EV/EBITDA Through the Cycle: Valuing Steelmakers

The Trough-Year Multiple

A last-twelve-months EV/EBITDA multiple on a steelmaker tells you where the cycle is, not what the company is worth. Take a hypothetical steelmaker with a $10B enterprise value that earns $30M of EBITDA in a trough year: its trailing multiple is over 330x. The filed record shows such years happen. For FY2025 Cleveland-Cliffs reported $18.6B of revenue but adjusted EBITDA of just $37M, about $2 per ton on 16.2M net tons shipped.

A multiple in the hundreds is obviously absurd as a valuation statement. Nothing about the asset base changed in 2025. The blast furnaces, the pellet plants, and the automotive contracts are all still there. What changed is the steel spread, which collapsed to the point where a year’s production generated almost no EBITDA. Nobody pays hundreds of years of trough earnings for a steel business; the price is for the assets through the trough.

The same year makes the point across companies. Filed FY2025 EBITDA was $6,541M at ArcelorMittal, $4,174M at Nucor and $37M at Cliffs. Trailing multiples struck on those three denominators would differ by a factor of a hundred or more, and the gap would say nothing about which business is cheaper. It would say the LTM denominator is broken.

The Denominator Is the Cycle

Steelmaking is a spread business, and the spread is violently cyclical. A steelmaker’s EBITDA is roughly the steel spread minus conversion cost, multiplied by shipments. Conversion costs and volumes move slowly; the spread does not. So EBITDA swings from feast to roughly zero across a cycle while the underlying asset base barely moves.

The FY2025 numbers span that whole range. Nucor earned about $210 per ton of EBITDA on its 19.8M external tons (around $165 on total tons including internal transfers), ArcelorMittal $121 per tonne, and Cliffs about $2 per ton. Same year, same industry, the per-tonne ladder running from healthy to break-even. Put a trailing multiple on any of these and you are valuing one point on a sine wave.

This cuts both ways, and the peak is the more dangerous end. At the top of the cycle a steelmaker prints enormous EBITDA and the multiple collapses under it: an enterprise worth five times its mid-cycle earnings screens at 2.5x in a peak year that doubles them. That looks like a bargain right until the spread mean-reverts. Misread a trough multiple and you pass on assets that still earn through the cycle; misread a peak multiple and you overpay right before EBITDA rolls off.

Build the Normalised Number Instead

The fix is to replace the LTM denominator with a normalised, mid-cycle EBITDA and value the enterprise against that. The build has three inputs: a mid-cycle spread, a conversion cost, and shipments.

Our planning spread is $525 per short ton: mid-cycle US Midwest hot-rolled coil (HRC) at $900, minus mid-cycle shredded scrap at $375. HRC is the flat sheet steel that goes on to become cars, appliances and pipe; a short ton is 2,000 lb, the unit US steel trades in, about 9% lighter than a metric tonne. These are planning assumptions, anchored to data: $900 HRC matches the 2022-2025 average of monthly CRU assessments (~$901/st). Screen spreads can run far above planning when tariffs or a tight scrap market lift HRC without a matching scrap move; that is exactly the kind of number normalisation exists to exclude.

From there, for an electric arc furnace (EAF) producer, the kind that melts scrap:

1. Mid-cycle spread: $525/st.
2. Less conversion cost plus overhead: $350/st (our worked-example assumption; energy, labour, electrodes, alloys, and corporate costs).
3. Mid-cycle EBITDA per ton: $175/st.
4. Times shipments: at 10.0M st/yr, normalised EBITDA is $1,750M.

The $175/st output passes the sanity check against the filed FY2025 points: it sits between Nucor’s $210 per short ton and ArcelorMittal’s $121 per metric tonne, which is about $110 on the same short-ton basis. Apply a 5.0x multiple and the worked example carries an enterprise value of $8.75B; with net debt at 1.0x mid-cycle EBITDA ($1.75B), the implied equity value is $7.0B. None of those outputs move when the spot spread does, which is the point.

For an integrated producer, one smelting iron ore in a blast furnace instead of melting scrap, the spread input changes (HRC minus an iron ore and coking coal basket) but the architecture is identical: mid-cycle spread, minus conversion cost, times tonnes.

The Band: Roughly 4-6x, and Why It Is Low

On normalised EBITDA, our working band for steelmakers is roughly 4-6x, with 5.0x as the central anchor. It is an illustrative convention. No exchange, index provider, or standard-setter publishes an authoritative through-cycle steel multiple, so any band you see, including this one, is somebody’s judgement about where diversified steel franchises clear across a full cycle.

Why so far below the broad market? Because steel EBITDA is not free cash flow, or anywhere close. The capital intensity of keeping a steelmaker running is brutal, and ArcelorMittal’s FY2025 disclosure is the cleanest exhibit: $4.3B of total capex, of which $2.9B was sustaining (the company’s “maintenance/normative” bucket). Spread over 54.0 Mt shipped, that is roughly $54 per tonne of sustaining capex against $121 per tonne of EBITDA. So about 44% of every EBITDA dollar went on holding volumes flat, before growth spend, decarbonisation spend, interest or tax. The capex barely moves with the cycle; the share of EBITDA it eats moves a lot. At the $175 per ton of mid-cycle earnings built above, that same $54 charge is nearer 30%. Either way a multiple on EBITDA has to clear the drag, which is why the multiple is low.

Within the band, position follows quality, and quality has identifiable sources: the production route, the market sold into, leverage and customer concentration. The filed FY2025 figures show how far apart those can sit. Nucor runs EAFs for 100% of its steelmaking production, a flexible, lower-fixed-cost route that flexes down in a trough. It sells into the tariff-sheltered US market, carried net debt at about 1.1x FY2025 EBITDA against ArcelorMittal’s 1.2x, while Cliffs carried $7.2B of it against $37M, and earned $210/ton in a year the integrated US peer earned $2. When you pick a point inside the 4-6x band, that is the kind of evidence that should move you toward one end or the other.

Where the Band Does Not Apply

The 4-6x band is for steelmakers, full stop. The temptation is to stretch it down the value chain to the suppliers: iron ore, and the metallurgical coal that a blast furnace burns as coke. They sit in the same screens and the same sector ETFs. Resist it.

Vale, Fortescue, and Warrior Met are not conversion businesses earning a spread; they are price-taking resource extractors whose economics are set by cost-curve position, reserve life, grade, and jurisdiction. An iron ore producer whose cash cost of mining and shipping a tonne sits near the bottom of the global curve has an earnings profile, a margin structure, and a depletion problem that a steel converter simply does not have, and the market values it accordingly, off the cost curve and asset-level cash flow. Hand an iron ore miner a steelmaker’s 5.0x and you will be wrong in one direction; hand an integrated steelmaker an iron ore miner’s framework and you will be wrong in the other.

The same discipline applies inside diversified groups. ArcelorMittal owns captive iron ore (48.8 Mt of production, 72% self-sufficiency), and that mining stream deserves its own lens in a sum-of-parts before the steel band goes on the rest.

The normalisation framework, the full worked examples for both EAF and integrated routes, and the cost-curve approach for the iron ore and metallurgical coal names are built out in the Steel & Bulk Commodities Sector Primer, alongside the Excel model that computes all of it from one price deck.