Is the ban leaking? Solar modules after the US Xinjiang + AD/CVD controls

Trade-flow companion to the price wedge (R72). Most cases in this corpus ask whether a controlled material re-enters the buyer through a single laundered origin (antimony → Thailand). Solar shows a new signature: serial trans-shipment — the origin hops country-to-country, staying one step ahead of the trade remedy. China's direct solar line was severed a decade ago; the Chinese-controlled supply chain simply re-labels through whichever country the US has not yet tariffed. As of 2024 that was the Southeast-Asia four; in 2025, with duties finally landing, the line jumped again — to Laos and Indonesia, two countries that make essentially no solar wafers.

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 (a country exporting multi-GW of modules it has no feedstock to make) plus public corporate/plant records and a final US Commerce circumvention determination — never asserted as smuggling on any single shipment.

Live chart + method: [/situation-room → "Is the ban leaking?"](/situation-room).

Verdict

The US spent a decade walling off Chinese solar: AD/CVD orders on Chinese cells (2012) and modules (2015) cut China's direct share of US module imports to under 1% (CSIS, US Census data); the 2021 BIS Entity List designation of five Xinjiang polysilicon/silicon producers (Hoshine, Daqo, GCL, East Hope, XPCC) and the UFLPA (enforcement from June 2022) then targeted the ~35–45% of world polysilicon that originates in Xinjiang. The controls did not sever the dependency — they moved its flag. US customs shows the textbook serial-relabel fingerprint:

Source: UN Comtrade (reporter USA, annual, HS 854143 "PV cells assembled in modules/panels"), free public preview API, pulled 2026-07-19. 2025 is a partial year (world $6.7 bn vs $15.2 bn full-2024), so read the 2025 column as shares, not annualised totals.

Two facts do the work. First, the Southeast-Asia four collapsed in lockstep with the duties: Cambodia went from a $1.38 bn (9.1%) supplier in 2024 to $0.9 M (0.0%) in 2025; Vietnam, Thailand and Malaysia roughly halved their shares. Second, the volume did not go home to China (still 0.2%) and did not vanish — it re-appeared under two new flags: Laos 1.8% → 15.1% and Indonesia 2.9% → 36.8%, making Indonesia the single largest supplier of US solar modules in 2025. Laos + Indonesia combined: 4.7% (2024) → 51.9% (2025). Independent confirmation: BNEF (via ING) puts the Laos + Indonesia combined module import share at 3% → 40% and cell share at 5% → 53% across 2024→2025.

Neither Laos nor Indonesia produces solar-grade polysilicon or wafers at scale — China makes ~97% of the world's wafers. A country cannot ship multi-GW of modules built on feedstock it does not make. The surge is Chinese wafers and cells assembled one border further from the tariff.

The corporate pipe (the common-ownership tell)

The re-labelling is not anonymous. The US Commerce final anti-circumvention determination of 23 Aug 2023 (88 FR 57419) made it a legal finding: cells made in Cambodia/Malaysia/Thailand/Vietnam from Chinese wafers, and modules using such cells that also contain more than two Chinese components (silver paste, frames, glass, backsheet, EVA, junction box), are circumventing merchandise. Commerce named BYD (H.K.) (via Cambodia), Canadian Solar and Trina Solar (via Thailand), Vina Solar (LONGi affiliate, via Vietnam) and New East Solar (Cambodia) among the affirmative/adverse-facts findings. That is the pipe for the SE-Asia leg — traced to a Federal Register determination, not inference.

The 2025 hop to Laos/Indonesia is the same firms moving again, each traceable to a named parent, a prior SE-Asia plant, and a new-origin exporting entity:

The pattern is the graphite/nickel mode-B (capacity relocation under Chinese ownership) fused with mode-A (Chinese wafer/cell content trans-shipped): a Chinese-controlled producer builds real assembly capacity in the next un-tariffed country and ships genuine-origin modules built on Chinese feedstock. The US has already responded — the Alliance for American Solar Manufacturing and Trade filed a new AD/CVD petition against India, Indonesia and Laos on 17 Jul 2025 (alleged AD margins Laos ~246%, India ~214%, Indonesia ~90%; combined 2024 imports ~$1.6 bn) — the fourth country-round of the same chase.

Why it matters for the buyer

1. Stated origin lies serially, not once. A risk model keyed on country-of-origin will keep marking the dependency "diversified" as it walks China → Vietnam → Cambodia → Laos → Indonesia. Each flag is genuine at the point of assembly and false as to control and feedstock. The implausible-origin cross-check (does this country make wafers?) is the only correction that survives the hop. 2. The control is binding *and* leaking at once. China's direct line is truly gone (<1%, a real severance); the Chinese-controlled supply chain is intact behind new flags. Both are true — the FEOC/origin blind spot is that "no Chinese-origin panels" ≠ "no Chinese-controlled panels," exactly as Indonesian nickel/bauxite showed. 3. It front-runs the next enforcement break. The relabel country is where the supply chain snaps next: Cambodia went $1.38 bn → $0.9 M in one duty cycle. A buyer sourcing "Lao" or "Indonesian" modules today should price the ~250% AD margin the 2025 petition already alleges — the corridor is a year, not a decade, from closure.

Method & honesty rails

• Trade data: UN Comtrade, US imports, annual, HS 854143 (PV cells assembled

in modules) — the module line used in the US AD/CVD/safeguard cases (10-digit 8541.43.00.10). Value in USD. 2025 is partial-year; treat as shares. Bare-cell line HS 854142 and the petition's "cells whether or not in modules" scope give slightly different per-country dollar totals (e.g. the petition's India/Laos 2024 values) — noted, not merged.

• Alternative-track only: never touches buyerRelativeScore or the base exposure —

it sits beside them, like the China–West price wedge.

• Inference, not accusation: relabelling is inferred from an origin's ≈zero

wafer/polysilicon capacity plus public plant records and a final Commerce circumvention determination. No single shipment is asserted illegal beyond the cited public record.

• Corrections applied (do not conflate): Hounen is a **2025 Cambodia AD/CVD

respondent, NOT a party to the 2023 circumvention determination. There is no public evidence Jinko, LONGi, JA Solar, Canadian Solar or BYD built Laos/Indonesia plants* — the confirmed relocators are SolarSpace (Laos), Trina, Thornova/Yuncheng and New East (Indonesia). The Cambodia "3,521%" combined-duty headline is press-reported; the primary Commerce figures are separate (Cambodia 125.37% AD + 3,403.96% CVD AFA). Some Indonesian capacity sits in JVs with genuine local partners (Trina's TMAI with Sinar Mas / PLN), so origin there is a mix* of Chinese-controlled and local — a labelled nuance, not a clean single-owner pipe.

• Generalises (R72): the two-layer detector — (a) imports from origin X of a

controlled good exceed X's plausible feedstock baseline post-control; (b) exporter-of-record ↔ Chinese-parent common ownership — here yields its first serial / whack-a-mole instance: the fingerprint recurs country-by-country as each new flag is tariffed. Watch the Laos/Indonesia line for the next collapse when the 2025 petition's duties land, and for the flag after that.