On 31 December 2025, Fastmarkets assessed the premium American buyers were paying for low-carbon domestic hot-rolled coil at exactly zero dollars. Two days later, the equivalent European differential sat at €100 to €170 a tonne. Same steel, same decade, and a gap that explains why the question in the headline has no single answer.
Key takeaways
- The EU’s carbon border mechanism entered its definitive phase on 1 January 2026, with enforcement from 1 April 2026.
- The first quarterly CBAM certificate price, published 7 April 2026, was €75.36 per tonne of CO2; the Q2 figure was €75.28.
- Free allocation only starts unwinding in 2026, at 2.5%, reaching 100% in 2034. The bill is scheduled, not immediate.
- ArcelorMittal walked away from a €1.3 billion grant-supported German project in June 2025, which tells you the subsidy was not the binding constraint.
The price of a tonne of carbon is now a published number
For twenty years the honest answer to “what does a tonne of CO2 cost us” was a shrug and a forecast. That changed on 1 January 2026, when the EU Carbon Border Adjustment Mechanism entered its definitive phase, covering iron and steel, cement, aluminium, fertilisers and hydrogen. A grace period ran to 31 March 2026, and from 1 April only authorised CBAM declarants could import those goods.
The European Commission now publishes a certificate price each quarter, calculated as the weighted average of EU ETS auction clearing prices over the preceding three months. The first, published on 7 April 2026, came in at €75.36 per tonne of CO2. The second, published on 6 July 2026, was €75.28. From 2027 the price moves to a weekly cycle and certificates are bought through a central platform from February of that year.
That stability matters more than the level. A capital committee can discount a number that is published on a schedule. It cannot discount a rumour.
The premium buyers will pay depends entirely on where you sell
Regulation sets the cost side. The revenue side is set by what customers will actually pay for a lower-carbon tonne, and here the divergence is stark rather than gradual.
| Market | Green steel differential (Fastmarkets) | Assessment date |
|---|---|---|
| Europe, flat-rolled vs HRC index | €100 to €170 per tonne | 2 January 2026 |
| China, reduced-emission flat steel | 0 to 500 yuan per tonne | 5 January 2026 |
| United States, domestic HRC | $0 per short ton | 31 December 2025 |
Through 2025 the European premium held a fairly narrow band of roughly €120 to €180 per tonne. A European mill therefore has something an American mill does not: a market-observable price for the thing it would be spending billions to produce. That is the whole ballgame.
The arithmetic on one tonne of steel
Worth doing explicitly, because the headline numbers are routinely misused. Fastmarkets puts conventional blast furnace and basic oxygen furnace production at roughly 1.8 to 2.2 tonnes of CO2 per tonne of steel, with a global average near 1.9. Take the Q1 2026 reference price of €75.36 and hold it flat, purely as an illustration.
At full exposure, 1.9 tonnes of CO2 at €75.36 comes to about €143 per tonne of steel. That lands inside the European green premium band, which is the optimistic reading and the one most decks stop at. But full exposure is not the 2026 situation. Free allocation to CBAM sectors only begins unwinding this year, at 2.5%, then 5% in 2027, 10% in 2028, 22.5% in 2029, 48.5% in 2030, and 100% in 2034.
Run the same calculation at 2026 rates and the carbon cost is closer to €3.60 per tonne of steel. By 2030 it is around €69. The gap between those two figures is the entire investment problem: the revenue premium is available now, the cost pressure that justifies the capex mostly arrives after 2029, and the plant has to be built in between. Our calculation, our assumption of a flat carbon price, and both should be stress-tested against your own emissions intensity rather than a sector average.
What the cancellations actually tell us
In June 2025 ArcelorMittal cancelled its electric arc furnace and direct reduced iron project for Bremen and Eisenhüttenstadt, walking away from around €1.3 billion in German grant support. Across the European sector, more than twenty-five decarbonisation projects have been delayed or halted, and SSAB paused an electric arc furnace plan.
The instructive part is that public money was available and was declined. Capital grants cover construction. They do not cover the operating spread between hydrogen or clean electricity and the coke it replaces, year after year, for the life of the asset. A company that cannot see that spread closing will refuse free construction money, which is a rational decision and a fairly damning verdict on how the incentives are sequenced. We traced the wider version of this shift in our piece on how old-school industries are being rewired.
The counter-example is Stegra, which closed €1.4 billion of additional financing in April 2026 led by the Wallenberg family, taking total project capital to roughly €7 billion, and is targeting production in 2026. One greenfield project at that scale does not settle the argument. It does show that the money exists when the offtake and the power contract line up.
Three terms that keep getting muddled
Carbon-neutral
Balancing emissions to a net figure of zero, whether by eliminating them or by purchasing offsets against what remains. It is not the same as emitting nothing, and the share of the claim carried by offsets is the first thing worth asking about.
Green premium
The amount a buyer pays above the conventional index for a lower-carbon equivalent. It is an assessed market price, not a policy target, and as the table above shows it can be zero. Any business case that assumes a premium without naming the market and the assessment date is assuming its own conclusion.
Abatement cost
The cost of avoiding one tonne of CO2 by a specific measure. It varies enormously between routes, which is why blended sector averages are close to useless for a single site. Efficiency retrofits usually sit at the cheap end, primary process changes at the expensive end.
So, can industry go carbon-neutral without going broke? In the European Union, for a producer with access to cheap clean power and customers inside the regulated perimeter, the numbers are starting to work and will work considerably better after 2029. For a producer selling into a market where the premium is assessed at zero, the same project is a bet that the regulation arrives before the balance sheet gives out. That is not a technology problem, and no amount of engineering optimism will make it one.
Curious where the replacement energy is supposed to come from?
The operating spread that killed those steel projects is, at bottom, a hydrogen cost question.
Sources: CMS legal update on the first CBAM certificate price announcement, covering the Q1 2026 price of €75.36 per tonne published 7 April 2026 and the definitive-phase timetable; European Commission quarterly CBAM price for Q2 2026 published 6 July 2026; Fastmarkets, “Global green steel markets in 2026”, 16 January 2026, for regional differentials and blast furnace emissions intensity; EU ETS free allocation phase-out schedule 2026 to 2034; reporting on ArcelorMittal’s June 2025 cancellation of the Bremen and Eisenhüttenstadt project and on Stegra’s April 2026 financing round. Price differentials are market assessments on specific dates and move continuously. The per-tonne arithmetic is our own illustration at a constant carbon price, not a forecast. Updated August 2026.
