Steel Without Coal

Coal can leave ironmaking. The oxygen still has to leave the ore—and every replacement route creates a new scarcity somewhere else.

Last updated September 2026
Figure 1 · The industry baseline

Steel begins by removing oxygen

Iron ore is already oxidized. Primary steelmaking must spend chemical free energy to reverse that reaction, then melt, refine, cast, and roll the metal. The climate problem is concentrated in the first transformation.

69.4%Share of world crude steel made in oxygen converters in 2025, overwhelmingly the blast-furnace–BOF route.
30.3%Share made in electric furnaces in 2025, using mixtures of scrap and ore-based metallics.
2.34 tCO₂/tWorldsteel's 2024 BF–BOF route average, versus 0.69 for scrap–EAF and 1.47 for DRI–EAF.

Process shares come from World Steel in Figures 2026. Route intensities are worldsteel member data using 2024 activity and a boundary covering scopes 1, 2, and some scope 3. They are route averages, not physical constants.

The answer in one paragraph

The blast furnace is an extraordinary optimized reactor, but its carbon is both fuel and chemistry. Scrap–EAF avoids repeating reduction and is therefore the first low-emission route wherever enough clean scrap exists. New primary iron still needs ore. Hydrogen DRI swaps carbon monoxide for hydrogen but inherits direct reduction's demand for high-grade pellets and adds a vast requirement for low-cost clean electricity. Electrolysis may eventually accept a different feedstock envelope and remove the hydrogen intermediate, but remains less mature. Decarbonizing steel is not a furnace substitution. It is a reordering of ore, scrap, reductant, power, quality control, and capital.

  • Steel's 2024 global average was 20.95 GJ and 1.92 tCO₂ per tonne of crude steel; averages barely moved because the route mix barely moved.
  • Scrap–EAF uses less than half the route-average energy of BF–BOF, but demanding sheet grades require control of copper, tin, and other residuals.
  • The stoichiometric hydrogen floor is about 54 kg per tonne of pure iron. Real plants need margin, heat, gas recycle, and downstream melting.
  • Hydrogen changes the reducing agent, not the need to beneficiate ore, make pellets, melt iron, refine chemistry, cast, and roll.
Figure 2 · The route map

The furnace is only one layer

Each route pays for oxygen removal, melting, impurities, and feedstock quality in a different place.

Global incumbent

Blast furnace–BOF

Reported CO₂
2.34 tCO₂/t
Reported energy
23.88 GJ/t
Metallic feed
Iron ore + coke + ~10% scrap
Reducing agent
Carbon and carbon monoxide

Binding constraint: Coke is reductant, heat source, and structural support; a mature integrated stack is difficult to unwind.

Worldsteel's 2024 route averages are shown for the three commercial incumbents and use a methodology spanning scopes 1, 2, and some scope 3. Hydrogen DRI and electrolysis are described separately because no comparable global operating average exists.
Part I: The mature incumbent

The blast furnace is a chemical ecosystem

Coke does at least three jobs: it makes reducing gas, supplies heat, and supports a permeable burden while tonnes of material descend through the furnace.

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.
01

Mine and beneficiate ore

Rock is crushed and separated to raise iron content. Every percentage point of gangue becomes mass that must be heated, moved, fluxed, and removed as slag.

Measure
Fe grade, silica, alumina, phosphorus
Failure boundary
DRI needs a narrower, cleaner feed than the bulk ore market
Where the frontier moves

Finer grinding, flotation, magnetic separation, and routes tolerant of lower grades.

02

Agglomerate the feed

Fine concentrate becomes sinter or pellets strong enough to survive transport and the reduction reactor without choking gas flow.

Measure
Pellet strength, reducibility, gangue, yield
Failure boundary
Extra processing adds energy, capital, and iron loss
Where the frontier moves

Lower-carbon pelletizing and higher-yield DR-grade products.

03

Remove oxygen

A reductant binds the oxygen in iron oxide. Carbon leaves as CO₂ or CO; hydrogen leaves as water; electrolysis moves oxygen ions with electrons.

Measure
Metallization, reductant use, reduction rate
Failure boundary
Incomplete reduction carries oxygen into melting
Where the frontier moves

Hydrogen shafts, fluidized beds, and electrochemical cells.

04

Melt and refine

An EAF or BOF melts metallic feed, adjusts carbon, removes phosphorus and sulfur, and sets temperature for casting.

Measure
kWh/t, tap-to-tap time, slag mass, yield
Failure boundary
Low-carbon DRI contains little chemical energy and can raise EAF electricity
Where the frontier moves

Hot charging, preheating, oxygen control, and better slag practice.

05

Control residuals

Copper and tin do not oxidize away conveniently once dissolved in steel. Sorting before the furnace and dilution with virgin iron protect demanding grades.

Measure
Cu, Sn, Ni, Cr by grade
Failure boundary
Surface cracking and loss of flat-product quality
Where the frontier moves

Sensor sorting, design for disassembly, and selective copper removal.

06

Cast, roll, and finish

Liquid chemistry becomes slabs, billets, sheet, rail, or bar. Reheating and rolling still require energy after low-carbon iron is made.

Measure
Yield, thickness tolerance, defects, GJ/t
Failure boundary
A low-emission melt that misses specification becomes high-cost scrap
Where the frontier moves

Near-net-shape casting, electrified heat, recovery, and digital control.

Part II: Scrap is the first alternative feedstock

The cheapest reduction step is the one already paid for

Steel scrap contains metallic iron. An EAF can remelt it without repeating the ore-to-iron reaction, which is why the route's energy and emissions are structurally lower.

Recycling changes the route average

9.84 GJ/tWorldsteel 2024 scrap–EAF energy intensity.
0.69 tCO₂/tWorldsteel 2024 scrap–EAF CO₂ intensity.
>70%Scrap share used to define the reported route.

These are not zeroes. EAFs consume electricity, oxygen, carbon, electrodes, flux, and often DRI or pig iron. Their emissions depend heavily on power supply and charge mix.

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.

Part III: Hydrogen changes the reductant

Oxygen leaves as water

In a hydrogen shaft, hot H₂ diffuses into porous pellets, removes oxygen, and exits as water vapour. The iron remains solid sponge iron until the EAF.

Fe₂O₃ + 3H₂2Fe + 3H₂O

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.

Figure 3 · Interactive input model

What does hydrogen expose the steelmaker to?

The model isolates three route-sensitive inputs per tonne of crude steel: hydrogen, non-electrolyser electricity, and the premium for direct-reduction-grade ore. It is not a total steel cost.

Reference case$312/t steel

Route-sensitive input burden

Hydrogen
$232/t
EAF and process power
$40/t
Ore premium
$40/t
Electricity demand incl. H₂
3.82 MWh/t
Power-linked emissions
0.19 tCO₂/t

The stoichiometric minimum for Fe₂O₃ + 3H₂ → 2Fe + 3H₂O is 54.1 kg H₂ per tonne of pure iron. The default 58 kg/t steel follows IEA's 55 kg/t DRI assumption plus 5% conversion allowance. Electrolyser electricity is represented at 52 kWh/kg H₂.

Calculation and boundaries

Exposed input burden = hydrogen use × delivered hydrogen price + EAF/process electricity × power price + DR-grade ore premium. Electricity demand adds illustrative electrolyser consumption to process electricity; emissions multiply that total by the selected power carbon intensity.

Excluded: base ore value, scrap, alloys, electrodes, flux, labour, oxygen, capital recovery, utilization, hydrogen storage, transport, casting, rolling, taxes, and finance. A delivered hydrogen price already contains its production electricity cost, so that electricity is not charged a second time; it is shown only for infrastructure and emissions scale.

At 58 kg per tonne, each $1/kg movement in delivered hydrogen changes steel's input cost by about $58/t before capital and other operating effects. Ore quality and plant utilization can be equally decisive.

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.

Part IV: The ore-grade bottleneck

A shaft furnace does not have a slagging blast furnace beneath it

Direct reduction keeps iron below its melting point. That avoids the blast furnace, but it also means unwanted minerals are not separated in a liquid ironmaking step.

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.

RouteMetallic feedReductantMaturityAdvantageBinding constraint
Blast furnace–BOFOre, coke, sinter/pellets, some scrapC/COCommercial incumbentOre flexibility and integrated energy recoveryHigh process CO₂; long-lived capital
Scrap–EAFSorted scrap; DRI or pig iron for dilutionReduction inherited from first lifeCommercial incumbentLowest current route-average energy and CO₂Clean scrap quantity and residuals
Gas DRI–EAFDR-grade pellets, natural gas, scrapH₂ + COCommercial incumbentAvoids coke ovens and blast furnaceFossil gas and high-grade ore
H₂ DRI–EAFDR-grade pellets, hydrogen, scrapH₂Pilot / early commercialWater instead of CO₂ at reductionHydrogen cost, clean power, pellets
Iron electrolysisRoute-specific ore or oxide feedElectronsPilot / demonstrationPotentially bypasses reductant productionCell materials, scale, feed prep, cost
Part V: Can electrolysis change the feedstock equation?

Use electrons without making hydrogen first

Electrolytic iron routes apply current to an iron-oxide feed and produce iron at a cathode, potentially removing the intermediate electrolyser and hydrogen network.

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.

First plants: where the projects stand in 2026

€6.5BStegra, Boden: funded and ramping from 2025–26 as the first large-scale green steel plant (Measured); 5 Mt/yr by 2030 is a Target, not yet a nameplate.
5,000+ tHYBRIT H₂-reduced iron produced at the Luleå pilot (Measured). The demonstration phase targets 1.2 Mt/yr, about 25% of Swedish steel (Target), with a 600 MW electrolyser planned at Gällivare.
+60 MtAnnounced DRI additions by 2030 lift output to 185–190 Mt/yr, about 13% of BF+DRI ironmaking (Projected). Roughly 80 Mt of expansions are announced: Europe/Russia 34, MENA 32, Asia/other 14 (Projected, announced capacity).

Other measured and targeted steps: JFE Steel produced its first green-hydrogen DRI prototype of about 1 t between June 29 and July 1, 2026, on a 15 kg/hr NEDO-funded pilot at Chiba (Measured pilot); POSCO targets its HyREX pilot at 300 kt/yr in 2028, with the site approved in spring 2026 (Target); Salzgitter's SALCOS programme targets up to 2 Mt/yr of green steel in 2027 (Target). The constraint signal runs both ways: ArcelorMittal has delayed final investment decisions on European DRI despite €3.5B in subsidies (Measured). The mechanism is unchanged—H₂ DRI plus EAF, with the stoichiometric hydrogen floor below. What now binds is feedstock and capital: DR-grade pellet supply (LKAB's upgrade estimated at $4.7B), electrolyser power, and capital discipline measured in delayed FIDs.

Part VI: Asset turnover is a physical fact

A blast furnace relining is a climate decision

Integrated mills are long-lived, site-specific systems. Their replacement cadence is set by refractory campaigns, downstream equipment, local jobs, logistics, contracts, and capital—not laboratory learning rates.

DR-grade ore stays scarce

If beneficiation, pelletizing, and premiums rise faster than hydrogen-route learning, the nominally clean shaft becomes feedstock-constrained. Planned supply responses such as LKAB's DR-grade pellet expansion must actually scale.

First plants miss commercial yield

If Stegra has not reached commercial yield by 2027, or demonstration projects slip repeatedly, the learning mechanism never engages and every later target inherits the delay.

The green premium stays above 30%

If H₂-DRI steel still carries more than a 30% price premium in 2030, buyers keep specifying blast-furnace material and announced capacity runs below plan.

Hydrogen remains expensive

At roughly 58 kg/t steel, each additional $1/kg adds about $58/t before capital, storage, and integration.

Clean power is not additional

Hydrogen and EAFs move emissions onto electricity. A high-carbon grid can erase much of the route advantage.

Scrap quality degrades

More obsolete scrap without better sorting raises copper and tin, forcing more virgin iron into the EAF charge.

Plants run below design utilization

Capital-intensive shafts, electrolysers, EAFs, and casting lines need coordinated high utilization. Variable hydrogen supply can make cheap energy expensive steel.

Existing assets outlive the transition

A technically superior process cannot cut fleet emissions quickly if blast furnaces are relined and operate for another investment cycle.

The optimistic view, with conditions

Steel becomes a portfolio of metallics

The likely transition is not one universal replacement furnace. It is more scrap, better sorted scrap, gas DRI shifting toward hydrogen, traded HBI from resource-rich regions, electrolytic iron as a clean diluent, and selective retrofits while integrated assets turn over.

Now

Use every clean tonne twice

Collect, sort, identify, and blend scrap while lowering EAF power emissions.

First commercial wave

Build complete H₂-DRI systems

Contract power, hydrogen, pellets, storage, EAF capacity, and offtake as one utilization problem.

Next architecture

Widen the ore envelope

Smelting steps and electrolysis compete to tolerate lower-grade feeds without recreating the carbon stack.

What low-carbon steel actually needs

  1. Low-residual scrapCollection and chemistry-preserving sorting, especially for demanding flat products.
  2. Ore-based metallicsDRI, HBI, pig iron, or electrolytic iron to meet growth and dilute residuals.
  3. Clean firm energyElectricity and hydrogen delivered at industrial utilization, not annual-average abundance.
  4. Product qualificationAutomakers, builders, and machinery customers accepting material from new routes.
  5. Turnover financeContracts and policy that justify retiring a working carbon-based asset before its technical end.

Sources, method, and boundaries

Route intensities use worldsteel's published averages and boundaries. Hydrogen stoichiometry is derived from molar masses; system use comes from IEA and DOE assumptions. Announced projects are not counted as operating. The interactive model exposes selected input costs only and must not be read as a steel-price forecast.

BF–BOF
Blast furnace ironmaking followed by basic oxygen steelmaking.
DRI / HBI
Solid-state direct reduced iron; hot briquetted iron is compacted for storage and trade.
Residual element
A metal such as copper or tin that remains dissolved and is difficult to remove during refining.
FID
Final investment decision: stronger than an announcement, but not proof of operation.