Summary of Key Points
Samsung’s profits soared by 18 times in the second quarter, yet its stock price plummeted, reflecting that investors are no longer solely focused on financial performance but are more concerned about the sustainability of the huge investments in AI. Sima Jing, Chief China Investment Strategist at BCA, believes that the AI bubble lies not in valuation (whether the stock price is high or not), but in profitability (whether profits can be maintained in the long term). She argues that current AI investment is in a “vertime period” (the decisive moment has not yet arrived, but time is running out) and suggests paying attention to four warning signs to determine if the bubble will burst. If it does, the U.S. stock market could fall by 30%-50%, with impacts far exceeding those of the 2000 internet bubble.
Detailed Analysis
1. The AI Bubble: Not So Much About High Stock Prices, but More About “Inflated” Profits
Sima Jing’s reference to a “profit bubble” means that many AI-related companies are currently generating high profits, which may be temporary and unsustainable.
- Example Comparison: Similar to the banking and real estate sectors before the 2008 financial crisis, their profits were strong at the time, but they collapsed once the factors supporting them (such as mortgage defaults) disappeared.
- The “Illusion” in the Semiconductor Industry: Semiconductor companies have high profit margins, but their price-to-earnings ratios are not high. This is because the market has realized that these high profits are fueled by short-term AI investments that may not be sustainable in the long run.
- The Hidden Danger: Depreciation: By 2030, major cloud providers (such as Amazon Web Services and Alibaba Cloud) will have purchased AI equipment (e.g., GPUs) worth approximately $2.5 trillion, with annual depreciation of $200 billion. This year alone, depreciation will amount to $500 billion, while these companies’ total profits last year were only $400 billion—meaning that future depreciation will erode all their profits, although this has not yet fully reflected in their financial reports.
- **The “Artifice of Double-Counting”: When you buy an AI chip for $100, the company selling the chip counts it as revenue (profit), and you count the money you spent on the chip as a capital expense (not an expense, but an asset), resulting in a profit appearance in their financial statements. In essence, the same amount of money is counted twice, creating a illusion of profitability that cannot be sustained in the long run.
2. Why a “Tvertime Period” Rather than a “Penalty Shot”?
The term “tvertime period” indicates that AI investment has not yet reached its final outcome, but there is little time left for companies to prove the viability of their investments. Three factors support this view:
- Profitability: The issue of depreciation has not yet emerged, but signs are already visible (future depreciation could exceed profits).
- Demand: Major companies’ free cash flow will peak in 2024 and be nearly zero by the end of this year. With cash flow dwindling while profit margins reaching new highs, this contradiction will eventually lead to a collapse.
- Supply: Manufacturers of storage chips are rapidly expanding production. Past experience with the internet shows that technological advancements increase output per unit of investment, so AI investment growth may not continue indefinitely. Once capacity catches up with demand, prices and profits will decline.
Sima Jing believes that even if companies can demonstrate the profitability of their investments, the growth rate of AI investment will likely be slower than this year, and it could even stagnate by 2028. Investors are concerned about the rate of growth, not the absolute amount; a slowdown next year would still cause panic.
3. Four Warning Signs to Identify a Bubble Burst
She provides four indicators that anyone can understand:
- GPU Rental Rates: Rental rates for older GPUs have decreased (small companies are cutting budgets first), while new models remain stable (major customers are waiting for the new versions). This is an early warning, but the situation is not yet critical.
- Storage Chips: There is still a shortage in the short term, but production is expanding rapidly. Prices are rising slower than expected, indicating that the market is already responding to changes in supply and demand; prices will fall before a true surplus occurs.
- AI Adoption Rate: 20% of major customers account for 80% of spending, while most small customers spend only $11 per month (using cheaper models). Many companies are reducing their AI expenditures as more affordable open-source models become available.
- Token Prices and Agent Downloads: The price of tokens (used to run AI models) has dropped, and the number of AI programming agents being downloaded has stagnated. This is because cheaper open-source models are becoming more widespread, leading companies to prioritize cost control over premium options.
4. The Potential Impact of a Bubble Burst: A 30%-50% Drop in the U.S. Stock Market, Worse than the 2000 Crisis
Sima Jing predicts that the impact of an AI bubble burst would be more severe than the 2000 internet bubble:
- High Proportion of Tech Investment: In the first quarter of this year, tech-related investments in the U.S. accounted for 5% of GDP (the highest in history). A sharp decline in these investments would significantly reduce overall demand.
- Deep Integration of Family Wealth: American households hold $70 trillion in stocks, half of which are in tech companies, accounting for 230% of GDP (compared to 130% during the 2000 bubble). A 1-dollar drop in stock prices would reduce consumer spending by 4 cents; a 20% market decline would decrease consumption by 3%, equivalent to a 2% reduction in GDP.
- Greater Impact on the Rich: The U.S. has a “K-shaped economy” (where the wealthy get richer while the middle class relies on wages). A bubble burst would cause the wealth of the top 1% to shrink, shifting the economic structure from K-shaped to L-shaped (where even the wealthy face significant losses).
She predicts that if the AI bubble bursts, the U.S. stock market could fall by 30%-50%, which sounds alarming but might not be extreme given widespread panic and selling.
Conclusion
AI investment appears promising at present, but there are underlying concerns about unsustainable profitability. Investors should monitor these warning signs closely. If the bubble does burst, the impact on the U.S. economy and stock market would be more severe than that of the 2000 crisis. Ordinary investors can use these indicators to decide whether to adjust their investments accordingly.