MG-silicon supplier is fiddling at the edges. The leverage is in the energy mix of the purification and crystal-growth stages. A life-cycle model is what tells you that, and tells you by how much, before you commit capital to a factory location or a supply agreement.
The 2026 shift: carbon footprint is now a tender qualifier, not a talking point
For years, a low manufacturing footprint was a nice-to-have, a line in a sustainability report. That is over in Europe, and the change is recent enough that many manufacturers haven’t repriced it.
France set the precedent. Since 2019, the French energy regulator (CRE) has required a carbon footprint assessment, the Évaluation Carbone Simplifiée (ECS), for modules bidding into its public tenders. Carbon is not a footnote in the score; non-price criteria can account for up to 30% of the total tender evaluation, and modules above the threshold (historically around 550 kg CO₂e/kWp for certain segments) can be excluded regardless of how cheap they are. The market responded by certifying: REC’s TwinPeak 4 at around 450 kg/kWp, Canadian Solar’s HiHero below 400, JinkoSolar’s Tiger Neo at 448–450. A 10% improvement in the non-price score has been worth bidding 2–4% higher on electricity price and still winning. South Korea built a parallel grading system. The carbon number became a commercial instrument.
The EU has now generalised the principle. Under the Net-Zero Industry Act, the non-price criteria provisions (Article 26) apply from 30 December 2025, with the first compliant auctions expected through 2026. Member States must apply non-price criteria to at least 30% of annually auctioned renewable capacity, or 6 GW per year, with those criteria weighted at 15–30% of the evaluation. Of the eight sustainability criteria the implementing act lists, life-cycle carbon footprint is the first. Because solar PV is over 80% import-dependent, it also triggers the Act’s resilience criterion. SolarPower Europe is pushing for carbon footprint to be a minimum environmental requirement for PV in public procurement, with a harmonised methodology under PV Ecodesign that accepts supplier-specific electricity data.
Read those two developments together and the message to a manufacturer or developer is blunt. The regulation sets the deadline. But the driver is access: if your module’s footprint isn’t quantified and defensible, you don’t get scored, or you get scored badly, on a growing share of the European market. A spend-based estimate or a generic industry-average figure won’t survive a tender review. The buyer’s procurement team, and the certifying body behind it, will want to know how the number was built.
What a credible answer actually requires
This is where most published guidance stops, at “manufacturing has a footprint, polysilicon is the biggest part, decarbonise the grid.” True, but not actionable. A footprint that holds up under tender scrutiny or third-party certification needs more than a direction of travel. A few principles, drawn from doing this work under ISO scrutiny:
Model the system, not the panel. The carbon is upstream, so the assessment has to reach upstream, to the specific silicon purification route and the specific grid powering it, not a global-average placeholder. Cradle-to-gate boundaries that start at “module assembly” miss the two stages that actually decide the result.
Use a recognised methodology, declared openly. ISO 14040, 14044 and 14067 set the framework; the EU’s Environmental Footprint (EF 3.1) method is the reference for European compliance contexts; a transparent background dataset such as ecoinvent underpins the rest. Which method, which dataset, which functional unit (per m², per Wp, per kWp): these aren’t formalities. They are what makes one manufacturer’s number comparable to another’s, and what a certifier checks first.
Carry the uncertainty. A single point estimate invites the question “compared to what?” Professional LCA work reports the sensitivity, the range you get when you vary grid mix and feedstock source, because that range is the insight. It shows where the impact reduction lives and demonstrates that the modeller understands the system rather than the spreadsheet.
Treat the footprint as a roadmap, not a report card. Once the two hotspots are managed, the model lets you tackle the next tier (heat generation, chemical inputs like sodium hydroxide and argon, aluminium, glass, metallisation paste) in order of impact, against a target. That sequence is the difference between a number filed for compliance and a number that drives where you build and who you buy from.
Modelling this in XYCLE: the reason the grid mix swamps the feedstock choice only becomes visible when you can hold the supply chain constant and swing one variable at a time. In XYCLE, the silicon purification and crystal-growth stages are built once as connected unit processes, then Scenario Analysis lets you run the same module against a coal-heavy and a hydro-powered grid and read the difference straight off the contribution analysis, which is how you get the 145–420 kg CO₂e/m² spread rather than a single misleading point. Parameter Management drives wafer thickness, efficiency and purification energy as inputs you can flex without rebuilding the model, and because the outputs are EF 3.1-aligned and traceable through the Data Navigator, the number you hand to a CRE or NZIA tender is one a certifier can follow back to its assumptions.
Here is where 2026 leaves a manufacturer. A solar panel’s footprint is one of the real paradoxes of the energy transition: a low-carbon technology that still carries a fossil-fuel cost at birth, concentrated in a few energy-hungry silicon stages in a few coal-heavy places. That paradox is shrinking as grids clean up. It is also being turned into a market filter right now. The manufacturers who can quantify their position credibly, and act on the hotspots that actually move the number are the ones who will keep market access as the filter tightens.
