EU Steel 'Melt and Pour' Proof Goes Live: 50% Duty Without Mill Certificate

The mechanism itself is small. One extra country field on the entry declaration, attached to the goods as a physical thing rather than to the transaction as a commercial thing. That difference is the entire point of melt-and-pour. An origin rule asks where the last substantial transformation happened, which tells you nothing about where the metal came from. A re-roller can take slab from one economy and turn out coil that carries its own flag. Ask about the furnace and you bypass the paperwork layer completely.

Regulation (EU) 2026/1384 took over from the safeguard on 1 July 2026. The shift from safeguard to standing quota instrument is more than housekeeping. A safeguard expires; a company can plan around an end date, defer investment, wait out two bad quarters. A standing regime with a review calendar does not expire, it gets revised. The instrument carries three dates that should already be in somebody's planning calendar: 30 September 2027, when tolerance for alternative evidence runs out; 30 June 2027, when the Commission has to report on extending the system to steel-containing finished goods; and 30 June 2028, when it has to consider whether melt-and-pour should replace origin as the basis on which quota is allocated.

Now do one subtraction that the press releases do not do for you. Of the 18,345,922 tonne headline, roughly 9.15 million tonnes is set aside for free-trade-agreement partners. Call it a reservation, an ear-mark, whatever you like — the arithmetic consequence is that everyone outside those agreements is competing for close to 9.2 million tonnes, not 18.3 million. About half of a number that most coverage treats as the whole.

Then look at the floating range. The Commission can move the total anywhere between 14.4 million and 22.2 million tonnes. If that total is tightened towards the floor while the reservation stays at 9.15 million in absolute terms, the unreserved pool does not shrink by the 21% the headline does; it falls from roughly 9.2 million to roughly 5.25 million, a drop of 43%. If instead the reservation scales with the total, the pool lands nearer 7.2 million. Nobody has told us which it is. So the honest planning range for a supplier without a free-trade agreement is somewhere between 5.25 and 9.2 million tonnes, and anybody budgeting against 18.3 million is budgeting against the wrong number by a factor that ends careers.

Why now, and why this instrument. The safeguard ran out on 1 July and something had to stand in its place — that part is calendar. The interesting part is why the replacement chose melting as its anchor instead of refining the origin rule once more. Origin-based quotas have a known weakness: they chase transformation, and transformation is mobile. Steel poured in one economy, shipped as slab and finished in another leaves the officer looking at a perfectly valid origin declaration with no way to see behind it. Melting is the one step in the chain that is expensive to move and cheap to record. Mills already log heat numbers for metallurgical reasons that have nothing to do with customs. Re-using that field is the cheapest available way to see through the rerouting of the past decade.

Put another way, this measure changes the unit of observation from the consignment to the heat. Every problem that follows comes out of that change, and none of them appear in a tariff line.

Who gets hurt, and by how much. The obvious layer first: importers of the 26 covered categories who clear into free circulation. Their exposure is binary and immediate — inside the quota at concessional treatment, or outside at 50%.

Then the layer that attracts the least sympathy and involves the most arithmetic: Chinese suppliers. The arrangement gives China 22 sub-categories of quota and specifically denies access to the unused quota of other origins. Under the old design that sort of residual draw-down was a familiar release valve; if one country did not fill its allocation, somebody else picked up slack. That valve is shut. An exporter whose sub-category fills out in week six of a quarter has nowhere to go until the next quarter opens, no matter how much paper sits unused somewhere else.

Traders who built their sourcing around melting in one place and finishing in another are barely visible today and probably the largest group eventually. Their commercial structure has not changed. Its visibility has.

Steel-containing downstream goods sit outside today's perimeter deliberately, though only until somebody finishes a report. Fasteners, wire, welded tube, machinery, white goods, structural components — goods containing steel are not covered now, and the Commission is due to report by 30 June 2027 specifically on whether they should be.

Then there is the group nobody budgets for anywhere: whoever has to obtain the document. Usually the importer, from a mill he has never met, through a trader who was never obliged to pass it on.

One operational detail deserves more attention than the tariff does. The quota is managed by quarter. A quarterly window turns this from an annual budget into a sequence of sprints. If your line runs hot early in the quarter, it can be exhausted before your vessel arrives and the rest of that quarter is spent at 50%. The quantity written down for the year tells you nothing at all about week seven.

Combine that with first-come allocation and you get the familiar rush. Cargo that was going to move anyway now moves earlier. Forward loading into the first weeks of each quarter becomes rational, which drains the window faster, which makes the pattern reinforce itself. Anyone planning January 2027 on the assumption that January behaves like a normal month will learn something uncomfortable in the second week of it.

On timing there are two clocks running, and confusing them is expensive. The duty clock started at 00:00 on 1 October and does not care when your goods were loaded or when the contract was signed. Anything released for free circulation from that minute is inside the new regime. Given Asia-Europe transit times, cargo booked early in September is clearing right now under terms negotiated in the spring, carrying documentation nobody discussed.

Running alongside it is the evidence clock, and that is the valuable one. There is tolerance until 30 September 2027 for evidence other than the Mill Test Certificate. That is nearly twelve months of slack for anyone whose supplier cannot produce a certificate today. It does not stop the 50% landing. It stops the document problem from being fatal for eleven months.

Now the part I was brought here for. Take one declaration and walk the chain: classification, then the data elements, then verification risk, then duty assessment, then release. Nothing here touches classification — your TARIC line is unchanged. The change lands squarely on the data elements and then ricochets into the duty assessment.

There are two failure modes, and they are not the same thing. The first is a missing element: you do not have the melt country, so you cannot assert the quota claim that depends on it. The outcome is boring and expensive, the claim fails and the rate defaults to the out-of-quota level. The second is a present but unsupported element: you assert a country and cannot back it. That one is worse, because it stops being arithmetic and becomes liability. A declared country that does not survive a check costs you the higher rate, interest on recovery, and something harder to price — a standing problem at that particular customs office that follows every declaration you file afterwards. My line on this has not moved in fifteen years: do not guess. Guessing at a country field to avoid a large duty is not clever, it is a larger liability bought on instalments.

The reason all this lands hard is granularity. A Mill Test Certificate is issued against a heat, and a heat is a particular batch out of a particular furnace. Commercial shipments do not respect that. One booking can span several heats. One vessel can carry product from two mills. One bill of lading almost never corresponds to one certificate. So the evidence has to be rebuilt at heat level and then mapped onto the declaration line, quantity for quantity.

That mapping is the actual work. Any half-competent broker can do it. Almost nobody has been doing it, because until 1 October nobody asked.

Now the two things the headlines have not touched, and the two that will cost somebody real money this quarter.

The structural one first: origin and melt-and-pour now point at two different countries for a large share of trade, and your declaration has to tell both stories at once, correctly, in the same breath. If steel was melted in China and processed in a third country that holds a free-trade agreement with the EU, your preferential origin claim may properly say that third country while your melt-and-pour element says China. Both are true. Both matter. And if the China sub-category covering that line has run dry for the quarter, the preference you validly earned does not save you from the 50%, because China cannot draw on anybody else's residual quota. Run it the other way and the same trap closes: metal melted inside a partner economy and finished somewhere neutral can fall outside the reserved pool, because that pool tracks the flag you claim rather than the furnace you used. One shipment, two country answers, two rulebooks, and neither of them knows the other exists.

The other one is a procurement problem wearing a customs costume. A great deal of the cargo clearing this month was bought in March or April. Those purchase orders say nothing about supplying a heat-level certificate, because in March or April nobody was asking. The seller is not being difficult — the seller was never bound. Now the goods are on the water or already in a yard, and the buyer needs a piece of paper that does not exist in anyone's file. That is the picture I expect on most desks this month, and it is expensive in a way that no amount of customs expertise repairs after the fact.

Where there is an exception, and there is one, it sits in the applicable precondition for this obligation attaching — a point Ingrid has set out in detail before — namely that the goods are released for free circulation. Steel arriving under inward processing, into customs warehousing, or onward for re-export is not standing in the same queue. If your cargo genuinely moves through one of those regimes, the melt declaration is not today's problem for that shipment, and settling that question first can take cargo off the emergency list.

Let me put arithmetic on the table with the assumptions stated, because everyone needs different numbers.

Assume one shipment of 500 tonnes of hot-rolled coil. Assume an invoice value of 400,000 US dollars, which works out at 800 per tonne. Assume it clears at a northern European port in October, the line is one of the 26 covered categories, and assume for the baseline that this line carries no anti-dumping or countervailing duty.

Case one: documents complete and quota available. Duty at the in-quota rate, which for this illustration I take as zero. The cost of the rule to this shipment is the staff hours to assemble the pack. Call it a few hundred dollars.

Case two: same shipment, the relevant China sub-category exhausted for the quarter. Fifty per cent of 400,000 is 200,000 US dollars, which is 400 dollars a tonne on top of what everyone thought the steel cost. Under the safeguard that ended in June the identical failure produced a 100,000 dollar bill. Same cargo, same paperwork gap, double the exposure.

Case three: same shipment, and now drop the assumption that no trade remedy applies, because on plenty of lines one does. The 50% does not absorb that duty, it stacks on it. So the real grid has two variables, not one, and they multiply rather than line up politely.

Scale it once, for somebody else's point of view. Assume 8,000 tonnes of covered product on one sailing at the same 800 dollars. That is 6.4 million dollars of value, with a theoretical 3.2 million riding on whether the heat numbers were kept.

The ratio that should drive behaviour, and rarely does. Assume you work on a 5% gross margin — stated as an assumption, because steel margins are nobody's business but yours. That is 20 dollars a tonne earned against 400 a tonne exposed: twenty to one. There is almost no amount of chasing, calling and paying a mill to re-issue a certificate that fails to clear that bar. People will not do it anyway, because they have not internalised the denominator.

Where this analysis is wrong, and what would flip it.

If you supply through a free-trade-agreement partner drawing on the reserved allocation, most of the above is noise for you. Your problem remains the old one: proving you are who you say you are.

If your clearing office accepts substitutes, the tolerance until 30 September 2027 changes everything for one population. For eleven months no certificate is a nuisance rather than a catastrophe. It becomes a catastrophe at midnight on 30 September 2027. The planning implication matters: the duty has no runway, the paperwork has nearly a year.

If the Commission moves in its 30 June 2027 report towards finished goods containing steel, the affected population stops being steel importers and becomes everybody. A machinery maker does not know his heat numbers; his supplier's supplier does, two tiers up, another language, no contractual hook. Same twenty-to-one ratio, far worse data.

If melt-and-pour becomes the allocation key itself after 30 June 2028, every country-specific quota including China's 22 sub-categories gets re-based on where metal was melted rather than where it now comes from. Some winners would become losers without changing a thing.

And the condition that quietly decides whether the twenty-to-one ratio is real at all: how your own member state's office handles heat-level evidence in practice. That depends on the office. Before committing real money to chasing documents, get that office's answer in writing.

So, what gets done, by whom, and by when.

For cargo already clearing, build the pack before entry is filed. Confirm the TARIC line, state the melt-and-pour country explicitly, attach whatever evidence you actually hold, then check two cross-references. Does the country named on the certificate match the country being declared? Do the heat numbers on that certificate cover the quantity being declared? Those two questions catch most of the trouble I expect to see.

Where the evidence cannot be assembled before arrival, make the decision openly instead of letting it happen by default. Declining the claim costs 200,000 dollars on our example shipment. Asserting it without support costs that amount plus recovery, plus interest, plus your standing at that office. Take the first one, then go fight the supplier.

For anything not yet loaded, change the purchase order this week. Four clauses: the seller delivers a heat-level Mill Test Certificate with the shipping documents; the certificate states the country of first melting and pouring; the seller warrants that statement; if the document is not delivered at least five working days before loading, the buyer may delay shipment without penalty. Then add the money clause — who carries out-of-quota duty if the sub-category is exhausted, written as a price adjustment rather than a claim. You cannot negotiate the allocation. You can negotiate who pays for losing it.

Then build the quarterly habit. Track fill per sub-category per quarter, set a trigger at 80%, and act on it: stop shipping into that quarter or move tonnage to an origin with headroom. Put the three dates where everybody can see them — 30 September 2027, 30 June 2027, 30 June 2028.

On alternatives and their traps. Moving the shipping country achieves nothing if melting stays where it was. Inward processing only helps if the goods really are re-exported, and it buys you a bond and a traceability obligation along the way. Asking the customer to accept delivery duty unpaid shifts cash rather than competitive position, because he will reprice against you at the next tender. And the move I expect to see most often — declaring your best guess at a melting country — is not an alternative at all. Documents may be tedious, but they are the only thing in this chain that never lies to your face.

• From this week, file no EU steel entry with a blank melt-and-pour field: confirm the TARIC line, state the country, and cross-check that the certificate's heat numbers cover the declared quantity before the declaration goes in.
• Amend every open steel purchase order by 31 October 2026 to require a heat-level Mill Test Certificate, a stated melt country, and delivery at least five working days before loading.
• Stand up a per-sub-category, per-quarter fill tracker with an 80% trigger, and move tonnage to an origin with quota headroom the same week that trigger is crossed.
• Book the out-of-quota exposure into the price of any shipment whose evidence is not in hand before arrival, rather than asserting an unsupported quota claim — 200,000 dollars on a 500-tonne, 400,000-dollar example.
• Put 30 September 2027 (alternative evidence expires), 30 June 2027 (finished-goods review) and 30 June 2028 (melt as quota key) into the compliance calendar this month and assign an owner to each.
• Add a price-adjustment clause to every Q4 and Q1 steel contract stating which party carries out-of-quota duty when the relevant sub-category has exhausted for the quarter.