Cleveland-Cliffs (CLF)

From Pellet Plant to Automotive Contract

Cleveland-Cliffs is the most vertically integrated steelmaker in the United States. It mines its own iron ore, pelletises it, and feeds those pellets into its own blast furnaces and basic oxygen furnaces, internally sourcing the vast majority of its pellet requirements. On top of that BF-BOF core sit a hot-briquetted iron plant in Toledo, some electric arc furnace (EAF) capacity at Butler, and the Canadian mills acquired with Stelco. The customer book is equally distinctive: direct automotive sales were $5.0 billion in FY2025, 28% of steelmaking revenue, making Cliffs the domestic incumbent for exposed automotive sheet, the visible outer panels that imported coil struggles to serve.

Integration cuts both ways, and FY2025 showed the sharp edge. Owning the ore means Cliffs never pays the seaborne price for its main input; an integrated mill buying at market would need roughly 1.37 tonnes of iron ore and 0.78 tonnes of coking coal per tonne of crude steel on worldsteel averages. But it also means the cost base is largely fixed. Mines, pellet plants, and blast furnaces run whether or not the steel price cooperates, and a blast furnace can't simply be switched off for a slow quarter. An EAF producer's biggest cost, scrap, tends to move with the steel price and cushions the margin on the way down. Cliffs gets no such cushion. When the spread compresses, almost all of it lands on EBITDA.

FY2025: Operating Leverage at the Bottom

The FY2025 numbers are the cleanest demonstration of steel operating leverage among the US producers. Cliffs shipped 16.229 million net tons and booked $18.6 billion of revenue, yet adjusted EBITDA came to just $37 million. That is about $2 per ton (the precise arithmetic is $2.28). The GAAP result was a net loss of $1,428 million, or $2.91 per diluted share.

Now put that next to the peers operating in the same spread environment. Nucor earned roughly $210 per ton of EBITDA on external shipments in FY2025 (about $165 on total internal-plus-external tons), and ArcelorMittal earned $121 per metric tonne across its global book, about $110 on the same short-ton basis. Same year, same US tariff regime for Nucor, broadly similar steel prices. The gap is not about who sells steel for more; it is about what happens to a fixed-cost integrated producer when the margin over its cost base thins out. Our EBITDA per tonne guide walks through the full ladder and the unit traps (Cliffs and Nucor report short tons, ArcelorMittal metric tonnes, and the external-vs-total shipments denominator moves Nucor's figure by $45/ton).

The flip side of $2 per ton is that recovery arithmetic gets dramatic quickly. On 16.2 million tons, every $50 per ton of margin Cliffs recaptures is roughly $810 million of incremental EBITDA. Against a market capitalisation of about $7.3 billion (9 June 2026 close), that is the whole high-beta case in one line.

What to Watch in the Financials

The spread, not the volume. Shipments scale whatever per-ton margin exists; they do not manufacture one when the spread is thin. Watch the right spread, though. The $525 per short ton the primer runs for a scrap-fed mill is hot-rolled coil (HRC, the flat sheet that goes on to become cars, appliances and pipe) minus shredded scrap, and purchased scrap is not what a Cliffs blast furnace runs on. Its spread is HRC over a basket of iron ore and coking coal, wider at planning prices, but the coke ovens, sinter plants and blast furnaces sitting between the basket and the coil consume far more of it than an EAF melt shop consumes of its own. Where realised prices settle against that cost base decides whether Cliffs prints hundreds of millions of EBITDA or another rounding error. The mechanics are in the steel spread guide.

Automotive volumes. A 28% direct automotive mix is a premium book in good times and a concentration problem in bad ones. Auto contracts are typically annual and lag exchange prints, which smooths revenue but also delays the benefit when market prices rally. Watch North American light-vehicle build rates and contract renewal pricing.

The balance sheet. Net debt was roughly $7.2 billion at 31 December 2025: $7,253 million of long-term debt against cash of just $57 million. That is approximately equal to the entire market capitalisation, so the equity is about half the capital structure. Equity holders own a leveraged slice of any recovery and absorb a leveraged share of any further deterioration.

The tariff wall. Section 232, the US national-security tariff on metal imports, has charged 50% on steel since June 2025, and an April 2026 proclamation widened it further by applying the duty to a product's full customs value rather than to its metal content alone. That is a large part of why US HRC has been trading well above the $900 per short ton the planning deck uses as mid-cycle. Cliffs' realised prices, and the spread above, rest partly on that policy floor.

The Valuation Problem: Multiples at the Trough

Trough EBITDA breaks the usual screens. Enterprise value is about $14.5 billion ($7.3 billion of equity plus $7.2 billion of net debt), so EV against FY2025 EBITDA of $37 million produces a multiple near 390x. Meaningless. The assets did not become worthless or infinitely expensive in twelve months; the denominator collapsed. The same failure infects the leverage screen: net debt over trough EBITDA suggests Cliffs can never repay anything, which is not a statement about the company, it is a statement about using a cyclical trough as a run-rate.

So normalise. Build a mid-cycle EBITDA from the planning spread and shipped volumes, then apply the roughly 4-6x band our through-cycle EV/EBITDA guide works to, which is a working convention rather than a published standard. Cliffs is that guide's worked example precisely because its FY2025 print makes the trailing-multiple lesson impossible to miss. Two judgement calls drive the output: which spread counts as mid-cycle in a tariff-supported US market, and how much of the normalised EBITDA the debt claims before equity sees any of it.

The Higher-Beta Recovery Trade

If the market spread holds above the planning assumption rather than mean-reverting, the operating leverage that produced $2 per ton at the trough works in reverse, and it works through a capital structure where equity is only half the enterprise. A given improvement in steel economics moves CLF's equity far more, in percentage terms, than it moves Nucor's, whose margins are already fat and whose balance sheet is conservative. That is what high beta means here: amplified exposure to the same variable, not a different thesis.

The same mechanism punishes the equity if spreads grind lower or the tariff regime softens. Nothing in the structure protects the downside; the fixed costs and the debt are still there. Whether amplified steel exposure fits a US steel-price thesis is a question about that thesis, not about Cliffs as a company.

Key Risks

Balance sheet through a prolonged trough. With $57 million of cash at year-end against $7.25 billion of long-term debt, Cliffs has little room for an extended stretch of near-zero EBITDA. Refinancing windows, maturity walls, and working-capital swings matter more for this equity than for any peer in the set.

Automotive concentration. A downturn in North American vehicle production hits roughly 28% of steel sales directly, and the high-margin exposed sheet within it hardest. EV transition shifts in steel content per vehicle add a slower-moving version of the same risk.

Policy dependence. The US price premium over landed imports is a policy choice, and policy choices reverse. A lower rate, a country deal, a quota arrangement or a fresh set of product exclusions would each compress the spread the entire recovery case depends on, with nothing changing inside Cliffs.

Fixed costs in a falling market. Idling a blast furnace is expensive, restarting one more so. If demand weakens, Cliffs faces the integrated producer's dilemma: run at a loss or pay to stop. EAF competitors can flex output at much lower cost, which is part of why the market pays them a structurally higher multiple.