Why Is Clean Steel So Hard to Make?

Sixty years of optimization

Worldsteel reports industry energy intensity falling from roughly 50 GJ/t in the 1960s to 20.95 GJ/t in 2024, a 58% decline. Larger furnaces, better burden preparation, oxygen enrichment, top-gas recovery, continuous casting, process control, and heat integration extracted enormous gains.

That maturity is precisely the problem. Incremental efficiency is valuable but cannot remove emissions inherent to carbon reduction. In worldsteel's 2024 route data, BF–BOF averaged 23.88 GJ and 2.34 tonnes of CO₂ per tonne of crude steel. Its coke ovens, sinter plant, blast furnace, BOF, gas network, and power recovery are physically and financially coupled.

The incumbent is not one furnace waiting for a clean replacement. It is a site-wide metabolism built around carbon.

Scrap has memory

Old steel carries the alloying and contamination history of cars, buildings, appliances, coatings, motors, and wiring. Copper and tin are difficult to remove once dissolved. Worldsteel notes that even small residual amounts can cause surface cracking during hot rolling, forming, and stamping in demanding flat products.

The answer is not “steel cannot be recycled.” It is information and separation: dismantle copper-rich parts, sort alloys, track chemistry, blend scrap recipes, and dilute remaining residuals with low-residual DRI or pig iron. DOE is funding AI-assisted sorting and selective copper-removal work because clean scrap is becoming a feedstock class, not generic waste.

The stoichiometric floor is small enough to state

Three hydrogen molecules remove the three oxygen atoms in hematite. By molar mass, 6.048 grams of hydrogen produce 111.69 grams of iron: 54.2 kg H₂ per tonne of pure iron. IEA uses 55 kg per tonne of DRI and adds 5% when translating DRI to steel. DOE process studies span wider values depending on route, recycle, metallization, and system boundary.

The mass is modest; the electricity behind it is not. At 52 kWh per kilogram, 58 kg of electrolytic hydrogen embodies about 3.0 MWh before EAF melting, auxiliaries, casting, or rolling. One million tonnes per year becomes a continuous industrial power and hydrogen infrastructure problem.

Hydrogen price passes almost directly into steel

At 58 kg/t, a $2/kg delivered hydrogen input costs $116/t steel; $4/kg costs $232; $7/kg costs $406. That sensitivity is why project announcements without hydrogen supply, electricity price, utilization, and offtake are incomplete evidence.

IEA's 2025 Breakthrough Agenda estimates early commercial plants using 100% hydrogen blends at 50–140% higher cost than BF–BOF, depending on region. Meanwhile, its 2026 Global Hydrogen Review says the low-emissions hydrogen pipeline has shrunk as projects were delayed or cancelled. Steel competes with fertilizer, refining, fuels, and other buyers for the first bankable molecules.

Gangue moves downstream

Silica and alumina in the pellet accompany sponge iron into the EAF. There they require flux, create more slag, consume heat, reduce yield, and can limit productivity. IEA therefore describes H₂-DRI as requiring high-grade ore. Industry convention commonly places DR grade around 67% Fe with low gangue, a small subset of the seaborne ore market.

Beneficiation can upgrade lower-grade ores, but finer grinding and separation consume energy and can lose iron. DOE's current R&D portfolio includes work targeting 25% lower energy for DR-grade pellet production and 3–5% higher iron recovery. “More high-grade ore” is not a free geological assumption; it is another processing industry.

The category contains materially different machines. Molten oxide electrolysis operates at high temperature and can produce liquid metal while oxygen evolves at an inert anode. Aqueous and alkaline routes operate colder and can produce solid iron for later melting. Their ore preparation, impurity tolerance, current density, electrodes, and products differ.

DOE describes molten oxide electrolysis as being scaled from laboratory toward industrial pilot operation. Its 2026 research selections still fund the practical questions around electrolytic iron: how pure iron melts in an EAF, how it changes productivity, and whether it can dilute copper-contaminated scrap. This is promising evidence of a route being engineered, not evidence of a bankable global intensity or cost curve.

The maturity ladder matters

1. Operating

   Scrap–EAF and gas DRI–EAF

   Commercial routes with measured global energy and emissions data.

2. Pilot validated

   Hydrogen reduction

   HYBRIT has produced more than 5,000 t of H₂-reduced iron at its Luleå pilot (Measured), with pellet, storage, and EAF melting trials through 2024; the demonstration aims at 1.2 Mt/yr, about 25% of Swedish steel (Target).

3. FID / construction

   First commercial hydrogen-capable plants

   Project evidence must include financing, construction, hydrogen supply, ore, and offtake, not only a target start date.

4. Demonstration

   Iron electrolysis

   Scale-up is testing cell materials, feedstocks, operating life, and integration with steelmaking.

Who is building what

Hydrogen direct reduction (H-DRI), molten oxide electrolysis, and electric arc furnace scrap conversion eliminate coal from primary metallurgy. Search the record, or filter by reduction technology.

Announced green steel production milestones often reflect short trial campaigns with trucked hydrogen; commercial cost depends on multi-gigawatt continuous renewable power contracts.

China and India set the size of the transition

Worldsteel records 960.8 Mt of crude steel from China and 164.9 Mt from India in 2025, against 1,848.9 Mt globally. China’s output fell 4.4% while India’s rose 10.4% from 2024. Together they made about 61% of the world total, so route changes and new capacity there dominate the global curve. Growing scrap availability can expand electric-arc-furnace output as earlier steel stock reaches end of life, but recycled steel cannot instantly replace primary iron while buildings and vehicles still hold metal.

The EU carbon border adjustment entered its definitive regime on 1 January 2026 for covered imports including iron and steel, with a gradual adjustment as free ETS allowances phase out. This changes the cost of carbon-intensive imports, but does not establish that a specific hydrogen-DRI project has reached commercial parity. That comparison requires a delivered hydrogen price, power price, DR-grade pellet price, scrap quality and plant utilization.

A product-level premium can be smaller than the steel premium: if a product uses one tonne of steel, a hypothetical $200/t steel premium adds $200 to that product before markups, redesign or tax. This is arithmetic, not a current steel contract price. The cost at a finished car or building should be reported separately from the factory-gate percentage premium.

Route mix and carbon policy move the investment window

Worldsteel’s 2026 figures put China’s 2025 output at 960.8 Mt with 10.6% made in electric furnaces, and India’s at 164.9 Mt with 57.7% in electric furnaces. Furnace share is not the same as a low-emission share: electricity mix, scrap quality and, in India, coal-based direct-reduced iron inputs determine emissions. Global progress cannot be inferred from European hydrogen projects alone.

The EU Carbon Border Adjustment Mechanism entered its definitive regime on 1 January 2026, alongside the gradual phase-out of EU ETS free allowances. A break-even carbon price for hydrogen DRI is site-specific: it equals the extra cost per tonne of steel divided by avoided tonnes of CO₂ per tonne, with ore/pellet quality, electricity, hydrogen, scrap and utilization held explicit. A 30% steel premium, for illustration, adds $300 to a product containing one tonne of steel bought at $1,000/t; it is not a measured car or building premium. Published project offtake prices, DR-grade pellet supply and furnace relining dates are too sparse for a defensible global series, so project announcements remain targets until output and cost are disclosed.