A quarter of what Armenia sends Russia is now electronics
The dual-use concentration is real. The supply behind it is not German.
I found this while measuring something else.
I was building a dependency model of the Eastern Mediterranean and needed one number: how much of Armenia's imports come from Iran. The answer looked like a story about decoupling. Iran's share of Armenian imports falls steadily across four years, from 8.22% in 2021 to 3.74% in 2024.
It is not decoupling. Iranian imports rose over the same period, from 437 to 628 million dollars. The share fell because the denominator moved. Armenian imports from the world went from 5.3 billion dollars in 2021 to 16.8 billion in 2024, with the inflection immediately after February 2022.
A landlocked economy of under three million people does not triple its imports in three years through domestic consumption. Something is arriving and something is leaving.
The pattern across four countries
The standard candidates for post-2022 transshipment are Armenia, Kyrgyzstan, Kazakhstan and the UAE. Comparing a 2019 to 2021 baseline against 2022 to 2024, all four show trade with Russia rising, and in every case exports to Russia rise faster than imports from Russia.
| Imports from Russia | Exports to Russia | |
|---|---|---|
| Armenia | +243% | +327% |
| Kyrgyzstan | +59% | +222% |
| Kazakhstan | +13% | +59% |
| UAE | +160% | +260% |
That asymmetry is suggestive rather than conclusive. Commodity prices, currency movements and ordinary post-pandemic recovery would all move both directions, but not necessarily by the same amount. On its own this is a pattern, not a finding.
The test that separates them
If the surge were general economic growth, it would be spread across everything these countries trade. If it were transshipment of controlled goods, it would concentrate in the chapters export controls were written for.
So the test is narrow: do HS 84 (machinery and mechanical appliances) and HS 85 (electrical machinery and electronics) grow faster than the country's total exports to Russia? A ratio above 1.0 means the chapter outgrew everything else.
It concentrates. Exports to Russia, three-year means.
| HS | Baseline | 2022–24 | Growth | vs total | Share before | Share after | |
|---|---|---|---|---|---|---|---|
| Armenia | 84 | 12.0m | 223.4m | 1,763% | 4.4x | 1.7% | 7.4% |
| Armenia | 85 | 9.0m | 800.8m | 8,786% | 20.8x | 1.3% | 26.5% |
| Kyrgyzstan | 84 | 1.3m | 132.5m | 9,996% | 31.4x | 0.4% | 13.9% |
| Kyrgyzstan | 85 | 5.9m | 43.6m | 643% | 2.3x | 2.0% | 4.6% |
| Kazakhstan | 84 | 133.8m | 984.2m | 636% | 4.6x | 2.3% | 10.5% |
| Kazakhstan | 85 | 109.7m | 710.6m | 548% | 4.1x | 1.9% | 7.6% |
| UAE | 84 | 110.3m | 585.1m | 430% | 1.5x | 11.2% | 16.5% |
| UAE | 85 | 309.1m | 2.0bn | 556% | 1.8x | 31.4% | 57.3% |
Eight country-chapter pairs. Eight increases in dual-use share.
The share columns matter more than the growth rates. A large percentage increase from a tiny base proves nothing, which is why they are here. Armenian electronics exports to Russia went from one fortieth of the total to more than a quarter. Kyrgyz machinery went from four parts in a thousand to nearly a seventh. The UAE now sends Russia more electronics than everything else it sends Russia combined.
And the absolutes rule out any base-effect objection. Eight hundred million dollars. Two billion. Nine hundred and eighty-four million. These are not rounding artefacts.
Two different things are happening
The table contains two phenomena that should not be blurred.
Armenia and Kyrgyzstan begin at one or two percent and end at fourteen and twenty-seven. Their trade with Russia has changed composition, not just volume. That looks like new routes opening.
The UAE begins at 31.4% and reaches 57.3%. It was already a substantial electronics conduit before 2022 and has roughly doubled its share. That looks like an established hub intensifying, which is a different problem with different remedies.
Kazakhstan sits between the two, with the largest baselines and the most modest ratios.
Where the goods are not coming from
I ran the same test on the inbound leg: imports from Germany, as a proxy for European supply. It goes the other way.
| HS | Baseline | 2022–24 | Growth | Share before | Share after | |
|---|---|---|---|---|---|---|
| Armenia | 84 | 63.7m | 74.8m | 18% | 27.7% | 17.9% |
| Armenia | 85 | 10.8m | 27.2m | 151% | 4.7% | 6.5% |
| Kyrgyzstan | 84 | 14.7m | 42.8m | 191% | 24.0% | 15.3% |
| Kyrgyzstan | 85 | 3.2m | 9.1m | 187% | 5.2% | 3.3% |
| Kazakhstan | 84 | 398.4m | 870.2m | 118% | 47.2% | 38.7% |
| Kazakhstan | 85 | 108.6m | 180.0m | 66% | 12.9% | 8.0% |
| UAE | 84 | 1.6bn | 2.0bn | 22% | 17.8% | 19.7% |
| UAE | 85 | 819.9m | 885.5m | 8% | 9.1% | 8.9% |
German dual-use supply into these countries grew modestly, between 8 and 191 percent, and its share of their imports mostly fell.
The arithmetic is stark. Armenia exported 800 million dollars of electronics to Russia while importing 27 million dollars of electronics from Germany. Direct German supply cannot account for the volume, and it is not close.
I want to be careful about what that does and does not establish. It does not show that German export controls are working, because goods of German origin can reach Armenia through an intermediary and appear in the data as arriving from that intermediary. What it shows is narrower and still useful: direct German exports are not the channel. The volume is arriving from somewhere the German bilateral series does not see.
That turns the question from the one I started with into a better one. Not "is this transshipment", which the concentration test answers, but what is the actual supply chain. China, Turkey, Hong Kong, the wider EU, or onward from the UAE, whose own electronics imports from Germany barely moved while its exports to Russia went to two billion.
What this does not show
The single weakest part of this is that inbound is measured against Germany alone. That was a request-budget decision after the UN Comtrade public tier blocked my address for exceeding its allowance, not an analytical one. Goods arriving from any other origin are invisible to it. A low inbound ratio therefore neither supports nor refutes anything, and the outbound leg carries the finding by itself.
Chapter-level codes contain controlled and uncontrolled goods together. HS 85 covers both a radiation-hardened microcontroller and a toaster. Nothing here identifies a single controlled item, an end user, a licence status or an intent. Establishing any of those requires customs-level data, corporate records and shipment documentation that Comtrade does not carry.
And these are mirror statistics compiled by the reporting countries themselves. Under-reporting of politically awkward re-exports is a known problem, which if anything means these figures understate rather than overstate.
What would settle it
Three things, in order of tractability.
Run the inbound test against China, Turkey, Hong Kong and the UAE as source economies, which would locate the supply chain rather than eliminating one candidate from it.
Descend from chapter to subheading. HS 8542, integrated circuits, and HS 8471, computing equipment, are far more diagnostic than the chapters that contain them.
Compare mirror statistics. Armenia's reported exports to Russia against Russia's reported imports from Armenia, where the gap is often more informative than either figure.
Method and data
UN Comtrade public preview API, annual series, retrieved 16 August 2026. Aggregate rows only, filtered on customs procedure C00, transport mode 0 and no second partner, because the API returns every breakdown combination alongside the total and truncates at 500 records per year, which silently drops the aggregate for some years. All responses cached and committed. Code, cached data and the full output are at github.com/egallantosint/medgraph.