Hydrogen is not clean by default. What makes it clean, or fails to, is how it is produced — and, as a study published in Nature Communications in 2026 argues, how that production is counted. Gunther Glenk, Nikolas Holler and Stefan Reichelstein examine how the rules of CO₂ accounting shape the economics of hydrogen production, and reach a conclusion that should unsettle anyone who treats accounting as paperwork: the bookkeeping can matter almost as much as the technology. Utilities not built, and thus not replacing fossil CO₂ move the baseline up from zero towards the emissions that continue when the rules are too rigid.
The starting point is familiar. Most hydrogen produced today comes from natural gas or coal and carries substantial CO₂ emissions unless that carbon is captured and permanently stored. Electrolysis shifts the question to the power supply: renewable electricity can deliver hydrogen with very low emissions, fossil electricity cannot. So far, so uncontroversial.
Two identical plants, two different numbers
The central argument goes further. The CO₂ intensity attributed to a batch of hydrogen does not follow only from what physically happens at the plant. It follows from the accounting framework: which emissions are inside the boundary, how electricity consumption is valued, over what period the balance is struck, and what weight is given to guarantees of origin and other contractual instruments.
Two technically and physically identical installations can therefore end up with different CO₂ figures, depending on the system used to count them. That is not a rounding error. Under a threshold-based support regime, it is the difference between a bankable project and a stranded one.
Average or marginal: which power plant answers the electrolyser?
The sharpest of these choices is between average and marginal emissions. An average approach assigns the electrolyser the mean CO₂ intensity of the regional electricity mix. A marginal approach asks a harder question: which additional generation actually responds to this additional demand?
Run the electrolyser during a renewable surplus and it may absorb electricity that would otherwise have been curtailed. Run it at times of high demand and it may pull additional fossil generation onto the grid. Same machine, same efficiency, radically different system effect — and, depending on the method chosen, radically different reported emissions.
The clock: annual balances versus hourly matching
The same tension appears in the time dimension. An annual balance can show that enough renewable electricity was procured for the hydrogen produced, while the electrolyser was in fact running at hours when that electricity was not available. Hourly matching describes reality more accurately.
It also has a price. Strict temporal requirements raise the credibility of the emissions balance and simultaneously reduce electrolyser utilisation, which pushes up the cost of every kilogram produced. Credibility and cost pull in opposite directions, and no rule set escapes that trade-off.
Additionality
A comparable conflict runs through the additionality principle: the renewable electricity claimed for hydrogen production must come from newly built generation. Without that condition, a producer could buy certificates from existing renewable plants while the electrolyser draws ordinary grid power — a renewable balance on paper that produces no additional renewable electricity in the system.
Strict additionality rules prevent that. They also make projects more expensive and slower, and they bite hardest in exactly the phase when an industry is trying to get off the ground.
The emissions of the plant that never gets built
This is where we would add a point the regulatory debate almost always leaves out. Rules are usually assessed by asking what could go wrong if a project proceeds: a weak balance, an overstated claim, public money spent on something that is not as clean as advertised. Fair enough. But there is a second failure mode, and it is invisible in every impact assessment because it produces no project to inspect.
If rules are so rigid that an electrolyser is not built at all, the emissions it would have displaced simply continue. The refinery keeps running on grey hydrogen. The truck keeps burning fossil diesel. The heat keeps coming from fossil gas. Nothing appears in any register, because nothing happened — and that is precisely the problem. The baseline is not zero. The baseline is the fossil status quo, running on undisturbed.
Any serious assessment of a carbon accounting regime therefore has to ask both questions: what happens if we act, and what happens if we do nothing.
Thresholds decide where the money goes
The stakes rise as public support is increasingly tied to CO₂ intensity thresholds. Tax credits, contracts, guarantees of origin and procurement standards can all depend on a calculated emissions value. Where those thresholds rest on assumptions that do not reflect how the power system actually responds, projects can be certified as clean and receive funding while delivering little real climate benefit. Rules that are too rigid do the reverse: they exclude technologies that would in fact run with low emissions under specific local conditions.
Regulation has to reconcile environmental integrity and verifiability with the economics of building a new industry. Neither objective survives if the other is ignored.
Beyond the colour code
The study also questions the familiar colour labels — green, blue, grey. A more informative approach would assess the full production pathway by its actual life-cycle emissions: upstream methane losses in the gas supply chain, the origin of the electricity used, and the timing and location of production. Buyers could then compare hydrogen on measurable emissions rather than on technology categories or marketing terms.
That is a point worth holding on to across the whole Power-to-X value chain. What matters for e-fuels, e-methanol and e-SAF is not the label on the molecule but the emissions attached to it — and the same accounting questions run through all of them, because they all start with hydrogen and recycled CO₂.
What this means for Switzerland
Switzerland is not writing these rules from scratch. It is deciding how closely to follow the European framework, where RFNBO criteria already encode answers to every one of these questions — additionality, temporal correlation, geographic correlation. Those answers are not neutral. They determine which Swiss projects can qualify, at what utilisation, and at what cost.
The overall conclusion of the study is blunt enough to be worth repeating: well-designed accounting rules steer investment towards projects that genuinely reduce emissions, while badly designed ones can direct billions towards projects that look low-carbon on paper while emissions are merely relocated elsewhere in the energy system. And, we would add, towards a third outcome that no ledger records at all — the projects that were never built, and the fossil emissions that continued in their place.
Source: Gunther Glenk, Nikolas Holler and Stefan Reichelstein, on carbon accounting rules and incentives for hydrogen production, Nature Communications, 2026. Reported by Scienmag.
🤖 AI transparency: This text and its visual were created using AI, reviewed and approved by a human.
Power-to-X Congress Switzerland 2026 – 22 September 2026, Kursaal Bern: “Reality Check with Net Zero”. Programme and registration.