Market Note — 2026-08-30
Market commentary on the global ferrous scrap trade. Not trading advice.
The read
Chinese steel firmed through August on a cost rally in coke, not on a recovery in demand. Turkish scrap — the price that actually sets our world — did not move.
That combination is the whole note. When steel rises because buyers are buying, scrap follows, because mills that are selling more need more feed. When steel rises because coke got expensive, the mill’s cost base moves and its appetite for scrap does not. The two look identical on a price screen and mean opposite things to anyone selling scrap.
What moved
Chinese port coke has had a hard summer. Quasi-first grade coke delivered to north China ports was assessed around CNY 1,530 a tonne in late April. By the assessment dated 21 August it was around CNY 1,790 — up roughly CNY 260 a tonne, about 17%, with CNY 110 of that added in a single week. On the Dalian Commodity Exchange, the most-traded coke contract gained CNY 133.5 a tonne week-on-week into the same date, and the most-traded January coking coal contract closed at 1,609 yuan a tonne on 27 August. Producers pushed through consecutive rounds of increases, not one.
Now set that beside the scrap prints across the identical month.
- US-origin HMS 80:20 CFR Turkey sat at $375.21 on 19, 20, 21 and 24 August, then $376.23 on 26 and 27 August. A dollar and two cents, in a month.
- The general 80:20 CFR Turkey mark went the other way: $373.25 → $373.00 → $372.00.
- The published range held $368–376 CFR and did not break either end, on any session, all month.
Chinese steel meanwhile did firm: the China HRC export index printed $494.38 on 27 August. And iron ore did not lead any of it — 62% Fe front-month futures at $95.92 and the 61% Fe CFR Qingdao assessment at $97.79, both easing on the day.
The mechanism, and why sellers should care
Coke is a blast-furnace input. Scrap is an electric-arc input. A coke rally raises the cost of making steel from iron ore; it does not, by itself, create one extra tonne of scrap demand. It can raise finished-steel prices enough that scrap-based producers find room to pay more — but that is second-order, and only if the increase sticks.
So the framing for a scrap seller is this. A cost-push rally in China is not a bid. It is a change in a competitor’s economics, and it becomes relevant to a Turkish cargo only when a Turkish mill actually pays more. Through August none did: coke ran 17% while the Turkish legs moved a dollar either way.
When a rising Chinese number is quoted at you as grounds for a higher scrap price, the question is not whether it is real — it is. The question is whether it is a demand signal or a cost signal, and whether it has reached the destination that would be paying you.
The transmission question
The broader question is whether a Chinese cost rally can set a floor under Turkish scrap at all.
The case that it can: dearer coke lifts blast-furnace costs, which supports finished-steel prices globally, which eventually supports what electric-arc mills pay for feed. That chain is real but long, and every link can absorb the move instead of passing it on.
The case that it cannot: China’s marginal cost is not Turkey’s. Turkish mills buy deep-sea scrap against European and Baltic collection, their own finished-steel demand, and freight. None of those three is set in Dalian.
August was a reasonably clean test, and it did not favour the first.
What would prove this wrong
Stating the falsifier plainly, because a view without one is just a preference.
Coke has run up hard since April — roughly 17% on the port assessment — while Turkish scrap stayed directionless through the same period. If scrap stays flat through another leg higher in coke, the China cost chain does not set the Turkish floor. That would be settled evidence, not opinion, and it would mean cost-push arguments sourced from Chinese raw materials can be set aside in a Turkish scrap negotiation.
The reverse counts too. If the next coke leg is followed within a few weeks by Turkish scrap breaking the top of its range on actual bookings, not offers, then the chain does transmit on a lag — also worth knowing.
No forecast here and no range. This is a mechanism note: the point is telling apart a market that is bidding from one that is merely expensive to supply. Only one of those pays for a cargo.
All figures are published exchange or assessment prints on the dates stated. Prices move; a figure quoted here is accurate to its date and no later.