Is the wall leaking? Magnesium — two suppliers that ship more than they make

Trade-flow companion to the price wedge (R72). DUAL-SCORE / alternative-track — never folded into any Tier-1 exposure score. The wedge measures scarcity; this desk measures the other half: is the control actually holding, or is the restricted trade re-entering the buyer under a new flag? Research, not investment advice. Origin-relabelling is INFERRED from statistical implausibility plus public corporate and enforcement records, never asserted as evasion of any particular consignment.

Verdict

Magnesium is the corpus's cleanest capacity-arithmetic case, because the buyer has no domestic production at all and the two flags that replaced China are small enough to count.

China smelts 950 kt of the world's 1,100 kt of magnesium metal (86%, USGS MCS 2026). The US has been walled off from it since 12 May 1995 (AD order A-570-832, pure magnesium, continued again 25 May 2023), with the alloy line covered by A-570-896. And since 2022 the US has produced none — USGS records US smelter production as nil against 64 kt of idle capacity, and US Magnesium LLC, the sole Rowley (UT) producer, filed Chapter 11 on 10 September 2025.

So 100% of US primary magnesium is imported, and ~93% of it arrives from just two countries — Israel and Türkiye — which together hold 49 kt of nameplate capacity and in 2023 shipped the US 74.7 kt. That is 152% of everything they can physically make, sent to one customer, before either country served its own industry or any other buyer.

Three findings, and the detector splits them cleanly rather than tarring all three:

1. Türkiye — adjudicated leak (mode A). CBP found evasion: Chinese AZ91E transshipped through a Turkish firm classified as a management consultancy. Türkiye's own export statistics never recorded ~40 kt that the US booked as Turkish, and by 2025 Türkiye's Chinese input line equals 98% of what it ships the US. 2. Israel — large residual, but the input test clears it. Israel exports two to three times its output, yet imports essentially no magnesium from anywhere. With no feedstock line there is nothing to relabel; this reads as a measurement or stockpile artefact, and is logged as an open question, not a leak. 3. Korea — the rung the flow moved to (candidate mode C, untested). Zero primary capacity, 99.9% Chinese feedstock, and a US-bound alloy line up 18× since 2019 — arriving duty-free under KORUS, outside both AD orders.

USGS states the rotation in its own words: for Jan–Aug 2025 vs 2024, US imports of magnesium metal (99.8% purity) fell 27% while imports of magnesium alloys rose 37% and waste and scrap rose 48%. The walled line shrank; the two unwalled lines grew.

Layer 1 — volume implausibility: the suppliers exceed their own capacity

US imports of HS 810411 (unwrought magnesium, ≥99.8% Mg — the pure line), net weight in tonnes, against each origin's USGS-recorded smelter production:

Trade data: [UN Comtrade](https://comtradeapi.un.org/public/v1/preview) free public preview API, reporter USA (842), flow imports (M), annual, HS 810411 — see Method. World rows for 2019, 2020 and 2024 were not returned by the preview tier and are shown as n/a rather than estimated; where World is returned it reconciles with the sum of named origins to within 0.5%. Production and capacity: [USGS Mineral Commodity Summaries, February 2026, magnesium metal](https://pubs.usgs.gov/periodicals/mcs2026/mcs2026-magnesium-metal.pdf), world smelter production 1,100 kt in 2025. Israel's and Türkiye's figures carry the USGS "reported" footnote — they are company/government-reported, not USGS estimates, which makes the gap below harder to attribute to a bad estimate.

The arithmetic:

Israel and Türkiye were 93.0% of US pure-magnesium imports in 2023 (74,691 of 80,294 t) and 93.7% in 2025. Russia — a genuine 60 kt producer — is the control group: its line runs 4.2 kt/yr through 2022, then goes to 209 t and then to zero, the shape of a supplier that actually stops. Neither Israel's nor Türkiye's line has that shape.

Layer 2 — the mirror, and the entity that isn't a smelter

Türkiye. Compare what the US says arrived from Türkiye with what Türkiye says it sent, same HS line, same years, both in tonnes:

Same Comtrade API, reporter Türkiye (792), flows X and M. Türkiye's "World" partner row is unreliable in the preview tier for this reporter and is not used; the partner-USA and partner-China rows are.

Two things fall out. First, 40,229 t cumulative (2021–2025) entered the US declared as Turkish that Türkiye never recorded exporting — the signature of material moving through a Turkish free-zone or transit regime, which does not enter Turkish export statistics, while the US importer nonetheless declares Türkiye as the country of origin. Second, the Chinese input line into Türkiye is flat at ~25–75 t through the years of the biggest US-bound surge, then jumps to 12,316 t in 2025 — equal to 98% of the 12,576 t Türkiye shipped the US that year. The feedstock became visible only after enforcement arrived.

The entity. CBP did not need the arithmetic. On 9 May 2025 it issued an affirmative determination of evasion in EAPA Case 7969: Thompson Aluminum Casting Company (Cleveland, OH) entered Chinese-origin AZ91E magnesium alloy covered by A-570-896, transshipped through Türkiye and declared as Turkish origin, so no cash deposits were collected. The alleger was Trinity Metals (allegation filed 6 March 2024); CBP imposed interim measures on 13 August 2024 including "live entry," and issued a final administrative determination on 18 September 2025.

The named Turkish shipper is Tigemer Danışmanlık ve Dış Ticaret Ltd. Şti. — and its own public record is the tell:

A consultancy-registered trading house in a border town, selling the exact basket of Chinese Pidgeon-process commodities, is not a metallurgical origin. CBP agreed.

And the enforcement worked, partially. Türkiye's US line halves the year the allegation lands — 27,586 t (2023) → 13,596 t (2024, −51%) → 12,576 t (2025) — and Israel's falls with it, 47,105 → 28,147 → 26,340. Neither returns to pre-2022 levels. The door narrowed; it did not close.

Israel: the residual the input test does not convict

Israel is the corpus's largest unexplained volume gap, and the honest reading is that it is probably not a China relabel:

• Israel's own export statistics report 56,085 t of HS 810411 shipped worldwide

in 2023 (50,980 t of it to the US) and 33,586 t in 2025 — against ICL-reported production of ~17,800 t in 2025 and USGS-recorded 17 kt (2024) / 20 kt (2025). The gap is on Israel's own books, not only in the US mirror.

• But Israel imports essentially no magnesium: no partner exceeds 100 t/yr on the

pure line in any year pulled, and the one real import line — magnesium alloy from China — stops in 2023 (1,650 t in 2022 → 1.7 t in 2023 → 0.3 t in 2024). A transshipper needs an input. There isn't one.

• The mirror gap also runs both ways (Israel reports more than the US records in

2023, less in 2024), which is the signature of a measurement or classification artefact rather than a one-directional laundering channel.

Candidate benign explanations, none yet confirmed: multi-year stockpile drawdown at Dead Sea Magnesium; secondary/recycled ingot counted in exports but not in "primary smelter production"; re-export from bonded storage that never enters the import series; or a units/coverage difference in the Israeli series. Until one of those is evidenced, the residual stays labelled — it is a hole in the data, not a finding against Israel.

The next rung: Korea's alloy line

The pure line is walled and now enforced. The alloy line is neither. USGS records the rotation for Jan–Aug 2025 vs 2024 — 99.8% metal −27%, alloys +37%, waste and scrap +48% — and Comtrade shows where the alloy went:

And the input line, from Korea's own statistics:

Korea does not appear in the USGS world magnesium-metal table at all: its primary production is zero. Its magnesium is 99.9% Chinese, and Chinese feedstock into Korea (23,864 t) is 191% of everything Korea ships the US on the alloy line (12,475 t) — more than enough to account for every kilo, with room left for Korea's own die-casters. Korean alloy enters duty-free under KORUS and sits outside both A-570-832 and A-570-896, which are China-specific.

This is a candidate mode C — transformation-washing, where alloying pure Mg into AZ91/AM60 is a genuine industrial process that may lawfully confer Korean origin. It is untested: no Commerce circumvention inquiry on magnesium from Korea appears in the public record as of this writing. The plywood precedent (mode C⁻) is the reason to watch it — there, Commerce ruled essentially the same shape insufficient to confer origin. Czechia and Hungary show the same structure on a smaller scale, and Czechia's US line already collapsed (9,293 → 1,378 t) once buyers moved on.

Which mode is this?

• Türkiye: mode A, adjudicated. Same molecules, new flag; a trading company with

no smelter; CBP-confirmed. The rare case where the volume-implausibility detector and a regulator reach the same conclusion independently.

• Israel: no mode. Large residual, no input line, bidirectional mirror error —

logged as a data gap.

• Korea: candidate mode C, untested. Real transformation, real Korean industry,

100% Chinese feedstock, and an origin rule that has not been litigated here.

The transmission chain the buyer should hold in mind:

> China (86% of world smelting; walled off from the US since 1995) > → Türkiye (transit relabel, convicted) and/or Korea (alloying, lawful-until-tested) > → US importer, duty-free or duty-unpaid > → aluminium alloy for auto/aerospace lightweighting, EV battery enclosures, steel desulphurisation.

Why it matters for the buyer

A Tier-1 exposure score reading country-of-origin off customs declarations will show a US magnesium consumer sourced from Israel, Türkiye and Korea — three friendly jurisdictions, apparently diversified, no China. The capacity arithmetic says that picture cannot be true: two of those three ship more than they can make, and the third makes none at all. The blind spot is not marginal — it is most of the supply.

The practical read: US magnesium supply concentration on China is materially higher than declared origin implies, and the 2024–25 enforcement wave is what re-priced it. A buyer whose contract says "Turkish origin" carries an AD-duty liability that CBP has already demonstrated it will pursue retroactively against the importer, not the foreign shipper. Per the plywood corollary, a newly-charged origin is the safer origin: Türkiye is now the watched door, Korea is the un-watched one.

Method & honesty rails

• Trade data: UN Comtrade free public preview API

(https://comtradeapi.un.org/public/v1/preview/C/A/HS), annual, netWgt in kg converted to tonnes. Reporters used: USA (842), Türkiye (792), Israel (376), Korea (410), Czechia (203). Commodity codes: 810411 (unwrought Mg ≥99.8%), 810419 (unwrought Mg alloy), 810430 (raspings/turnings/granules). Where the preview tier returned no World row, the cell is n/a — never estimated or back-filled.

• Production and capacity: USGS Mineral Commodity Summaries, February 2026,

magnesium metal chapter. Israel's and Türkiye's entries carry the USGS "reported" footnote.

• Enforcement facts: CBP EAPA Case 7969 notices (13 Aug 2024 interim measures;

9 May 2025 determination as to evasion; 18 Sep 2025 final administrative determination). The two CBP PDFs return HTTP 403 to automated fetches from this host; the facts above are taken from CBP's own document-landing pages and from trade-press reporting of them, and the case number is given so any reader can pull the primary text directly.

• Inference, not accusation. No individual shipment is alleged to be evasion here

except the entries CBP itself adjudicated. The capacity gaps are statistical implausibility; the Israeli residual is explicitly not treated as a leak; the Korean line is labelled untested and may be entirely lawful.

• Alternative-track only. Nothing in this case is folded into any Tier-1 exposure

score. The divergence between declared origin and physical capacity is the signal.

What would change the verdict

• A Commerce circumvention inquiry (or a finding of no circumvention) on magnesium

alloy from Korea would resolve the C/C⁻ question directly.

• An ICL or Israeli-customs statement reconciling ~56 kt of 2023 exports against

~17–20 kt of production would close the Israeli residual — in either direction.

• If Türkiye's US-bound line stays near its Chinese input line (≈98% in 2025), the

Turkish route has been documented rather than closed, and the duty question moves from origin fraud to substantial transformation.

Register note

Anchor action: 2026-05-07-brazil-gecex-892-magnesium-china-ad — Brazil's GECEX 892 conversion of its Chinese-magnesium AD duty to a specific USD/kg rate (8 May 2026), the register's in-force magnesium trade-remedy instrument. The US orders that this case tests (A-570-832, A-570-896) predate the IPTM filing horizon; both are cited above with their Federal Register continuations.