[Edaily Reporter KIM YOON JI ] As global capital converges on the U.S. artificial intelligence (AI) industry, analysts have pointed out that a cyclical pattern is emerging in which funds from countries such as South Korea and Taiwan—which have reaped massive export profits from U.S. AI investments—are flowing back into U.S. financial markets. Experts have warned that if the AI investment bubble bursts due to this cycle, the shockwaves could spread beyond the U.S. to Asian countries, including South Korea.

According to Bloomberg News on the 12th (local time), an analysis by Bloomberg Economics of U.S. Treasury data found that AI-related sectors accounted for approximately 20% of U.S. stocks held by foreign investors last year. This represents a doubling from the approximately 10% recorded in 2019. Applying this ratio to the end of June this year, the value of U.S. AI-related stocks held by foreign investors is estimated at approximately $4.9 trillion (about 6,582 trillion won). This figure is equivalent to Germany’s annual gross domestic product (GDP).
Investment in AI by U.S. companies is also expanding to unprecedented levels. In a recent paper, Columbia University economist Stein van Nieuwenburg estimated that cumulative investment in AI-related infrastructure will reach $9 trillion (approximately 12,089 trillion won) by 2032. If this comes to pass, it would represent the largest capital expenditure in U.S. history—exceeding even the 19th-century railroad construction boom in terms of investment relative to GDP.
Corporate debt is also piling up rapidly. Obligations not reflected on the balance sheets—such as lease agreements, purchase commitments, and payment guarantees—entered into by major hyperscalers like Amazon, Microsoft (MS), Alphabet, Meta, and Oracle alone amount to approximately $3 trillion (about 4,029 trillion won).
The U.S.’s massive AI investments are generating enormous profits for South Korea, Taiwan, and other countries. As the U.S. purchases large quantities of advanced semiconductors needed for data center construction, cash is piling up at major semiconductor manufacturers such as SamsungElectronics(005930)and Taiwan’s TSMC.
The U.S.’s monthly imports of semiconductors and electronic components surged from $11 billion (approximately 14.77 trillion won) in December 2023 to $71 billion (approximately 95.3388 trillion won) in July of this year. Driven by increased AI-related exports, South Korea’s current account surplus is also projected to expand from $33 billion (approximately 44.3124 trillion won) in 2023 to $45 billion (approximately 60.42 trillion won) this year. South Korea’s savings-to-GDP ratio rose from 33% at the end of 2023 to 40% in June of this year.
Part of these accumulated funds is flowing back into the U.S. financial markets. Net purchases of U.S. stocks by South Korean investors surged from $17 billion (approximately 22.8 trillion won) in 2023 to $69 billion (approximately 92.66 trillion won) in July of this year. In effect, a virtuous cycle has formed in which U.S. AI investment boosts Asian semiconductor exports, and funds accumulated in Asia flow back into the U.S. stock market, thereby fueling the AI investment boom.
Frederick Neumann, HSBC’s chief economist for Asia, warned, “Asian capital is the fuel driving the U.S. AI boom,” adding, “If AI investment falters, a painful correction will follow—not only for the U.S. and the dollar but also for Asia and the global economy as a whole.”
Bloomberg noted that these capital flows resemble those seen just before the 2008 global financial crisis. At that time, massive excess savings from Asian countries such as China and Japan flowed into the U.S., exacerbating global economic imbalances.
The nature of these investments has also changed. In the early 2000s, excess savings from Asia and elsewhere flowed primarily into U.S. Treasury bonds and other debt securities through central banks. This lowered U.S. borrowing costs and supported credit expansion. Recently, however, purchases of U.S. Treasury bonds by foreign central banks have declined, while private investors—such as hedge funds and pension funds—are concentrating their funds on relatively riskier assets like stocks and corporate bonds.
This is ultimately leading to rising U.S. Treasury yields. Rising Treasury yields increase the interest burden on developed countries—such as the U.S., which already carries massive debt—worsening their fiscal situations and potentially reducing their capacity to respond in the event of a crisis. Furthermore, there are concerns that some hedge funds, which use U.S. Treasuries as collateral to invest in risky assets, could sell off Treasuries to cover losses from AI-related investments, thereby amplifying shocks in financial markets.
Experts point out that the countries reaping the benefits of the AI boom may actually suffer the greatest damage when the bubble bursts. David Chao, a global market strategist at Invesco, noted, “Paradoxically, the countries that have invested massive amounts of capital in the AI industry could be hit the hardest when the AI boom ends.” However, he added that since the current AI investment frenzy is underpinned by the practical application of AI technology and the profitability of major companies, it is difficult to equate it with the 2008 financial crisis.