A $24.34bn monthly surplus from one product, and the won went the other way

Korea’s trade surplus in integrated circuits alone reached $24.34bn in July 2026, against $6.92bn in July 2025 — a factor of 3.52 in thirteen months. Over the same thirteen months the won weakened from a monthly average of 1,375.2 to 1,497.4 per dollar, 8.9 per cent in the wrong direction for anyone reading the surplus as a source of currency strength.

Why 2.66 and 1.56 produce 3.52

The surplus is a difference, and differences amplify. Exports of HS 8542 grew by a factor of 2.66 over the thirteen months while imports grew by 1.56. Neither number is extraordinary on its own; the gap between them is, because the surplus inherits the growth of the larger series without the damping of the smaller one.

The path matters as much as the endpoints. The monthly surplus crossed $10bn in December 2025, reached $17.69bn in March 2026, peaked at $24.74bn in June and stands at $24.34bn. Three consecutive months above $20bn from a single four-digit tariff line is, by some distance, the largest such concentration in Korean trade statistics.

It is worth naming what the customs data cannot settle here. Imports rise in a boom because production requires inputs, so a widening surplus is not automatically a sign of a strengthening position — it can equally mean the import side has not yet caught up. The unit value series shows that roughly five sixths of this surplus is attributable to price rather than shipped volume, which is the more consequential fact about its durability.

The channel that fails the test

The standard argument is short. Exporters earn dollars, sell them for won, and a $24bn monthly surplus is $24bn of supply pressure on the dollar, so the won should strengthen. If that channel dominated, thirteen months of a tripling surplus would have produced a stronger won.

Source: Korea Customs Service (HS 8542), Bank of Korea (ECOS). Own calculation.

The correlation confirms the mismatch rather than the story. On levels the Pearson coefficient is 0.864, an R² of 74.6 per cent — but the sign says that months with larger surpluses were months with a weaker won, not a stronger one. Strip out the common trend by testing month-on-month changes instead, and almost nothing remains: R² falls to 2.2 per cent. A level correlation between two series that both rose over the same period is close to uninformative, which is precisely why the growth-rate test is the one that matters.

The won did strengthen, afterwards

Won strength arrived, but out of sequence. The surplus peaked in June 2026 at $24.74bn. The dollar peaked on 2 July at 1,554.4 won, and by 25 August the rate had fallen to 1,380.6, a move of 11.2 per cent in under two months. The strengthening followed the surplus peak rather than accompanying the surplus climb.

That ordering is not evidence of a lagged channel either, on a sample of this length. What it does establish is narrower and firmer: across these thirteen months, the semiconductor surplus is not sufficient to explain the won, and a sentence pairing the two as cause and effect gets the sign wrong before it gets the magnitude wrong.

What to watch

• The composition of any decline. A $24.34bn surplus is $32.72bn of exports less $8.38bn of imports. A fall driven by weaker exports and a fall driven by heavier imports mean opposite things, and imports are a measure of activity.
• When the sign flips. Larger surpluses have coincided with a weaker won throughout this sample, which is the reverse of the textbook. The reference point for the next reading is the July 2026 monthly average of 1,497.4 won.
• Concentration risk in the aggregate. One tariff line now sets the direction of the national trade balance, and excluding it changes the sign of the bilateral balance with China. The monthly country split is published as each month clears customs.