How to Analyse a Steel Stock (Spread, EBITDA/Tonne)

In its March 2026 results, JSW Steel reported a quarterly net profit of roughly ₹19,243 crore (Inve data, Q4 FY26) — a number so large it would make the stock look like a screaming bargain on earnings. It wasn't. Buried in that figure was about ₹18,607 crore of one-time other income for the year (Screener.in, FY26), and on a clean operating basis the company's EBITDA per tonne for the prior quarter had sunk to ₹8,700 — what management itself called a level "impacted by multi-year low steel prices" (JSW Steel Q3 FY26 concall). The headline profit told you the opposite of the truth. (Illustration of how to read the numbers, not a view on the stock.)

That gap is the whole game. A steel company does not earn money the way a software firm or a bank does. It takes iron ore and coking coal in at one price, turns it into a tonne of steel, and sells that tonne at another — and the entire economic engine is the difference between those two prices, multiplied by how many tonnes it can push out of an expensive, debt-funded plant. Net profit, P/E, even revenue growth can all mislead you, because they sit at the end of a chain whose first link is a global commodity price the company does not control. Read steel the way the people who run it do, and you watch a different set of numbers entirely.

This is how to analyse a steel stock the way a cyclical analyst does: the handful of per-tonne numbers that decide the outcome, where the important ones hide (often in an investor deck, not the income statement), how to value a business whose earnings swing 5x across a cycle, and the one mistake that has cost more steel investors money than any blow-up — buying the peak as if it were the new normal.

A boundary first: you will not forecast the global steel price, and neither can management. What you can do is read whether a company's spread, its plant, and its balance sheet are built to survive the down-leg and compound through the up-leg.

What actually drives the economics of a steel company?

Picture a bakery that buys flour at a price set in a global auction it can't influence, sells bread at a price set in another auction it can't influence, and has borrowed heavily to build an oven so large it loses money unless it runs nearly full. That is a steel mill. The flour is iron ore and coking coal; the bread is hot-rolled coil; the oven is a blast furnace that cost tens of thousands of crores and must be fed whether demand is there or not.

Three consequences fall out of that picture, and they govern everything.

The spread is the business, not the selling price. A steel company can report rising revenue while its economics collapse, if input costs rise faster than what it charges. So you never look at realisation alone — you look at realisation minus the cost of coking coal and iron ore per tonne. That difference, the spread, is where the profit lives.

Volume is leverage in both directions. Because a huge share of a mill's cost is fixed — the plant, the interest, the people — every extra tonne sold at full utilisation drops almost straight to EBITDA, and every tonne not sold in a downturn bleeds. This is operating leverage at its most violent. A 10% swing in volume can move profit far more than 10%.

Debt turns a cyclical into a survivor or a casualty. Plants are built with borrowed money. In the up-cycle, leverage magnifies returns; in the down-cycle, the interest bill keeps coming while the spread vanishes. Whether a steelmaker lives to see the next upturn is mostly a balance-sheet question, settled before the cycle even turns.

Hold those three — spread, volume, debt — and the metrics below stop being a list and become a single story.

The metrics that matter — and where they hide

Here is the uncomfortable part for anyone used to reading a P&L: the numbers that decide a steel investment are mostly not on the income statement. Realisation per tonne, EBITDA per tonne, the coking-coal cost movement, utilisation, captive ore share, value-added mix — these live in the investor presentation and the concall, expressed per tonne, and you have to go and get them. The income statement gives you sales and operating profit; it does not tell you whether you sold more tonnes or got a better price, and those two have completely different meanings.

Realisation (net sales realisation, NSR) per tonne

This is the price the company actually got for a tonne of steel after discounts — what an analyst calls NSR or ASP (average selling price). It matters because it is half the spread, and it tracks the global steel price with a lag. Where to find it: almost never on the income statement as a line item; it sits in the investor deck or gets quoted on the call. JSW Steel, for instance, flagged a "₹3,300/tonne realization improvement" into Q1 FY26 helped by safeguard duty and seasonal demand (JSW Steel Q1 FY26 concall). What "good" looks like is less about the level — which the market sets — and more about a company holding realisation while peers discount, which signals a stronger product mix or customer book.

The spread (coking coal and iron ore)

The single most important relationship in the sector, and the one beginners skip. Coking coal is largely imported and priced in dollars; iron ore can be bought or mined captively. When coal costs rise faster than steel prices, the spread compresses and profit evaporates even at flat volume. Where to find it: the deck and the call, always as a change — companies guide it quarter to quarter. Jindal Steel told analysts it expected "coking coal prices to increase by $20 to $25 per tonne sequentially" into Q1 FY27 (Jindal Steel Q4 FY26 concall); JSW guided a "USD15 to USD20" rise around the same window (JSW Steel Q3 FY26 concall). A $20/tonne coal increase on a tonne of steel that needs roughly 0.7–0.8 tonnes of coal is real margin, and it lands before any price hike can offset it. The defence against this is captive raw material — Jindal had "ramped it up to 45% in Q2 from our own captive mines" on iron ore (Jindal Steel Q2 FY26 concall), and a company mining its own ore simply doesn't feel the same swings as one buying everything on the market.

EBITDA per tonne

This is the spread after all operating costs, per tonne of steel — the cleanest single measure of how profitable each tonne actually is, stripped of volume and accounting noise. It is to a steelmaker what NIM is to a lender: the one number that, tracked over time, tells you whether the core economics are improving or rotting. Where to find it: the deck, quoted explicitly. The contrast across the cycle is stark. Jindal printed an adjusted EBITDA per tonne of ₹15,680 in Q1 FY26 (Jindal Steel Q1 FY26 concall) — and then ₹6,981 two quarters later in Q3 FY26 (Jindal Steel Q3 FY26 concall), the difference being a one-time blast-furnace start-up cost and a softer market; excluding the start-up cost, the underlying figure was ₹8,516 (Jindal Steel Q3 FY26 concall). JSW's India operations ran at roughly ₹10,701 in Q2 FY26 before sliding to ₹8,800 in Q3 (JSW Steel Q2 and Q3 FY26 concalls). Read EBITDA per tonne across six to eight quarters, not in one — a single quarter is a snapshot of where the cycle is, not of the company.

Volume and capacity

Volume is how many tonnes were sold; capacity is the ceiling; the gap between them, expanding or contracting, is the operating-leverage story in motion. Where to find it: sales volume in the deck (per quarter and full year), capacity in the strategy section. Jindal sold 8.68 million tonnes in FY26, up 9% YoY (Jindal Steel Q4 FY26 concall), and guided FY27 sales of "10.5 million to 11 million tonnes" against production of "11 million to 11.5 million tonnes" (Jindal Steel Q4 FY26 concall) — a step-change as new furnaces ramp. JSW targets 50 million tonnes of India capacity by FY31 (JSW Steel Q3 FY26 concall). The thing to watch is the funding of that growth and whether new capacity arrives into a strong or weak price environment — capacity commissioned into a downturn is a fixed cost with no spread to cover it.

Utilisation

The percentage of capacity actually running. It matters because of the fixed-cost problem: below a certain utilisation, a mill loses money on every tonne; above it, each extra tonne is almost pure margin. Where to find it: the deck, sometimes the call. APL Apollo ran "nearly 90%" on its 5-million-tonne capacity (APL Apollo Q2 FY26 concall) — high utilisation that lets fixed costs spread thin. Watch utilisation alongside new-capacity commissioning: a company adding a plant will see reported utilisation dip as the new furnace ramps, which is normal, not alarming — provided demand is there to fill it.

Net debt (and net debt to EBITDA)

The survival metric. Absolute net debt tells you the burden; net debt to EBITDA tells you how many years of peak cash flow it would take to clear it — and the danger is that EBITDA is itself cyclical, so a ratio that looks fine at the peak can double when the spread halves. Where to find it: the deck states net debt directly; the ratio is usually quoted on the call. Jindal carried consolidated net debt of ₹16,019 crore at a ratio of 1.66x at end-FY26 (Jindal Steel Q4 FY26 concall), with management reiterating its "commitment to cap the net debt-to-EBITDA at 1.5x" (Jindal Steel Q2 FY26 concall) and pledging in an earlier call to keep it "below 1.5 across all cycles" (Jindal Steel Q4 FY25 concall). JSW sat far heavier at about ₹80,347 crore of net debt (Inve data, Q3 FY26). At the other extreme, APL Apollo closed FY26 with a "net cash balance exceeding INR 1,500 crores" (APL Apollo Q4 FY26 concall) — no net debt at all. The lesson: a steelmaker's leverage must be judged against trough EBITDA, not the comfortable peak number management quotes.

Value-added mix

The share of sales that is specialised, higher-margin product (coated, automotive, special grades) rather than commodity coil. It matters because it dampens the cycle — value-added steel earns a steadier premium and is less exposed to the raw commodity price. Where to find it: the deck, as a percentage of sales. Jindal hit "a record 73% of total sales" value-added in Q2 FY26 (Jindal Steel Q2 FY26 concall); JSW's value-added share reached a record 61% in Q3 FY26, "67% ex-JVML" (JSW Steel Q3 FY26 concall). A rising value-added mix is one of the few genuinely structural improvements a steelmaker can make — it shifts the business slightly away from being a pure price-taker. Across the listed universe the spectrum runs from integrated primary producers like Tata Steel and SAIL to value-added players such as Jindal Stainless.

How do you value a cyclical like steel?

This is where most retail investors lose money, and the error is mechanical: they value a steel stock on P/E.

A P/E divides price by current earnings. For a business whose earnings swing 5x across a cycle, that denominator is meaningless. At the bottom of the cycle, earnings collapse, so the P/E looks enormous — and the stock looks "expensive" exactly when it is cheapest. At the top, earnings are inflated by a fat spread, the P/E looks low, and the stock looks "cheap" exactly when it is most dangerous. A cyclical's P/E is highest when you should buy and lowest when you should sell. It is the perfect contrarian indicator, used backwards by almost everyone.

JSW Steel's reported P/E of 33.1x (Screener.in, FY26) shows the trap from the other side: the multiple looks expensive, but the earnings denominator was flattered by ₹18,607 crore of one-time other income (Screener.in, FY26). The P/E is lying in both directions at once — too high because operating profit is depressed, then artificially supported by a non-operating windfall.

So use two lenses instead:

EV/EBITDA strips out the capital structure and the accounting noise, comparing the whole enterprise value against operating cash generation. It is the cleaner cyclical multiple — but it must be read against normalised, mid-cycle EBITDA, not the peak. The discipline is to ask: what does this business earn per tonne in an average year, across the cycle, multiply by volume, and value that. Pay a peak multiple on peak EBITDA and you are double-counting the good times.

Price to book (P/B) anchors on the replacement value of the assets — the plant, the furnaces, the mines — which don't evaporate when the spread does. For a capital-heavy cyclical, P/B paired with through-cycle return on equity is often the most honest gauge. JSW trades at about 3.0x book with a 10.1% ROE (Screener.in, FY26); Jindal at about 2.2x book with an 8.2% ROE (Screener.in, FY26). The question is whether the through-cycle ROE justifies the premium to book — a steelmaker earning a single-digit ROE across the cycle while trading at 3x book is pricing in an up-cycle that has to actually arrive.

The owner's frame: don't ask "is this cheap on this year's earnings?" Ask "what does a tonne of this company's steel earn in a normal year, how many tonnes can it sell, and what am I paying for that mid-cycle stream?"

A worked case: APL Apollo and the spread that isn't a price

The cleanest way to feel why spread — not price — is the metric is to look at a steel company that barely makes steel at all. APL Apollo buys hot-rolled coil from the primary mills and converts it into structural steel tubes. Its economics are not the global steel price; they are the spread it adds over the coil it buys — pure conversion margin. That makes its EBITDA-per-tonne guidance unusually clean to track, and it shows the said-versus-did pattern beautifully. (Illustration, not a view on the stock; figures as reported by the company.)

At the Q4 FY25 call, management guided "EBITDA spreads should be between Rs. 4,600 to Rs. 5,000 per ton" for FY26, "significantly higher than FY25" (APL Apollo Q4 FY25 concall). Two quarters later, with the business running well, they raised the bar — "upgrading our sales volume growth guidance of 20%... with the EBITDA guidance of almost INR5,500 per ton" (APL Apollo Q2 FY26 concall). And by the year's close, the chairman could report "EBITDA per ton reaching upward of INR 5,500" with a record print of ₹5,525 in the quarter (APL Apollo Q4 FY26 concall) — guidance set, raised, and delivered.

But read the same call for what didn't hold. Pushed on durability, management reframed the number downward in tenure: "First, you say from INR5000 to INR5500 per ton, I think this on a long term basis" — quietly resetting ₹5,500 from a run-rate to a long-term aspiration (APL Apollo Q4 FY26 concall). And the volume target had already been walked back: the FY26 volume-growth guidance of 15–20% was cut to 10–15% within the year — "20% is difficult. If it gets worse then we [are] targeting 15%" (APL Apollo Q4 FY26 concall). An earlier commitment to drive employee cost down to "Rs. 600-Rs. 700 per ton" simply stopped being mentioned (Inve data) — guidance that quietly went silent.

The point is not that management misled anyone — this is, on the record, a well-run converter that delivered its headline spread. The point is the texture: the spread guidance was hit, the volume guidance was trimmed, and a cost target was dropped, all in the same business, across a few quarters. You only see that pattern by tracking each commitment against the quarter it was made — which is the entire job of Promise Tracker, and the kind of thing no one reconstructs by re-reading four transcripts by hand.

Red flags specific to a steel company

• Realisation rising while the spread falls. If prices go up but coking-coal and ore costs rise faster, the company is busier but poorer. Always net the two.
• Capacity commissioned into a weak price environment. A new furnace is a fixed cost. If it ramps as the cycle turns down, it converts operating leverage into a drag — high depreciation, high interest, no spread to cover them.
• Net debt judged against peak EBITDA. A 1.5x net-debt-to-EBITDA ratio at the top of the cycle can become 3x at the bottom when EBITDA halves, without a single rupee of new borrowing. Stress-test the ratio against trough earnings.
• A P/E that looks cheap. For a cyclical, a low P/E usually means peak earnings, which means you are buying the top. Treat a "cheap" steel P/E as a warning, not an invitation.
• One-time income dressing up the profit. As with the ₹18,607 crore other-income line above, a fat reported profit can be a non-operating windfall. Strip to operating EBITDA per tonne before you believe any earnings figure.
• Falling captive raw-material share. A company that mines less of its own ore over time is increasingly exposed to the commodity swing it least controls.

Frequently asked questions

A repeatable workflow

1. Anchor on the spread. Realisation per tonne minus coking-coal and iron-ore cost, tracked as a change quarter to quarter. Net the two — never read price alone.
2. Read EBITDA per tonne over time. Six to eight quarters from the deck; one quarter is the cycle, the trend is the company.
3. Check volume against capacity and utilisation. Is growth funded sensibly, and arriving into demand or into a downturn?
4. Stress the balance sheet. Net debt against trough EBITDA, captive raw-material share, and whether the leverage survives the down-leg.
5. Value on the cycle, not the quarter. EV/EBITDA and P/B on mid-cycle numbers — and treat a cheap-looking P/E as a red flag.
6. Audit the guidance. Check the per-tonne spread, volume, and capacity commitments against what actually happened next.

Inve's KPI Screener lines up realisation, EBITDA per tonne, net debt and volume across steel companies — value, trend and a data-confidence flag per number — so the per-tonne hunt takes minutes, not an afternoon of PDF-mining. For a sibling spread business read in a completely different shape, see how to analyse an NBFC, where the spread is interest, not steel.

Where this lens can be wrong. The strongest case against everything above is that the spread, the metric this whole guide leans on, is set by a global commodity price no analyst can forecast. You can read EBITDA per tonne, captive ore share and net debt perfectly and still be blindsided by a China demand collapse, a coking-coal spike, or an anti-dumping ruling that resets the spread overnight. Reading the operating numbers tells you whether a company is built to survive and compound through a cycle — a strong balance sheet, captive ore, a rising value-added mix. It does not tell you when the cycle turns, and at the trough even the best-run steelmaker earns little. The honest claim is narrower than it looks: this analysis lowers your odds of owning a fragile mill into a downturn and raises your odds of owning a durable one into the recovery. It cannot time the recovery, and a great balance sheet earns a poor return for as long as the spread stays thin.

The owner's question to sit with before buying any steel stock: across a full cycle — not this quarter's spread — what does a tonne of this company's steel earn, how many tonnes can it sell, and is the balance sheet strong enough that it is still standing, still investing, and gaining share when the price finally turns? If the answer leans on this year's fat spread holding forever, you have read the peak, not the business.

Related reading

• How to Analyse a Bank Stock (NIM, CASA, GNPA, ROA)
• How to Analyse a Building Products Company (Pipes, Tiles)
• How to Analyse a Cables and Wires Stock