KOSPI, which had risen to 9,385.59 intraday on June 19, fell to 5,593.56 on July 30, a drop of more than 40% from its peak. Concerns about a bubble in artificial intelligence (AI) investment by U.S. hyperscalers (operators of ultra-large data centers) spilled over into debate about a semiconductor peak-out, triggering a sharp market correction. Meanwhile, the won-dollar exchange rate, which had surged to the 1,550-won range in early July, fell by more than 120 won in a month to the 1,430-won range. While KOSPI and the won-dollar exchange rate typically move in the same direction, that synchronization has broken down recently as the AI investment boom has dominated financial markets.
The AI investment boom has turned the "exchange rate determination theory based on external interest rate differentials" into a stale tale. Traditionally, a widening U.S.-Korea interest rate gap was a bearish factor for the won, but the opposite trend has appeared since 2024. Although the gap, which once widened to 2 percentage points, has recently narrowed to the 1-percentage-point range, the won-dollar exchange rate, which traded in the 1,300-won range in 2024, climbed to the mid-1,500-won range in the first half of this year.
Yield differentials move exchange rates
The AI investment boom has changed the variables that determine exchange rates. Rather than interest rate gaps, where higher expected investment returns are found moved exchange rates. Global capital flows reacted more sensitively to the dynamic variable of investment returns than to the static variable of "country-by-country interest rate gaps." A prime example is domestic retail investors' overseas stock purchases. From Jan. 2024 to the first half of 2025, before the KOSPI rally fully took off, domestic retail investors were net buyers of $28.3 billion (about 33 trillion won) in U.S. stocks, while they were net sellers of 19.8 trillion won in domestic stocks. During this period, the S&P 500 and Nasdaq indexes rose more than 40%, while KOSPI's gain was limited to 4%. The overwhelming investment returns in the U.S. market drew domestic funds to the United States, boosting dollar demand and adding upward pressure on the won-dollar exchange rate.
The roughly threefold surge in KOSPI from the 2,900s in June 2025 to the 8,400s this June can be seen as a derivative effect of the U.S. AI investment boom. As U.S. hyperscalers raced to build AI data centers, demand for HBM and other products surged, and the memory semiconductor supercycle began in earnest. With a spike in DRAM and NAND flash prices, operating profits at Samsung Electronics and SK hynix swelled to record levels, and the sharp increase in their market capitalization led KOSPI's gains. The U.S. AI investment boom not only lifted U.S. big tech stocks, it also boosted the earnings and corporate value of Korean semiconductor corporations, fueling a historic rally in the domestic stock market.
The KOSPI rally pushed up the exchange rate
The U.S. AI investment boom drew massive inflows into the Korean stock market, but paradoxically, the KOSPI rally that exceeded normal speed spurred an increase in the won-dollar exchange rate. Foreign investors who had bought domestic stocks around the 2,600 level on KOSPI tended to switch to selling once their cumulative investment return exceeded 35%. An analysis using ChatGPT found that foreign investors' behavior began to change at the KOSPI 5,500 level. As investment returns surpassed target levels, mechanical rebalancing by institutional investors kicked in in earnest.
In fact, in May–June, when KOSPI rose to 7,700 and 8,400 and returns from year-end reached 88.4% and 107.9%, respectively, foreign investors were net sellers of roughly 100 trillion won in stocks. Statistically, from January this year to recently, the correlation coefficient between foreign investors' net stock purchases and the won-dollar exchange rate's rate of change was -0.47. This indicates a clear correlation in which foreign net selling drives a rise in the won-dollar exchange rate. In lag-adjusted analysis, the relationship in which foreign trading precedes changes in the exchange rate also appeared statistically significant. By contrast, the hypotheses that the U.S.-Korea interest rate gap or exchange rate moves dictate foreign trading did not show significance. This supports the view that foreign profit-taking and converting won into dollars added upward pressure on the exchange rate. In the end, KOSPI's over-speed rally triggered foreigners' mechanical rebalancing, which in turn led to won weakness.
Easing stock market volatility is the key to currency stability
Foreign investors, who unloaded about 150 trillion won of domestic stocks in the first half, were net buyers of more than 4 trillion won in domestic stocks on July 14–24, when KOSPI fell into the 6,000s on a correction in semiconductor stocks. After taking profits during the KOSPI surge, they used the pullback as a chance to rebuild net long positions. In this process, the won-dollar exchange rate also fell into the 1,400-won range.
This analysis suggests that exchange rates also cannot be viewed separately from the asset market. In the AI era, not only external variables like the global dollar's value but also capital flows driven by domestic stock market volatility can move exchange rates. The KOSPI surge triggered by the AI investment boom in May–June and foreigners' large-scale profit-taking are prime examples showing this new mechanism. It is necessary to recognize that, just as a sharp drop in the domestic market can, an over-speed rise can also amplify exchange rate volatility. In the AI era, where capital moves along global supply chains, the line between the domestic stock market and the foreign exchange market is increasingly blurred. Easing stock market volatility, which has grown three to five times larger than in major countries, has become an important task for exchange rate stability as well.
In that regard, delaying the National Pension Service's rebalancing, which could have eased KOSPI volatility, and instead introducing high-risk products like the double-leveraged Samsung Electronics–SK hynix ETF only amplified market volatility, working at cross purposes with the goal of exchange rate stability. It is a policy failure that should serve as a cautionary example. In the AI era, exchange rate policy should not stop at managing the foreign exchange market. It must shift to a new policy frame that jointly manages stock market flows and volatility.
Plus Point
Won-yen rate falls to the 891-won level per ¥100… Korea-Japan exchange rate decoupling
The won and the yen, which had long moved together as representative weak currencies, have recently diverged. While the won-dollar exchange rate has dropped into the 1,440-won range as foreign investors resumed net buying of domestic stocks, the dollar-yen rate has climbed to around ¥163, extending yen weakness. As a result, on July 28 the won-yen rate fell to the 891-won level per ¥100, the lowest in one year and eight months.
The difference is capital flows. In Korea, as semiconductor stocks that had surged on the AI investment boom corrected, foreigners resumed buying domestic stocks, increasing demand for the won. Inflows from the $25.6 billion issuance of SK hynix depository receipts (ADR) and the Bank of Korea's policy rate hike also supported won strength.
Meanwhile, with the Japanese government's expansionary fiscal policy intensifying and expectations that the Bank of Japan (BOJ) will raise rates only slowly, Japan continues to face strong yen carry trade pressure. In the AI era, while capital market flows in Korea have grown more influential in determining the exchange rate, in Japan, monetary policy and interest rate differentials remain the key variables for the yen.