The Tariff Didn't Fail — It Redirected: Reading the September US Steel Import Data

The headline that matters is not the headline

AISI's October 27 release on September 2026 imports leads with a month-on-month decline of 2.9% in total volumes and a modest 1.1% rise in finished steel. Both numbers are noise. The signal is in the country detail: South Korea shipped 519,000 net tons in September, up 21% on August, and 3,957,000 NT over the trailing twelve months — a 44% year-on-year increase into a market where the Section 232 tariff on steel is 50%.

That combination — a prohibitive headline duty and a sharply rising volume from one supplier — is the central fact of the US import picture in 2026. It does not mean tariffs have failed in aggregate; the year-on-year comparison is the truer measure of direction:

Finished volumes are falling roughly twice as fast as total volumes — meaning semi-finished and slab-type flows are holding up better than product destined directly for end users. And the finishing mix is shifting: finished share of total was 69.5% in September, down from a higher base earlier in the year, consistent with more slab and billet entering the system.

Korea: the anomaly the industry has already flagged

The scale is worth quantifying. Korea's trailing-12-month total is nearly 40% larger than Brazil's (3.26m NT) and larger than Canada's (3.18m NT), despite Canada's land-border logistics advantage and its historic position as the top supplier. Canada is down 40% and Mexico down 32% over the same window. Some of that gap is Mexico and Canada losing ground to the 50% duty applied to their previously quota-protected volumes; some is Korea gaining.

The industry has not been passive. On September 28, 2026 — three days before the reference month closed — eight organisations (AISI, SMA, CPTI, USOMA, ALPPA, AISC, AAM and the United Steelworkers) wrote to Commerce Secretary Howard Lutnick and USTR Jamieson Greer seeking a Korea-specific tariff-rate quota and an investigation. Their evidence: Korean long-product shipments (including rebar) reached a record 145,300 t in July 2026, sixfold the 22,100 t of July 2024, with flat-rolled and pipe/tube volumes also climbing.

Why would a 50% duty not bite? Several mechanisms plausibly explain it, and the data cannot yet distinguish among them:

• Derivative leakage. Commerce added 407 HTSUS categories as steel derivatives in August 2025, but the dutiable base is the steel content, not the transaction value. Fabricated and downstream goods can carry a low steel-content ratio and therefore a modest effective duty.
• Product-mix arbitrage. Korean suppliers may be concentrating on grades and specifications where domestic availability is thin and buyers absorb the duty — the same categories showing the largest import growth (see below).
• Quota exhaustion elsewhere. With Canada and Mexico volumes collapsing, Korea may be picking up share in accounts that previously sourced from North American mills and cannot be served domestically at the required specification or lead time.
• Vessel timing. Net tons are recorded on entry; a few large bulk or breakbulk cargoes can swing a monthly figure. The 60% jump in Japanese volume (165,000 NT) in a single month is a reminder that monthly country data is lumpy.

This is analysis, not reported fact. AISI's Census-based country data does not decompose by product-and-origin, so the mechanism remains an open question — one the coalition letter explicitly asks Commerce to investigate.

The product-level story: infrastructure and capex, not autos

The products with the largest increases tell a coherent story about where US demand is strongest and where domestic supply is tightest.

Two clusters stand out. The first is rail: standard rails and railroad accessories surging together points to a procurement cycle rather than spot buying — likely a Class I or transit authority programme placed with an offshore mill, since US rail supply is concentrated in a handful of domestic producers with limited surge capacity. The second is piling and heavy structural shapes, both up strongly over the full year. That is a construction signal — foundation work for data centres, power generation, grid and industrial projects — where domestic mills are reportedly running tight on certain wide-flange and piling sections.

The heavy structural shapes figure is not a one-month artefact: AISI's August release already showed +55% on its own trailing-twelve-month basis. The trend is entrenched.

Permits versus entries: read both, trust neither alone

AISI publishes two import releases each month, and conflating them is a common error. The SIMA permits release (published October 8) showed 2,492,000 NT of total permit applications and 1,671,000 NT finished, with September finished market share estimated at 19% and YTD at 16%. The Census-based release a fortnight later showed lower volumes and 16% YTD share.

The gap is instructive rather than contradictory. Permits are applications, not entries; AISI itself cautions they may over- or understate. The direction — permits running ahead of entries for the latest month — is consistent with subsequent downward Census revisions. For anyone tracking trend, the permits series has more value as a leading indicator of the next month's entries than as a current volume measure; the Census series is the better level estimate.

One small inconsistency deserves note: the September release describes August finished market share as 18%, while the August release itself said 17.5%. Revisions of this size are routine but mean the 16% YTD figure should be treated as an estimate with a few tenths of a point of uncertainty.

What to watch

The policy question is now more consequential than the volume data. Three developments will determine whether September's Korean surge is a blip or a structural break:

1. Whether Commerce opens a Korea-specific case. A TRQ request from a coalition this broad, arriving alongside a presidential letter urging no weakening of Section 232, is a strong political signal. A formal investigation would create quota uncertainty for Korean suppliers and their US buyers well into 2027.
2. The direction of slab and semi-finished flows. If finished share of total continues to erode while total volumes stabilise, US mills are importing feedstock — which is a margin-positive outcome for them, not a threat.
3. The rail and piling categories. Both are specification-driven and project-timed. If they hold at elevated levels through Q4, the story is genuine construction demand meeting constrained domestic supply — and buyers should plan for longer lead times and premium pricing on those sections regardless of tariff policy.

The aggregate decline in imports is real and reflects a genuinely protected market. But a market protected by a 50% duty that still cedes share to a single supplier at a 44% annual growth rate is not a settled market. It is one where the tariff has changed the shape of trade more than its volume — and where the next policy move will be targeted, not general.

Reporting basis: figures are AISI estimates from preliminary Census Bureau data, published October 27, 2026, covering September 2026. Month-on-month changes compare September with August 2026. Trailing twelve months (T12M) covers October 2025–September 2026. All interpretive judgments are the author's.