Hyundai Rotem Company (KRX:064350) earned an operating margin of 18.2% in the second quarter of 2025. Then 17.1%. Then 16.5%. Then 15.4%. Then 14.5% in the second quarter of 2026.
Five consecutive quarterly declines, totalling 3.7 percentage points.
Over the same five quarters revenue rose from ₩1,417.6bn to ₩1,606.1bn, so operating profit still grew — ₩257.6bn to ₩232.4bn quarter to quarter, which is actually a decline, though first-half 2026 operating income of ₩456.6bn was ahead of the ₩405.7bn earned in the first half of 2025.
A company delivering the largest arms export in Korean history, with a backlog above $20bn, is getting steadily less profitable on each won of revenue. The reason is not a problem. It is worth understanding anyway.
Where a margin falls tells you what caused it.
Gross margin: 25.3% in the second quarter of 2025, then 23.5%, 22.6%, 21.3% and 21.5% in the second quarter of 2026. Almost the entire operating margin decline happened above the overhead line.
That rules out cost inflation in the head office or a step-up in selling expense. What it points to is mix — the company is recognising revenue on work that carries a lower margin than the work it was recognising a year ago.
Management has said exactly that. Profitability declined as the second phase of the Poland K2 programme entered its early stage, while an increase in fourth-phase domestic K2 mass production reduced the export share of revenue.
Two effects, and they are different from each other.
The first is structural and permanent.
Korea's Defense Acquisition Program Administration buys domestic weapons on a regulated basis. The contractor's margin is set by formula against allowable costs, and it is not generous — the whole point is to prevent a monopoly supplier from extracting rents from its own government. Export contracts are negotiated commercially, with the customer's alternatives being American or German tanks at considerably higher prices and longer lead times.
So the same K2 tank, built on the same line, earns a materially different margin depending on which flag is on the buyer's paperwork. When Korean army orders rise as a share of output, blended margin falls, and nothing has gone wrong.
For a US reader, the closest analogue is the difference between a Foreign Military Sales case and a domestic cost-plus contract, except that the Korean domestic margin formula is tighter than the American equivalent.
The important consequence: Hyundai Rotem's margin is a function of the export/domestic split, and that split is decided by two governments rather than by the company.
The second effect is temporary and it is the more interesting one.
Long-term defence contracts are recognised as work progresses. Revenue and profit are booked in proportion to costs incurred against total expected costs, which means the margin recognised depends entirely on the estimate of what the contract will eventually cost.
At the start of a large new programme, that estimate is conservative. Engineering, tooling, supply chain qualification and first-article production all consume cost while the company is still learning what the contract will really take. Auditors will not let a contractor book optimistic margin on a programme that has barely begun. So the early quarters of a big contract dilute the blended margin, and the margin improves later as uncertainty resolves and the estimate at completion gets revised.
The second Poland contract — reported at $6.5bn, and elsewhere as around $6.1bn for the EC2 phase, a discrepancy worth noting and resolving from the company's own disclosure — is exactly that kind of programme. It is the second exercised contract under a framework covering up to 1,000 K2 tanks, and the first 180 tanks from the initial contract have already been delivered.
So the current margin reflects a large new contract at its least profitable stage, running alongside a mature contract that has finished delivering.
Total backlog was around $20.1bn at the first quarter of 2026, essentially unchanged from the end of 2025 and up sharply from about $12.7bn at the end of 2024.
Flat backlog with rising revenue means the company is delivering roughly as fast as it is winning. That is fine for a year. It is also why analysts trimmed price targets in July on order uncertainty: a defence stock trades on the next contract, not the current one, and Hyundai Rotem has already booked the two Polish tranches that drove its re-rating.
The defence division's own operating margin was reported at 27.2% in the first quarter, against a group figure of 15.4%. That gap tells you what the rail and equipment businesses contribute — which is revenue, and not much profit.
Take the constructive view, which I think is largely correct.
A margin falling because a company signed a $6.5bn contract is a far better problem than a margin falling because it lost one. The dilution is arithmetic, not deterioration, and it reverses on a schedule the company can predict even if it will not publish it.
The domestic mix effect is also not permanent. Korean army K2 orders come in defined phases; the fourth phase will complete. And a domestic order book has strategic value beyond its margin — it keeps the production line loaded between export tranches, which is precisely what allows Hyundai Rotem to promise Poland delivery dates that European manufacturers cannot match.
The cash position supports the patient reading. Cash and equivalents reached ₩2,527.3bn at 30 June, up from ₩908.4bn at the end of December, as advance payments on the new contract arrived. Total equity is ₩3,369.8bn and retained earnings ₩1,755.4bn, against a deficit of ₩156.6bn as recently as FY2019.
The gross margin trend across the next two quarters. It fell four quarters in a row and then ticked up slightly, from 21.3% to 21.5%. If the third and fourth quarters both come in above 22%, the EC2 dilution has passed its trough and the mix argument is confirmed. Another two quarters below 21% would mean something other than contract phasing is at work.
Second, any disclosure of the export share of defence revenue. That single ratio explains most of the margin, and the company does not routinely publish it. It should be asked for on every earnings call.
Third, the next order. With backlog flat and the two Polish tranches booked, the question for 2027 is what replaces them — a third Polish exercise under the 1,000-tank framework, a new customer, or nothing. The margin question resolves itself. The order question does not.
kstock reads DART every morning and writes up what moved — the contract, the buyback, the number that does not add up. The daily post and a Saturday roundup, by email.