Green Steel's Demand Problem, and How to Fix It

In Middletown, Ohio, a blast furnace that has been running for 73 years was supposed to be switched off. In March 2024 the Department of Energy picked Cleveland-Cliffs for up to $500 million to tear out the coal-fed ironmaking at its Middletown Works and put in a hydrogen-ready direct-reduced-iron plant with two electric furnaces. The plan would have cut the plant's emissions by roughly a million tonnes. On August 21, Vice President JD Vance and Energy Secretary Chris Wright stood at that same plant and announced the opposite. The $500 million will now help rebuild and extend the old coal furnace so it can run for another two decades, matched by $500 million of Cliffs' own money, with the rebuild finished in early 2030.

The reason the government gave is the part worth sitting with. In its own announcement, DOE said Cliffs had decided the original clean project no longer made business sense because customers were unwilling to pay a "green premium" for steel. Not that the technology failed. Not that hydrogen was unavailable, though Cliffs has said that too. The clean plant died because nobody wanted to buy what it would make.

That should unsettle anyone who assumed corporate climate targets would pull heavy industry into a cleaner shape on their own.

 

The premium is real, and it's tiny

 

The strange thing is that the premium is small where it actually lands. Green steel made through hydrogen and electric furnaces does cost more than coal-based steel, no argument there. The International Council on Clean Transportation puts a new hydrogen-based plant at about 34% dearer than simply relining an existing blast furnace. In Europe, buyers with sustainability commitments have been paying somewhere between €100 and €170 a tonne over the conventional price through late 2025 and into 2026, with a few long-term automotive contracts reported as high as €350. Those are big-sounding figures until you follow the steel into a finished thing. A car holds around a tonne of it. Analysts at Transition Asia put the added cost of green steel at roughly $200 per passenger car, against an average price near $28,000. That is well under one percent. A premium carmaker could absorb it and never touch the sticker.

 

Why nobody pays it anyway

 

So affordability is not the problem. The problem is that steel is a commodity, and green steel is molecule-for-molecule identical to the dirty kind.

The buyer who pays extra is competing against the buyer who doesn't, selling a product the customer cannot tell apart. In that setup the first mover is penalized rather than rewarded. Voluntary pledges bend the moment margins tighten. When European hot-rolled coil prices fell through 2024 and steelmakers' margins thinned, the assessed green premium dropped to €50–75 a tonne before recovering, because the premium was among the first things buyers stopped paying. Scope 3 targets are real commitments on paper. They are also the easiest thing in the world to defer for a year when the finance director is hunting for savings. This is the coordination failure at the center of the whole story, and it is exactly the kind of failure that markets, left alone, do not solve.

 

Europe is failing the same test

 

None of this arrived with the Trump administration. In June 2025 ArcelorMittal walked away from €1.3 billion in German state aid rather than build hydrogen-based plants in Bremen and Eisenhüttenstadt, saying plainly that CO2-reduced steel was not profitable and that German energy cost too much. Thyssenkrupp suspended the hydrogen tender for its Duisburg plant. Salzgitter pushed the later phases of its Salcos project back three years. SSAB delayed its Luleå conversion. By late 2025 roughly half the green-steel projects announced worldwide had slipped or been shelved. The pattern rhymes with Middletown down the line: governments were willing to fund the plant, and buyers were not willing to fund the product.

Here is the uncomfortable place to start if you want a fix. There is no voluntary version of this that works at scale. If a chemically identical product carries any premium at all, a commodity market will route around the firms that pay it. Good intentions do not survive contact with a procurement spreadsheet. What can work is a set of tools that either buys time for the best projects or changes the arithmetic so that clean stops being a premium in the first place.

 

Short term: what buys time

 

The strongest projects are staying alive by not touching the open market at all. Stegra, the Swedish company building a hydrogen-based plant at Boden, sold more than half of its first-phase output before construction finished, on binding multi-year contracts to Mercedes-Benz, Porsche, BMW, Volvo, Scania, and non-carmakers like IKEA and Microsoft. Those pre-sales are what made a roughly €7 billion project bankable, and they work for the reason the economics predict: a luxury automaker can bury $200 to $400 of premium in a $50,000 car without anyone noticing. The instruction for producers is blunt. Sell to the buyers whose product is expensive enough to hide the cost, and lock the contracts before breaking ground.

A second bridge is to unbundle the green attribute from the physical metal. Because green and grey steel are identical, you can sell the environmental benefit separately from the tonnes, the way renewable-energy certificates work. Stegra did exactly this with Thyssenkrupp Materials Services, which takes the physical steel but leaves the carbon benefit for Stegra to sell on to someone who wants it. Microsoft, which does not buy steel directly, instead pays its construction suppliers to use Stegra's steel and claims the attribute for its own targets. Book-and-claim lets scattered willingness-to-pay, a data-center operator here, a consumer-goods firm there, add up into real revenue for a producer who would otherwise be waiting on a single reluctant buyer. It is a genuinely useful bridge. It is also scaffolding. Done loosely it invites double-counting, where two companies claim the same clean tonne, and it will never generate enough demand to justify the tenth plant, let alone the hundredth. Treat it as a way to begin, not a place to end up.

The third short-term move is public money covering the gap outright. Germany's carbon contracts for difference, the Klimaschutzverträge, pay a producer the difference between clean and conventional production cost for up to fifteen years, with a clawback clause if clean production later becomes the cheaper option. The first round drew €5.3 billion in bids and signed fifteen companies in October 2024, and a second round of around €5 billion cleared European state-aid approval in 2026. It is a sensible way to stop a good plant from dying while the market matures. It is also, stated without dressing, the taxpayer paying the premium the customer refused. Defensible as a bridge. Indefensible as the permanent shape of things.

 

Long term: what changes the math

 

The permanent fixes are the ones that change the numbers, and there are two that matter.

The first is to make the dirty product carry the cost of its own carbon, so the clean producer stops being undercut. This is what the EU's carbon border adjustment mechanism is built to do. As its full phase begins in 2026, it charges the embedded emissions of imported steel at the same rate EU producers pay under the emissions trading system, adding something like $100 to $200 a tonne to conventional imports. That flips a green premium into a grey penalty. The company that cleaned up is no longer punished for it. The United States has no equivalent, and that absence is not a footnote. It is a large part of why Cliffs' spreadsheet came out the way it did. A carbon price that actually bites is the single most powerful lever available, and the fact that Washington has chosen not to pull it is the real cause of the Middletown reversal, more than any one grant decision.

The second permanent fix is to use the state's own buying power, and its rule-making power, to manufacture the demand the market will not. Public bodies buy around 14% of GDP in the EU. If public infrastructure, transport, and defense are required to use low-carbon steel, that builds a floor of guaranteed demand that depends on no single company's willingness to pay. The EU's Industrial Decarbonisation Accelerator Act, introduced in March 2026, begins writing low-carbon and "made in Europe" requirements into public procurement, alongside a low-carbon steel label and a wider overhaul of procurement rules due across the year. The cleaner and harder version is a mandatory ceiling on the carbon intensity of any steel sold into a market, lowered on a set schedule. Mandates get attacked as heavy-handed, and they are. They are also the one instrument that removes the free-rider problem completely, because when everyone has to pay, nobody is undercut for doing the right thing.

 

What Washington chose instead

 

It is worth noticing what the same Department of Energy did the day before Middletown. On August 20 it put another $500 million into critical-mineral processing, battery manufacturing, and recycling, framed entirely around supply-chain security and energy dominance. So Washington has not given up on industrial policy. It has decided which industries qualify. Batteries and critical minerals are strategic; green steel is a premium nobody asked for. That is a choice, and the consequences run long. Spending clean-industry money to reline a coal furnace does more than skip a decarbonization step. It ties Middletown to coal-based ironmaking into roughly 2050 and throws away the cheapest moment to switch, which was this one, with the furnace already due for a rebuild.

The Cleveland-Cliffs decision answered the question the entire green-industry thesis had been avoiding. If a clean plant with half a billion dollars of federal backing still cannot find customers, then heavy industry does not decarbonize because companies promise it will. It decarbonizes when the rules make the dirty version cost more, or when a buyer with margin to spare, or a government with a mandate, decides to pay. Offtake deals and certificates and contracts for difference are all ways of getting to that point, or ways of pretending you can get around it. Middletown picked pretending, and bought itself twenty more years of coal.