Summary of Key Points
In the past two days, AI-related stock markets around the world (mainly in the U.S., Japan, and South Korea) have experienced a series of sharp declines: AI semiconductor stocks in the U.S. market have plummeted (NVIDIA dropped nearly 5%, and SK Hynix' shares fell below their issue price), South Korean stocks triggered circuit breakers and closed down by 10.84% (SK Hynix lost 14.65% in a single day), and AI sectors in the Japanese and Chinese markets also suffered heavy losses. The reasons behind these declines include the South Korean government's sudden measures to reduce leverage, subpar returns from AI applications, China's breakthroughs in the semiconductor supply chain that challenge international giants' monopolies, as well as external factors such as Federal Reserve policies. However, this wave of decline is essentially a short-term adjustment of valuation bubbles and leverage levels; the long-term growth potential of AI remains unchanged.
Detailed Analysis
1. The South Korean "Leverage Boom": Sweetening the Deal Before Pulling the Ladder Away
In May this year, the South Korean government introduced 16 "double-leverage ETFs" that allowed ordinary investors to borrow money to buy AI stocks like SK Hynix and Samsung, potentially doubling their profits but also doubling potential losses. After the stock market soared, the government suddenly became concerned about high leverage levels and started to tighten regulations:
- Increased margin requirements for leveraged ETFs (more money was needed to borrow);
- Moved the launch of new products forward to July 31;
- Combined with a central bank interest rate hike, raising the cost of borrowing;
- Limited individual investors' leverage on any single stock to no more than 20% of their total assets.
As a result, 1.2 million leveraged accounts were required to pay additional margin, and over 300,000 accounts had their positions forcibly liquidated (when stock prices dropped below a certain level, brokers sold the stocks to cover debts, with some investors even ending up owing the brokers money). The number of circuit breakers triggered in South Korean markets this year exceeds the total for the past 26 years, reflecting a clear case of inflating bubbles through leverage before abruptly bursting them.
2. Growth Bottlenecks in the AI Industry
The deeper reason for the decline in the AI sector is the disconnect between "investment" and "returns":
- The demand for upstream computing power (such as NVIDIA chips) is high, but midstream cloud providers (like Amazon and Google) are spending too much on these chips, eroding their profits and putting pressure on valuations;
- The downstream application side (e.g., AI chatbots and generative AI) lacks "blockbuster" products that could generate significant revenue; investments of one dollar may not yield the same return (low ROI).
This has led to concerns that although NVIDIA is a key player, if cloud providers and end-users cannot sustain demand, its status as a market leader could change. Recently, the cost of NVIDIA's credit default swaps (CDS) increased, indicating growing concern about its debt risk.
3. China's Semiconductor Breakthroughs: Challenging International Giants
International AI giants like Samsung, SK Hynix, and ASML used to profit from their monopolies. However, this year, China's advancements have reduced their competitiveness:
- Memory Chips: ChangXin Technology's listing in the Chinese stock market has filled a domestic gap, allowing it to compete with Samsung and SK Hynix and challenge their price monopolies;
- Lithography Machines: China has begun mass-producing DUV lithography machines (although not the most advanced EUV models), impacting ASML's market share.
These breakthroughs have narrowed the profit margins of these international giants, leading to declines in their stock prices (e.g., ASML fell by over 8% and Tokyo Electron by over 10% last night).
4. This Decline is a Bubble Bursting, Not an End to AI's Future
Despite the sharp drop, it is mainly due to excessive short-term gains and leverage-related liquidations; the long-term fundamentals of AI are still strong:
- Market Potential: Global AI spending exceeded $2.5 trillion this year and is expected to grow by 30% annually. Companies are investing heavily, so the demand for hardware will not disappear suddenly;
- Performance Support: For example, SK Hynix's net profit quadrupled in the first quarter of 2026 and is expected to increase sixfold in the second quarter, with a forward P/E ratio of only 7.8 (indicating a potential return on investment in 7.8 years, which is relatively low); NVIDIA and Samsung's valuations are not considered inflated.
In short, this is a temporary adjustment of market bubbles, not an end to AI's trajectory. Just as the internet bubble burst in 2000, new leaders like Google and Amazon emerged afterward.
5. Future Risks to Watch: The Federal Reserve and Geopolitics
Two external factors could continue to influence the market:
- Federal Reserve Interest Rate Meetings (July 30): If interest rates rise further, borrowing costs will increase, potentially leading to more stock market declines;
- U.S.-Iran Nuclear Negotiations (July 28): A failure in these negotiations could drive up oil prices and affect the global economy and markets.
Conclusion
There's no need to panic; opportunities often arise from market downturns. This decline is similar to the leverage bubble burst in China in 2015, but the long-term trend of AI as a global technology driver remains unchanged. After the short-term turmoil, companies with core technologies will emerge. Once leverage is reduced and markets stabilize, new opportunities will present themselves. Just be patient.