When the AI bubble finally bursts in the United States, the fallout will be felt far beyond Silicon Valley. For Nvidia and Intel, it would be a bad day. For Seoul, Taipei, and Tokyo, it could be a national crisis.
Artificial intelligence has rapidly become the growth engine that Asia depends on, much like China once was. Investors from South Korea to Japan should be asking not how high Nvidia can climb, but what happens when the spending spree by American tech giants slows or crashes.
The recent announcement that Anthropic is going public sharpens the question. The IPO could be the largest in history, raising up to US$100 billion. To justify such a valuation, CEO Dario Amodei must convince investors that growth can stay near 1,300% a year, with net revenue retention around 900% and implied gross margins of 85%.
Bears point to Anthropic's US$80 billion in half-year net losses. Even if demand holds, the capital expenditure boom assumes vast, steady revenue for years to come.
The 94-page SEC filing is full of tripwires. It warns that AI poses “catastrophic or existential risks to humanity.” It also flags a heavy dependence on Taiwan Semiconductor Manufacturing Co. (TSMC), whose reliability is never certain given Beijing's plans for Taiwan and investor doubts about Washington's commitment under President Donald Trump.
And there is the threat of natural disasters to Taiwan's power grid. “Any restrictions or delays on the supply of advanced chips out of Taiwan or TSMC,” Anthropic wrote, “would adversely affect our business, financial condition and results of operations.”
Denny Fish of Janus Henderson called the IPO “a generational moment,” but added that “the devil comes down to the details, and we still don't have all the details yet.”
It could be bigger than that, as Anthropic appears to be “coming for finance,” said analyst Jame DiBiasio, author of the J.DB Report newsletter. “This is not simply a technology supplier offering banks a more intelligent chatbot,” he said. “Anthropic is moving into the applications through which financial professionals work, while supplying the infrastructure on which competing applications depend.”
South Korea's high-stakes bet
On AI, signals from Seoul are mixed. Last week, data-center firm Elice fell 32% in its Kosdaq debut, the exchange's biggest IPO this year. Yet chip maker SK Hynix reported quarterly profit up sevenfold to US$63 billion and plans to return about US$45 billion to shareholders, which is lifting the Kospi. The index is up nearly 42% this year.
That momentum helps bury the week's other headline: Goldman Sachs cut its 12-month Kospi target by 25% over AI-investment worries and valuation volatility. The index's forward P/E has dropped to roughly four, the lowest since 2006. Goldman said the market is “pricing an overly negative fundamental outlook” and still sees “attractive risk-reward.”
No major economy has remade itself around AI as fast as South Korea. For President Lee Jae-myung, it has paid off. He pledged before taking office in June 2025 to push the Kospi above 5,000, about double its level then, by fixing corporate governance and ending the “Korea discount.” Thanks to AI, the index got there without any significant legislation or reform.
The export boom followed. Chips now make up 41% of Korean exports, which rose 83.5% year-on-year in September. In the first 20 days of October, semiconductor exports jumped 263% and computer exports 435%. Even under Trump's tariffs, chip shipments to the US rose 154% in October, and shipments to China rose 155%.
Those are 1970s “Asian tiger” growth rates, not the numbers of the world's 12th-largest economy and an OECD member. And they look unsustainable. If Microsoft, Amazon, Meta and Alphabet decide next year that they need fewer chips, servers and data centers, the shock would travel at rapid speed from California boardrooms to Korean factories, the stock market and the Bank of Korea. As a recent report warns, circular financing is already signaling that the AI bubble is set to burst.
Taiwan's concentration risk
Taiwan, too, is riding the wave. AI-driven double-digit growth is eclipsing the hollowing out of manufacturing sectors like machinery, basic metals, textiles and chemicals. The economy grew a blistering 12.92% year-on-year in the second quarter, following 15.43% in the January-March period.
These are Taiwan's fastest growth rates since the 1980s. In September, Taiwan's exports posted their second monthly record in a row, jumping 60.9% from a year earlier. It marked the 35th straight month of year-on-year gains.
Even with US tariffs and trade curbs, Taiwan's exports to the US surged 106.2% in September year-on-year to US$27.46 billion, while exports to China jumped 30.2%. These, too, are Asian tiger-level increases.
Taiwan markets also arguably face the biggest concentration risk anywhere. TSMC directly drives as much as 15% of Taiwan's roughly US$1 trillion in annual GDP. Indirectly, though, roughly 1,600 domestic suppliers rely on TSMC, many of them large enough to affect unemployment trends if things went awry.
At the same time, TSMC also makes up at least 42% of the Taiwan Stock Exchange Weighted Index. Any bad news from TSMC could quickly reverse the index's 61%-plus rally this year.
TSMC is also something of a global indicator, sitting at the center of global AI infrastructure spending and whether hyperscalers increase or cut capex budgets. All this means global markets pay very close attention to granular data both from TSMC and Taiwan more broadly. Any zigs or zags by hyperscalers would affect foundries, chip packaging operations, equipment suppliers, industrial real estate trends and demand for banking credit.
Though Taiwan is technologically indispensable, Korea's risks are more systemic. The International Monetary Fund has warned about the concentration of AI-related exports in a few economies, and the potential for a sudden reversal. As Seoul and Taipei markets shine while Western fiscal haze lingers, the question is how long this can last.
The AI boom has been a boon for Asia, but it is a fragile one. The region's leaders would do well to prepare for the day when the music stops.
