EU Advances CBAM Transition Rules for 2026

The Analysis

Strictly speaking, nothing changed on 6 March 2026 in the sense that matters to a cash forecast. What the European Commission advanced that day is a set of transition rules for the 2026 reporting year under the Carbon Border Adjustment Mechanism, and what those rules do is clarify reporting obligations for importers. Clarify is the verb the Commission used and it is the right one. A reporting obligation is not a payment obligation, and the two arrive on different dates.

The two dates to hold on to are 2026 and 2027. The 2026 rules govern what an importer must declare. The 2027 date is the one Brussels says this step keeps on track: the full charge. A reporting duty tells you what you owe. A charge is the thing you pay. Between them sits a year in which the only thing that changes is the quality of the data you hold.

I want to get one misreading out of the way early, because it is the one I have watched do damage in every transitional regime I have worked through. Transition is not a grace period. The word describes a phase in which the reporting duty runs at full strength while the financial duty is switched off, and it exists precisely so that the data exists by the time the money falls due. Anyone who reads transition as the European Union not having decided yet will discover in 2027 that the decision was made and that the record they needed was supposed to be built during 2026.

Let me put the instrument in the shape I use when I am advising, which is: who owes what, to whom, on what base, and from when.

Who owes it. The obligation in a border carbon mechanism of this design attaches to the importer, not to the producer sitting outside the customs territory. That sounds like a small point and it is the largest one in this column, because the data the importer must declare is data the importer does not hold. Embedded emissions are calculated at the installation where the goods were produced. The importer stands at the border with a legal duty to report a number generated several thousand kilometres away, by a party who has no legal duty to supply it. That asymmetry is not a drafting oversight. It is the mechanism: the border is where the European Union can reach, so the border is where the duty is placed, and the commercial consequence is that the data request has to be built into purchase contracts rather than assumed.

To whom. Here is the part I rarely see in coverage of this mechanism, and it changes how you should plan. Instruments of this type in the European Union are administered at member state level. The Commission proposes and adopts the framework; the reporting declarer deals with a competent authority in the member state of import. That authority is the one that receives the report, asks the questions, handles corrections, and, where the instrument provides for it, applies penalties. So the practical answer to who you talk to is not Brussels. It is the authority in the member state where your goods are declared, and if you import through more than one member state you have more than one counterparty. I am not going to name yours. That is a lookup you should do, per member state, before assuming one process covers your European footprint.

On what base. This is the second thing people get wrong, and it is the one that breaks spreadsheets. The duty is not ad valorem. The tax base in a carbon border mechanism is embedded emissions, expressed in tonnes of carbon dioxide equivalent, not the customs value of the consignment. Two consequences follow immediately, and both are counterintuitive to anyone who has spent a career on ad valorem duty. A high-value, low-carbon consignment and a low-value, high-carbon consignment of the same commodity are not treated alike, and the difference runs in the opposite direction from the one your pricing intuition expects: a cheap, dirty tonne can carry more exposure than an expensive, clean one. And because the base is physical rather than monetary, a change in the market price of the goods does not change the bill. Your exposure moves with the carbon price and with the emissions intensity of the specific installation, not with your purchase price.

Which goods. The mechanism applies to goods listed in its annex, and scope in this instrument has always been defined by reference to the Combined Nomenclature. I am going to be careful here, because this is exactly the kind of list that changes and exactly the kind of list people recite from memory. I am not going to recite the annex from recollection. What I will say is structural: scope is a code question. If the code is in the annex, you are in. If it is not, you are out, and no amount of reasoning about the carbon content of your product brings you back in. Have your broker or your compliance function run your actual codes against the annex text as it currently stands, and put a date on that check, because the annex is amendable.

And here is the exception I would circle before anyone relaxes. Scope by code means classification is doing the work. A classification error here is not a rate error, it is an in-or-out error. I have spent a great deal of my professional life on classification disputes, and the pattern is always the same: the error surfaces years later, discovered by somebody else, and the correction runs backwards. If your covered and your uncovered codes sit adjacent in the same chapter, that adjacency is where your risk is, and it is worth a binding ruling rather than an internal opinion. Whether a binding ruling is available to you and what it costs is a question for your adviser in the relevant member state.

Now the burden reduction, and this is where I would slow right down, because relief is being read as retreat. The Commission's stated aim in this step is to reduce administrative burden while keeping the 2027 full charge on track. Those are two statements. They are not in tension, but they are not the same statement either. A burden reduction in a reporting regime can take several shapes, and each one moves the burden to a different place.

If the relief takes the form of a threshold below which reporting is not required, then the operative questions are what the threshold is measured in, over what period, and at what level of aggregation. Measured in mass, in value, or in emissions. Per consignment, or per importer per year. And whether crossing it exempts you from the whole declaration or only from part of it. Those are four different instruments, and they do not produce the same answer for the same importer. A per-importer-per-year threshold, for instance, is not something you can plan around by splitting volumes across entities, and I would want that confirmed in the text rather than assumed.

If the relief takes the form of making default values easier to use, then the trade moves in time rather than in amount. Using a default value is cheaper this year and may be worse in the year the charge starts, and the reason sits in the design rather than in any view I hold about policy: default values in mechanisms of this kind are set at a level that gives the importer an incentive to produce verified installation data. They are meant to be unattractive. So the cheap path in 2026 is not necessarily the cheap path in 2027, and the two should not be chosen by the same person on the same day.

If the relief takes the form of fewer data fields or a longer reporting interval, then the administrative cost genuinely falls and nothing else moves. That is the benign case. I do not know which of these this step contains, and I am not going to guess. What I will say is that all three leave the 2027 date standing. Burden relief in a transitional reporting regime is relief from reporting, and it is not relief from the eventual calculation. If your reading of this announcement is that the European Union has decided to go easier on carbon, read the second clause again.

Let me do the arithmetic, with every assumption written down, because the gap between a reporting duty and a charge only becomes visible once it is in numbers. I have no figures from the Commission's step to work with, so every input below is an assumption. Replace them with your own, and with the published price when one exists.

Assume an importer bringing 8,000 tonnes a year of a covered iron or steel product into the European Union. Assume the embedded direct emissions of that product, as produced by the installation supplying you, are 1.8 tonnes of carbon dioxide equivalent per tonne of product. Assume the carbon price relevant to the eventual charge is EUR 85 per tonne of carbon dioxide equivalent. None of those three numbers comes from the announcement. They are mine, for the arithmetic.

Embedded emissions: 8,000 tonnes times 1.8 is 14,400 tonnes of carbon dioxide equivalent a year. At EUR 85, that is EUR 1,224,000 a year of gross exposure before any adjustment for free allocation or for the phase-in of the mechanism. I am deliberately not applying a phase-in percentage, because that percentage is set in the instrument and I am not going to quote a figure I have not read in the current text. Run it three ways instead, because the spread is the point. At full coverage the number is EUR 1,224,000. At 75 percent coverage it is EUR 918,000.

At 50 percent it is EUR 612,000. The difference between the top and the bottom of that range is EUR 612,000 a year, which is larger than most other line items in a compliance budget by an order of magnitude, and it turns entirely on a parameter set by the regulation rather than by you. If you take one planning lesson from this column, take that one. Do not build a business case on your guess at the phase-in percentage. Build it on the range.

The second calculation is the one I think most importers are getting wrong, and getting wrong in the wrong direction. Assume that, purely for the arithmetic, the default value available to you for this product sits 20 percent above your actual verified installation figure. So the default is 2.16 tonnes of carbon dioxide equivalent per tonne rather than 1.8. On 8,000 tonnes, that is 17,280 tonnes instead of 14,400, a difference of 2,880 tonnes. At EUR 85, the annual cost of declaring on default rather than on verified data is EUR 244,800.

Against that, put the cost of having verified data. Assume you have six relevant installations supplying the European Union. Assume verification costs you EUR 15,000 per installation per year, which is an assumption and should be replaced with a real quote. Six times EUR 15,000 is EUR 90,000 a year. On these assumptions the net position in the first year of the charge is roughly EUR 154,800 in favour of verified data, it repeats every year, and the verification cost does not grow with your volumes while the exposure does. If your tonnage doubles, the argument for verified data doubles with it.

The third calculation is the one nobody puts in the board pack, because it concerns the reporting year rather than the charge year. Assume the 2026 reporting programme costs you the equivalent of one full-time analyst for two quarters, plus a data collection exercise across your suppliers. Assume a fully loaded annual cost of EUR 90,000 for that person, so two quarters is EUR 45,000, and add EUR 25,000 of supplier chasing and external help. Call it EUR 70,000 of effort in 2026. Set that against the EUR 244,800 annual cost of sitting on defaults, or against the low end of the exposure range above. On these assumptions the reporting programme pays for itself inside the first year of the charge, before any discussion of penalties for non-compliance. That is the argument for treating 2026 as a build year rather than a waiting year, and it is arithmetic, not enthusiasm.

I should be careful about penalties, because this is where I see the most confident nonsense. The instrument provides for consequences where reporting obligations are not met, and in European instruments of this shape the penalty regime is typically set and applied at member state level within a framework set at Union level. I have not verified the current penalty provisions and I am not going to quote a figure or a multiplier from memory. Get it from the competent authority in each member state where you declare, in writing, and note that the answer may differ between member states. What I will say without quoting anything is the structural point: a penalty for failing to file is a different exposure from a higher charge caused by filing on defaults, and only the second one is optional.

Who this lands on, and how hard. The layered answer is that legal incidence and economic incidence fall on different parties, and the distance between them is a commercial problem rather than a legal one. A large importer with its own customs function and a single member state of import has a manageable task: build the reporting line, go to the suppliers, hold the data. A smaller importer declaring through an indirect representative in a member state where it has no establishment has two problems at once. It still carries the reporting duty, and it is one step further away from both the producer data and the authority.

And the non-European producer sits in an odd position: no direct duty under the instrument, but commercial pressure to supply verified emissions data, because without it the importer's declared numbers get worse and the importer's willingness to keep buying at the same price falls. If you are negotiating 2027 supply terms during 2026, this is the year to put a data clause in, with a defined format, a defined deadline and a defined consequence, because a clause saying the supplier will use reasonable endeavours to provide emissions data is not a clause that produces data.

What I would actually do, with dates. This column is dated 6 March 2026, so the 2026 reporting year is already underway and the useful horizon is the rest of it.

By 31 March, have the trade compliance function produce a scope determination against the annex as it currently stands, listing every Combined Nomenclature code you declare that falls inside it, with annual tonnage and entered value against each, and with the member state of import recorded. No code parked in an undecided column. If a code is genuinely uncertain, it goes to a classification review with a date on it, not into a holding bucket.

By 30 April, identify the installation behind each in-scope code. Not the supplier. The installation. A trader buying from three mills does not have one data problem, it has three, and the emissions intensity of each is a different number. Record, for each: name, country, whether verified installation data is available, and who on the supplier side is accountable for delivering it.

Before 30 June, send the data request, in writing, to the suppliers covering the top 20 percent of your in-scope tonnage, with the format specified and a return date. Do not send a general letter asking for carbon data. Specify the parameter, the unit, the reporting period and the evidence you will accept, because a vague request produces vague answers, and vague answers are unusable in a declaration.

Before 30 September, decide, per installation, whether you are declaring on verified data or on default values for this reporting year, and write the decision down with the reason. The reason matters. A decision made to save effort in 2026 is visible in 2027, and it should be a decision rather than an accident.

Before 31 December, run the charge calculation internally on the year's actual tonnage and your own emissions figures, at the three coverage levels I used above, so that the first number your finance function sees in 2027 is one they have already seen in draft. And set a standing review: assign the Official Journal of the European Union and the competent authority's own publications to a named person, monthly.

The traps, in the order I expect people to hit them. The reporting perimeter, meaning which movements count: goods released for free circulation and goods placed under another customs procedure are not the same question, and if you have been told that a warehouse or an inward processing arrangement keeps you outside the reporting duty, ask for the basis in the text rather than accepting the assurance. The classification: covered and uncovered codes sitting adjacent in the same chapter, and a classification position taken for duty reasons years before carbon existed. The effective date: the reporting duty for 2026 and the charge from 2027 are two different start points, and a single mental model in which this mechanism begins in 2027 will produce a year of missing declarations.

The default value lock-in: declaring on defaults in 2026 because it is easy, then finding that the same decision is expensive in 2027 and that reversing it requires data taking two quarters to collect. And the contract trap: a clause obliging the supplier to provide data the supplier does not have, which reads well and delivers nothing.

Here is where I would concede the argument to someone who reads this differently. If the burden relief in this step turns out to be a threshold that puts you below the reporting line for 2026, then the build programme I described is optional this year, the money can wait, and I would not spend EUR 70,000 complying with a duty I do not have. If the 2027 charge is delayed, the reporting duty does not delay with it, so the data programme still stands while the cash forecast moves. If default values are set closer to actual intensities than I have assumed, the EUR 244,800 shrinks and the case for verified data weakens. All three are testable, and none of them should be settled by reading a summary.

The thing I would not concede is the distinction I opened with. A reporting obligation is not a payment obligation, and a transition period is not a holiday. What the Commission did on 6 March was make one of those clearer while restating its commitment to the other. If you remember one sentence, remember this one: the data you fail to collect in 2026 is the data you will be paying for in 2027.

↑ Back to the key points

— By Dr. Ingrid Voss