China's September Steel Data: A Weak Month Masked by a Softer Cumulative Decline

The National Bureau of Statistics' Jan–Sep 2026 release, reported on 19 October, has been read largely through its cumulative lines: crude steel down 2.9% year on year to 724.77 million mt, pig iron down 3.0% to 628.44 million mt, finished steel down 2.0% to 1.0671 billion mt. That framing is misleading. The cumulative decline actually narrowed by 0.2 percentage points versus January–August, and the monthly detail explains why that is not a sign of recovery.

September was below the run-rate, not above it

At 73.59 million mt, September crude steel output rose just 0.1% year on year and fell 1.4% month on month. The important comparison is against the year's own average. Dividing the nine-month total by nine gives a monthly mean of 80.53 million mt — September ran about 8.6% below it. Pig iron tells a sharper version of the same story: 65.34 million mt in September against a 69.83 million mt nine-month monthly average, roughly 6.4% below.

Source: NBS data as reported by SteelOrbis, 19 October 2026. Final column is my calculation.

The narrowing of the cumulative decline is therefore an arithmetical artefact of a weak August dropping out of the comparison, not evidence of a September rebound. On the reported month-on-month changes, August crude steel output can be back-calculated at approximately 74.6 million mt (my estimate).

Hot metal is where the signal is

The most useful information in this release is the gap between pig iron and crude steel. Over nine months they are almost indistinguishable (−3.0% versus −2.9%), but in September pig iron fell 1.4% year on year while crude steel edged up 0.1%, and pig iron dropped 3.4% month on month against 1.4% for crude steel. That divergence implies mills are trimming blast-furnace hot metal and covering the difference from other feed — scrap, or drawn-down semi-finished stocks.

For upstream markets this matters more than the crude steel headline. Iron ore and coking coal demand are levered to hot metal, not to crude steel, because the latter also counts electric-arc and scrap-fed output. A September pig iron rate roughly 6% below the year-to-date average is a meaningful pullback in seaborne iron ore and coking coal requirements, and it is consistent with the softer iron ore pricing environment that would normally accompany it — though this release alone cannot establish the price response.

The pig-iron-to-crude-steel ratio for the nine months works out at 86.7% (my calculation), a normal-to-high figure for China's furnace mix. There is no evidence here of structural scrap substitution; the September divergence looks cyclical, likely a response to margin compression and to the seasonal pre-winter construction slowdown and heating-season restrictions that typically begin to bite in October.

Finished steel is a poor guide to consumption

The finished steel series deserves careful handling. At 1.0671 billion mt against 724.77 million mt of crude steel, the implied ratio is 147% (my calculation). That is not a yield and should not be read as evidence of expanding downstream processing: China's 钢材 (finished steel) series double-counts product that is re-rolled or reprocessed, so the ratio routinely sits far above the ~95–100% physical yield of a single pass. Its value is in its direction, not its level.

That direction is worth noting. Finished steel fell 3.8% year on year in September — a steeper decline than crude steel's +0.1% — yet rose 3.0% month on month to close essentially at the nine-month monthly average. Read together, this is consistent with a downcycle in apparent demand being partly offset at the semi-finished or net export margin, rather than with genuine downstream restocking. SteelOrbis attributes the September softness to "sluggish demand from downstream users," which is trade-press framing rather than an NBS statement, but it aligns with the macro backdrop in the same release: industrial value-added grew 5.5% year on year, yet fixed asset investment and real estate development remained the drag.

The annualised picture

Annualising the nine-month run-rate gives roughly 966 million mt for 2026 (my estimate). If September's weaker monthly pace persisted through the fourth quarter, the year would land nearer 945 million mt. Either outcome implies a second consecutive year below one billion tonnes and a further step down from 2025, when the implied comparative base (my calculation, from the reported percentage changes) was around 746 million mt for the same nine months.

That framing is more useful to planners than the headline percentages. China is not cutting output deliberately to engineer a supply response; the data show production tracking demand down almost tonne for tonne. Q3 industrial capacity utilisation of 73.1% reported alongside these figures supports the same reading.

What to watch

The October and November prints will resolve whether September was seasonal or structural. Two markers are worth tracking: whether the pig iron–crude steel spread persists, which would confirm genuine hot metal retrenchment rather than a one-month scrap substitution; and whether finished steel continues to undershoot crude steel, which would suggest net exports absorbing the difference — a trade-policy exposure for Asian and European producers as much as a Chinese domestic story.

The tonnages here are single-sourced through trade press pending publication of the NBS industrial output table; readers relying on them for contract or hedge decisions should verify against the NBS release directly.