ArcelorMittal's Dunkirk EAF moves into permitting — and the project is smaller than the one it replaced

ArcelorMittal's mid-October update on its two northern French projects is, on the surface, routine progress reporting: €300 million committed at Dunkirk, a permitting clock started, five electrical steel lines at Mardyck converging on a spring 2027 commissioning. Read against the project's own history, though, the announcement marks two things more consequential than the headline numbers.

First, the Dunkirk scheme is no longer the project that was originally designed. It is a single 2 million t/y electric arc furnace replacing a plan that once included direct reduction and two furnaces. Second, the permitting stage is where the project's remaining risks now sit — not financing, which has been restructured around a different subsidy instrument, but electricity cost, scrap and metallics supply, and product mix.

What is actually happening at Dunkirk

The figures hold up against ArcelorMittal's own communications and the regulatory file, with some necessary corrections to how the trade press framed them.

The consultation is a three-month enquête publique-style procedure covering both the environmental authorisation and the building permit — not a single October event. That matters commercially: the building permit is a hard gate, and the January 2027 public meeting in Grande-Synthe is the most likely point for local objection to crystallise.

The project that was replaced

The most important context the routine reporting omits is that Dunkirk has already been re-scoped once, downward. As Senator Thomas Dossus set out in the French Senate on 25 February 2026, the original decarbonisation plan ran to roughly €1.8 billion and comprised a direct reduction unit plus two electric furnaces, backed by €850 million under France 2030. That configuration was suspended; the replacement is a single EAF at €1.3 billion, with the France 2030 grant suspended and support redirected to the energy savings certificates (CEE) mechanism.

The arithmetic of the change is worth stating plainly:

• Capital cost falls by roughly 28% (€1.8bn → €1.3bn).
• Primary metallics capacity falls much further: a DRI plant plus two furnaces becomes one furnace with no captive DRI.
• The CO₂ claim survives in weakened form. A single EAF consuming HBI/DRI and pig iron cannot capture the full benefit of green hydrogen-based direct reduction that the original integrated configuration was designed around.

ArcelorMittal has not published the tonnage split of its metallic charge, which is the single most important missing number. The 0.6 t CO₂/t figure will move materially with the HBI/DRI-to-scrap ratio and with the carbon intensity of the power contract. It is a design-point claim, not an operating guarantee.

On financing, the company states that CEE support will equal 50% of the €1.3 billion investment — roughly €650 million on its own arithmetic (my calculation; ArcelorMittal has not published the euro figure). That is a larger share of a smaller project, delivered through a mechanism tied to energy performance rather than a fixed capital grant.

Why the EAF decision is conditional, not strategic

ArcelorMittal has been explicit that the go-ahead rests on three external conditions: the EU steel import tariff-rate quota, reform of the carbon border adjustment mechanism, and a long-term low-carbon power supply contract signed with EDF. Europe CEO Geert van Poelvoorde has said further EAFs elsewhere in Europe will be studied cautiously.

That language should be read literally. Dunkirk is a template test, not the first of a fleet. The binding constraint is electricity: an EAF at 2 million t/y consuming roughly 350–400 kWh per tonne of liquid steel implies something in the order of 0.7–0.8 TWh per year at full run rate (analyst estimate, assuming typical scrap-based consumption and no hot-metal charging). The EDF contract is therefore as much a commercial prerequisite as the permits.

The tariff-quota condition is equally substantive. A single 2 million t/y EAF replacing blast furnace capacity inside the EU changes ArcelorMittal's own cost position only if imports are constrained at the same time; otherwise the company trades captive iron ore and coke for market-priced scrap and power while competing against imports that face neither carbon nor power costs.

Mardyck: the more commercially interesting asset

Mardyck, adjacent to Dunkirk, is further along and arguably the better business. Phase 1 — preparation, annealing-coating and slitting lines — began operating in March 2026; the first industrial coil was produced on 26 March, and the first commercially saleable electrical steel coil followed on 12 May. Phase 2, comprising the annealing-pickling line and reversing mill, is nearing mechanical completion, with initial tests targeted for spring 2027.

Two corrections to the trade report are worth noting: the "200 employees in March" figure is an end-state target, with roughly 175 in place at first coil and 140-plus cited in May; and the European release quotes US$500 million while French communications use €500 million — the euro figure is the one used in state-support documentation.

The strategic value here is different from Dunkirk. Mardyck is import substitution in a high-value niche: ultra-thin, high-polarisation, low-loss grades for traction motors and wind turbine generators, where European demand is rising with electrification and where non-EU supply is concentrated. A fully integrated French chain from hot band to finished electrical steel is a defensible position in a way that a single EAF is not. It is also the natural internal customer logic for Dunkirk's output quality — cleaner steel, tighter residuals.

What to watch

The permitting outcome in January 2027 is the near-term gate, but it is unlikely to be the deciding one. Three things matter more:

1. The EDF power contract terms. If the long-term supply deal is struck at a rate that holds the EAF's operating cost inside import parity, Dunkirk becomes a replicable model. If not, it becomes a demonstration asset.
2. The metallic charge split. Any disclosure of HBI/DRI versus scrap shares will reveal how much of the 0.6 t CO₂/t claim survives in practice — and whether ArcelorMittal intends to source DRI externally, which would embed imported green-premium cost in a commodity product.
3. Whether a second European EAF follows. Van Poelvoorde's caution is the most informative line in the whole announcement. A decision on the next unit, expected only after Dunkirk clears permitting and power pricing, will say more about the credibility of European green-steel capex than any single plant's commissioning.

For suppliers, the near-term signal is procurement: long-lead equipment orders are already being placed against the €300 million commitment, and the Mardyck Phase 2 commissioning in spring 2027 will pull through rolls, Pickling-line consumables, and electrical steel processing services. For competitors, the signal is that ArcelorMittal is building a French electrical steel franchise while hedging its green-steelmaking exposure down to a single furnace.

References

• ArcelorMittal France, ArcelorMittal confirme la construction d'un four électrique à Dunkerque (10 Feb 2026), via concertation-amf-decarbonation.fr
• Registre dématérialisé, Présentation de la consultation — Grande-Synthe environmental authorisation and building permit, 15 Oct 2026–15 Jan 2027
• Sénat, Séance du 25 février 2026 (Sen. Thomas Dossus on the re-scoped Dunkirk project)
• ArcelorMittal France, A Mardyck, ArcelorMittal France s'apprête à lancer une nouvelle filière de production d'aciers électriques (2 Dec 2025)
• ArcelorMittal Europe, ArcelorMittal prepares to launch new electrical steels production line in France (2025)
• SteelOrbis, ArcelorMittal commits €300 million to Dunkirk steel project, targets spring 2027 tests for new line in Mardyck (15 Oct 2026)