CBAM Default Values Cost 10% Extra in 2026 and 30% From 2028, First Filing Sep 2027

The Analysis

Strictly speaking, this is not a carbon tax. It is a charge for arriving without data, dressed up in a default value. On 1 January 2026 the Carbon Border Adjustment Mechanism moved from reporting to charging, and EU importers of cement, iron and steel, aluminium, fertilisers, electricity and hydrogen now owe certificates on the embedded emissions of what they bring in. The certificates go on sale in February 2027. The first annual declaration and surrender falls due on 30 September 2027. The BSI session in Foshan this week put the same slide in front of exporters that competent authorities have been showing since spring, and the core of it is this: if you cannot produce verified actual emissions, you pay on the EU default value, and that default carries a mark-up that climbs. Ten per cent in 2026, twenty in 2027, thirty from 2028. Fertilisers are the outlier at one per cent in every year.

I want to be exact about the direction of that mark-up before anyone quotes it back at me. It is not in the CBAM Regulation. Article 7, and the replaced Annex IV, refer only to a proportionately designed mark-up to be fixed in implementing acts. The actual figures sit in Implementing Regulation (EU) 2025/2621, as corrected by (EU) 2026/1740 with retroactive effect from 1 January 2026. The distinction is not academic. An implementing act can be amended on a much shorter clock than the parent regulation, so the thirty per cent from 2028 is a policy setting, not a treaty commitment. Do not build a ten-year supply contract on that number without a clause that lets it move.

So the interesting question is not what the mark-up is. It is whom the rule actually reaches, and the honest answer is narrower and wider than people assume, both at the same time. Narrower, because there are real exemptions and boundary conditions, and a surprising number of worried importers are not in scope at all. Wider, because one clause almost nobody reads turns the whole scheme into a cash-flow instrument that bites roughly a year before the carbon cost shows up. The boundaries first. That is where the money hides.

The scope test is three questions, and you can answer all three from your own customs records without asking anyone. Is the good listed in Annex I by CN code. Is it being released into free circulation in the EU, or entering inward processing that will later be released. And does your annual mass of covered goods exceed fifty tonnes net.

That third one is doing more work than any other sentence in the scheme, and it is routinely misread. The threshold is a single mass-based figure of fifty tonnes of CBAM goods per year, aggregated across everything in scope, not fifty tonnes per shipment and not fifty tonnes per CN code. If you ship four tonnes of steel fasteners a month, that is forty-eight tonnes and you are outside. Day traders and small assemblers who import a mixed bag of components often fall under it without realising, and importers who assume it is per consignment over-comply and spend money they did not have to spend.

The first question trips people too, because Annex I and Annex II answer different questions and get conflated constantly. Annex I is the list of goods in scope at all. Annex II is the list of goods for which only direct emissions are counted: iron and steel, aluminium, hydrogen, and, added by the Omnibus, electricity. Everything else in scope carries direct and indirect. Cement and fertilisers are the ones that pull in indirect emissions, and for a Chinese or Turkish cement exporter that is where the number doubles. If your trade compliance lead tells you the whole scheme is direct-only, they have read Annex II and mistaken it for Annex I. Get the two lists side by side.

Then there is the list that makes people relax too early. Annex III is the set of third countries and territories outside CBAM, and its logic is not origin in the abstract, it is whether the goods were already carbon-priced. Iceland, Liechtenstein and Norway are inside the ETS, which is why their goods are out. Switzerland is fully linked. That is the actual test, and it is why the same product from Turkey and from Norway are not equivalent even when the mill looks similar. The mechanism is designed to price carbon that was not paid elsewhere. If someone has already paid it, the adjustment is not owed.

Now the clause almost nobody reads, and the one I want importers to sit with. Article 22(2), as the consolidated text stands, requires you to hold certificates covering fifty per cent of the cumulative emissions of the current year by the end of each quarter, starting in 2027. Read that as a carbon cost and you have missed it. It is a cash requirement, and it arrives roughly a year before the annual surrender. You must have fifty per cent of the year's embedded emissions covered in certificates by 31 March, more by 30 June, more by 30 September, tested quarter by quarter. That is a working-capital line, not an emissions line. A mid-size steel importer with, say, twenty thousand tonnes of annual embedded emissions is holding certificates on ten thousand tonnes by the first quarter test.

And here is the part that surprises the finance side. The fifty per cent hold is computed on the default values without the mark-up. The mark-up is added later, at declaration. So the cash the treasury department is asked to find in March is not the same number as the tax that shows up in the profit-and-loss in September. Two numbers, two dates, two owners. If you briefed your CFO on one, you briefed them on the wrong one.

Let me put an order of magnitude on it, with the assumptions written down. Take a mid-size importer of iron and steel goods, twenty thousand tonnes of annual embedded emissions on the default-value, default-methodology route. Assume a CBAM certificate price around eighty euros per tonne, which is in the range the Q3 2026 market has been printing, and hold it flat for the arithmetic. On defaults, the mark-up alone is ten per cent in 2026, which is eight euros per tonne of default value, so eighty thousand euros a year of pure penalty for having no data. In 2027 the same volume on the twenty per cent mark-up is one hundred and sixty thousand euros, and from 2028 on the thirty per cent rate it is two hundred and forty thousand. Three years, three escalating numbers, all of it avoidable in principle by verified actual data.

That is the arithmetic the mark-up is built to force. It is not a revenue line the EU is counting on. It is designed to be unattractive enough that you switch to the actual-values route, which is why the difference between the two routes is the number that should sit in your budget, not the certificate price on its own. A clean mill's verified figure can be materially below the country default, in which case verified data saves you both the mark-up and some of the base. A dirty mill's verified figure can be above the default, in which case you may rationally choose the default and pay the mark-up. Model both, per supplier, and you will find the answer is not uniform across your vendor list.

The transmission timing matters, and the honest answer is that the binding constraint is not data, it is the verifier. Accreditation of the first CBAM verifiers was expected only around September 2026. Article 8(1), and the Omnibus change behind it, means verification by an accredited verifier is required only on the actual-values route, but that is exactly the route you want. So the scarce resource in the first declaration cycle is verifier capacity, not your supplier's meter readings. A supplier who tells you in June 2027 that the data exists but no accredited verifier is available until the autumn has an honest problem, and you inherit it. Book the verifier before you need it, in the first half of 2027, or the data is worth nothing.

Here is what I think the coverage of this story has missed, and it is the point I would put first if I were briefing a board. The real question is not whether CBAM raises your cost. It is whether the mark-up reaches you through the clause you least control, which is upstream in your supplier's data chain. Your supplier's clean, verifiable emissions figure is a document, and documents have lead times measured in months. The mark-up does not fall on you because you failed a compliance test. It falls on you because a supplier three tiers back could not produce a verified precursor figure in time, and precursor data has to come from an accredited verifier's report covering the relevant reporting period. That is the transmission channel nobody puts in a slide. Your exposure is partly your supplier's paperwork discipline, and you cannot verify it from a spreadsheet.

I want to put the price of waiting on the page, because "we will get the data next year" is the most expensive sentence in this whole file. Run the arithmetic forward for the same twenty thousand tonnes and the same eighty euros a tonne. Stay on the default route one more year and 2027 costs you one hundred and sixty thousand euros of pure mark-up; fetch verified data a year earlier and 2026 would have cost you eighty thousand. The year you waited is not free, and the ladder makes the penalty explicit: ten, then twenty, then thirty, each step locked by the calendar rather than by anything you can negotiate. The saving from getting data in place for the 2026 reporting year, rather than the 2027 one, is not a percentage point or two. It is the difference between the ten per cent step and the twenty per cent step, and it is recurring.

Then the counterfactual, the condition under which the whole thing flips. If your goods are not in Annex I, none of this reaches you, and no amount of carbon accounting changes that. If your annual covered mass is under fifty tonnes, you are outside, and the correct action is to document that position rather than to build a compliance programme you do not need. If your supplier is in Norway or Switzerland, the goods were already ETS-priced and the adjustment is not owed. If, on the other hand, your supplier will not share installation-level data, you will be on the default route whether you like it or not, and the mark-up is a certainty, not a risk. The right move in that last case is not to argue with the supplier in 2027. It is to start the data conversation in the fourth quarter of this year, while there is still leverage from next season's purchase order.

What I would do, concretely. Map every EU-bound line by CN code and mark it Annex I, Annex II or out of scope, signed off, by the end of December. Run the fifty-tonne test on a rolling annual mass ledger, in tonnes net, so the position can be proven on demand, not reconstructed after a query. Send every covered-goods supplier a written data request this quarter, installation-level identification and precursor-level emissions, with a hard reply date in the first half of 2027. Engage an accredited verifier for the 2026 reporting year before you need them, and treat their calendar as the constraint. Put the two routes, default-with-mark-up and verified-actuals, side by side in the 2027 budget as separate lines. And separate the cash from the cost in your treasury plan: a certificate-holding facility sized on the fifty per cent quarterly test, priced on defaults before the mark-up, refreshed every quarter, owned by treasury, not by compliance.

I want to name one more exemption that people miss, because it sits inside the boundary rather than outside it. Inward processing is not a free pass. If goods enter the EU under inward processing and are later released into free circulation, the CBAM obligation is triggered at that release, and the reporting period is keyed to the date of release, not the date of import. For a manufacturer who brings in a precursor, works it, and sells the finished good into the EU market, the charge can land a year or two after the container cleared, attached to a transaction that looked purely internal on the day it happened. That is exactly the kind of boundary condition that a compliance programme built on import dates will miss, and it is the one I would ask a client to test against their processing records before they conclude they are outside.

There is also a question of who the counterparty is, and it is worth stating plainly because it is a common source of dispute. The obligation sits on the EU importer, or on the indirect customs representative where one is appointed. The exporter does not owe the certificate. What the exporter controls, though, is the data that decides the importer's number, which means the exporter holds pricing power without holding the legal duty. If you are selling into the EU and you have not yet been asked for installation-level emissions, you will be, and the terms on which you answer are about to become a negotiating position rather than a courtesy.

Let me close the loop on the counterfactual, because it is the part of the brief that keeps the analysis honest. The claim that CBAM taxes everything entering Europe is false, and the claim that it taxes nothing because the numbers are small is also false. The rule reaches a defined list of goods, above a defined mass threshold, into a defined set of destinations, with defined carve-outs for territories already inside a carbon price. Change any one of those and the conclusion moves. Your job is not to predict which way the policy goes. It is to know, in your own file, which side of each of those lines you sit on, so that when the implementing acts move, as they will, you already know whether the change touches you.

One practical note on sequencing, since three dates are now in play. The data request to suppliers is a this-quarter action, because lead times on verified emissions run in months and the verifier calendar is the real bottleneck. The verifier engagement is a first-half-of-2027 action, because accreditation capacity is finite and everyone will want the same quarter. The certificate facility is a treasury action that starts before the first quarterly test in 2027 and runs all year. Ship the first two late and the third becomes a rushed, expensive borrowing decision, which is the one outcome the mark-up was designed to make you avoid, and the one the market will happily sell you.

The taxes are not the hard part. The dates are. This one tests something you hold either trivially or not at all, which is whether your supply chain can produce a verified document on a deadline you did not set. Fix that, and the mark-up becomes a line item you chose to avoid. Miss it, and it becomes a penalty you paid for a year you never got to revise.

One more thing about the calendar, because it is where importers lose the argument with their own finance team. The obligation is annual but the tests are quarterly, and the first quarterly test in 2027 comes before any 2026 declaration has been filed. You will be asked to put certificates behind emissions you have not yet declared, on a default value you have not yet confirmed, for a supplier whose verifier has not yet been accredited. Anyone who has run a treasury function will recognise the shape of that problem: it is not a tax problem, it is a working-capital and certainty problem, and it wants a facility, not a debate.

The number to size it on is fifty per cent of the year's cumulative emissions, computed on default values before the mark-up, refreshed each quarter as actual import volumes land. If treasury is told the number only once a year, in September, they are being asked to solve a March problem in September, and the answer will be a scramble.

There is a second reason to start the supplier conversation early that has nothing to do with goodwill. The mark-up is a function of missing data, and missing data is not a uniform condition across your vendor list. Some suppliers will have installation-level meters and an energy ledger already, because their home market made them. Others will have nothing beyond a utility bill and a production total. The two look identical in a purchase-order email and they are a decade apart in compliance readiness. Until you ask, your cost model treats them the same, and the mark-up will sort them out for you at the worst possible moment, in the declaration year, after the purchase orders are signed.

I would also separate the two exposures that get muddled into one worry. The first is the carbon cost itself, which rises as free allocation is phased out toward its 2034 end point and is largely a function of the price of an EU allowance. The second is the no-data penalty, which is a function of nothing but your own paperwork and your supplier's. The first you cannot control; the second you can, almost entirely. Importers who treat them as a single lump of uncertainty end up doing nothing on the half they could have fixed, and that is the most expensive mistake in the whole file, because it is the avoidable half.

So the discipline is to keep the two apart in your own books. Show the carbon cost, however large, as a cost of doing business in the EU and price it into the landed number. Show the mark-up as a waste line, something you pay for a document you did not obtain in time, and put an owner and a date against it. One is a tax you plan for. The other is a failure you can retire."""

↑ Back to the key points

— By Dr. Ingrid Voss