Is the licence leaking? Tungsten APT and the Taiwan share-shift
Third material in the circumvention track (R72), companion to the antimony and graphite cases. The price wedge measures scarcity (what a controlled material costs ex-China). This case measures the other half: is the control actually holding, or is Chinese-controlled material re-entering the US through a laundered origin? This is a DUAL-SCORE / alternative-track signal — never folded into any Tier-1 exposure score. Research, not investment advice; origin-relabelling is INFERRED from statistical implausibility plus public production data, never asserted as smuggling on any single shipment.
Read this one for the discipline as much as the finding. Unlike antimony (a hard US-specific ban) and graphite (a ban plus a 93.5% duty), tungsten is under a global export licence — friction, not a wall. The detector should, and does, return a fainter, partial fingerprint here. We label it GATE 0 and say exactly what is missing, rather than dress an emerging share-shift up as a proven leak.
Verdict
On 4 February 2025 China's MOFCOM and Customs (Announcement No. 10 [2025]) put tungsten — ammonium paratungstate (APT), carbide, and metal/alloys with W ≥ 97% — under dual-use export licensing, reinforced on 26 Oct 2025 by Announcement No. 68's state-trading-enterprise quota regime for 2026–2027. Crucially, tungsten was not in Announcement 46's Dec-2024 US ban, so it was not suspended by the Nov-2025 truce (Announcement 72): the licence stands. China holds ~79% of mined and ~90% of refined tungsten (USGS MCS 2026), so it is the textbook single-origin chokepoint.
US customs data shows an emerging trans-shipment tell in the refined-APT line — but a muted one, and different in shape from antimony/graphite:
| Tungstates / APT (HS 284180), US imports, gross wt | 2023 | 2024 | 2025 (post-licence) | 24→25 |
|---|---|---|---|---|
| China (the controlled origin) | 852 t | 939 t | 1,015 t | +8% |
| — China share of US APT imports | 39% | 39% | 30% | share ↓ |
| "Other Asia, nes" (code 490 — predominantly Taiwan; ~0% W) | 291 t | 345 t | 875 t | +154% |
| — its share of US APT imports | 13% | 14% | 26% | doubled |
| Philippines (~0% W) | 0 t | 191 t | 246 t | +29% |
| Germany (secondary refiner) | 562 t | 472 t | 635 t | +35% |
| Viet Nam (genuine producer — Nui Phao) | 370 t | 217 t | 164 t | −24% |
| World total | 2,203 t | 2,395 t | 3,388 t | +41% |
Source: UN Comtrade (reporter USA, annual, HS 284180 tungstates), free public preview API; committed artifact `data/intelligence/tungsten-circumvention.json`. Production shares: USGS Mineral Commodity Summaries 2026.
The signal here is a share shift, not a collapse-and-reroute. China's direct APT line to the US did not vanish — it rose in tonnes (939→1,015 t); its share fell only because the world total grew 41% and one origin surged. That one origin is "Other Asia, nes" — the UN Comtrade catch-all dominated by Taiwan, which appears in neither the USGS tungsten mine table nor the refining table. A ~0%-capacity origin roughly doubling its share of US APT imports (14%→26%) in the first post-licence year, while China's own line holds, is consistent with re-export/relabelling of Chinese-controlled APT to dodge licence friction and the US 50% Section 301 tariff — the same molecules under a new flag, rather than a banned origin going dark.
The upstream ore line (HS 261100) shows the collapse leg, but small and legitimately filled: China ore to the US fell 84 t → 7 t (−91%) across 2023–2025, while the gainers are real producers — Bolivia (~2% mine), Spain (Barruecopardo), Portugal (Panasqueira), Rwanda (~1.5%). That is genuine diversification of raw feed, not relabelling, and we do not count it as a circumvention tell.
Why the fingerprint is muted (and why that is the honest read)
Two structural reasons this looks nothing like the antimony collapse:
1. A licence is friction, not a wall. Announcement 10 makes exporters apply; it does not forbid US-bound tungsten. So Chinese APT keeps flowing directly (throughput ↑ 8%), and the circumvention pressure shows up only at the margin — as a zero-capacity transit gaining share, not as a direct line going to zero. The "banned-origin-collapses" leg that anchors the antimony and graphite cases is structurally absent for a licence regime. 2. Tungsten has real ex-China supply. Vietnam (Nui Phao — the largest mine outside China), Bolivia, Spain, Portugal, Rwanda, Austria (refining), and a ramping Almonty/Korea leg mean a surge from those origins would be plausible and legitimate. The implausibility test must therefore isolate origins that cannot produce tungsten — which is exactly why the Taiwan/"Other Asia, nes" and Philippines lines (both ~0% USGS capacity) are the ones that matter, and why the Vietnam line falling is not a red flag.
What is missing for GATE 1 (the labelled gap)
- Disambiguate code 490. "Other Asia, nes" is a Comtrade catch-all; it is
predominantly Taiwan but the preview API will not resolve it to a Taiwan-specific customs line. US Census USA-Trade HTS-10 (behind a free key) or Taiwan's own customs export data would confirm the origin. Until then the headline surger's identity is inferred, not traced.
- No common-ownership pipe. The strongest tell in the antimony case was the
corporate pipe (Youngsun → Thai Unipet → Youngsun & Essen). No named Chinese-tungsten-producer ↔ Taiwan/Philippines affiliate was traced this wake. Layer-2 is an open question, not a claim.
- One post-control year only. 2025 is ~11 months of licence and may revise.
A second year of data would confirm whether the Taiwan share-shift persists or was a one-off restocking artifact.
Because the identity of the headline surger and the ownership pipe are not traced, this case is GATE 0 — every figure in the tables is sourced (Comtrade + USGS), but the central interpretation rests on a labelled inference, so it is filed as an emerging signal to watch, not a proven leak.
Why it matters for the buyer
1. Even a licence bends the origin data. A risk team reading "US APT imports from China are stable" is right on tonnage but wrong on the trend: a quarter of US APT now enters via an origin that mines and refines no tungsten. The dependency is intact; part of it wears a Taiwanese label. 2. It flags where enforcement tightens next. US Section 301 and anti-circumvention scope over third-country routing make the Taiwan/Philippines APT leg the probable next pressure point. A buyer sourcing "Taiwanese" APT should know it almost certainly did not start as Taiwanese ore. 3. It calibrates the detector's honesty. Not every controlled material leaks loudly. Showing tungsten's faint fingerprint next to antimony's stark one is the point — the method distinguishes a ban from a licence instead of manufacturing a signal for both.
Method & honesty rails
- Trade data: UN Comtrade, US imports, annual, HS 284180 (tungstates/APT) —
the refined chokepoint China dominates ~90% — plus 261100 (ore), 810110 (powders), 284920 (carbide) as context. Net weight is gross product weight.
- Production baseline: USGS MCS 2026 (Tungsten): China ~79% mine / ~90%
refine; Taiwan and Philippines absent from both tables.
- Two modes, labelled (from the antimony/graphite cases): (A) *trans-shipment
relabelling — surge from a ~zero-capacity country (the Taiwan/Philippines APT lines here); (B) capacity relocation to an owned offshore leg (graphite→ Indonesia). Tungsten shows a weak (A) at the margin plus a small, legitimate ore-diversification that is not* circumvention.
- Alternative-track only: never touches
buyerRelativeScoreor the base
exposure — it sits beside them, like the China–West price wedge.
- Inference, not accusation: relabelling is inferred from USGS ~zero
production plus Comtrade. No individual shipment is asserted illegal, and the identity of code-490 and any ownership pipe are explicitly left as open questions.