[Edaily Reporter Park Sun-Yeop ] Analysts have suggested that the sharp decline in semiconductor stocks—despite strong earnings reports from major companies such as SamsungElectronics and TSMC—is due to the market’s focus shifting from current earnings to whether investments in artificial intelligence (AI) will continue. They also note that the key to alleviating volatility in the domestic stock market lies not in the earnings of semiconductor companies, but in the earnings reports of major U.S. tech firms.
Kim Jae-seung, an analyst at HYUNDAI MOTOR SECURITIES, stated in a report on the 20th, “It will be difficult for the semiconductor sector to alleviate investor concerns on its own through these second-quarter earnings announcements,” adding, “We need to verify the performance of existing businesses, AI revenue growth rates, and capital expenditure guidance from hyperscalers.”

Although semiconductor companies’ earnings are strong, stock prices have failed to keep pace. The Philadelphia Semiconductor Index plummeted 10% last week, falling 20.2% from its peak on the 22nd of last month. A market is generally considered to have entered a bear market when it falls more than 20% from its peak.
Among individual stocks, memory semiconductor companies saw the sharpest declines. Compared to their previous highs, Kioxia fell 52%, SK hynix(000660). dropped 38%, and Western Digital declined 36%. SamsungElectronics(005930). and Micron also fell 30% each.
The market’s concern is not the immediate earnings of semiconductor companies, but whether investment in AI infrastructure can be sustained over the long term. Capital expenditures by U.S. hyperscalers are projected to reach $752.9 billion this year and $950 billion next year. As competition for AI investment intensifies, these companies’ ratio of capital expenditures to operating cash flow has risen from less than 50% in the past to nearly 100% recently. This means they are reinvesting most of the cash generated from their operations back into AI infrastructure.
Until now, large-scale investment has been driven by the expectation that the AI market would take shape as a winner-takes-all structure dominated by a handful of companies. The reasoning was that falling behind in the competition could mean losing the entire future market, so the risk of underinvestment was deemed greater than that of overinvestment.
However, the proliferation of affordable, high-performance open-source AI models from China is casting doubt on this premise. If companies can switch between multiple models as needed rather than relying on a single one, the market dominance and profitability of existing top-tier model providers could decline.
The “Kimi K3,” recently unveiled by China’s Moonshot AI, has further fueled these concerns. The Kimi K3 ranked among the top performers in metrics evaluating the intelligence and task-execution capabilities of major AI models. Although it is not one of the existing ultra-low-cost Chinese models, it has drawn attention because the performance of this open-source model has come close to that of U.S. closed-source models.
The democratization of AI models is also shifting perceptions regarding the AI industry’s beneficiaries. Analysts note that market interest is shifting from suppliers—such as model and semiconductor manufacturers—to application companies that have secured consumer touchpoints, data, and customer loyalty. This creates a structure where companies capable of selecting various AI models, applying them to services, and monetizing them gain relatively higher value.
Semiconductor companies are distant from end consumers, so even small changes in the AI market can lead to significant stock price volatility. This is because if the profitability of AI services declines, hyperscalers will reduce their capital expenditures, which in turn could lead to a decrease in semiconductor orders.
Accordingly, Analyst Kim emphasized that three key factors must be examined in hyperscalers’ earnings reports. First, profits and cash flow from existing businesses must be maintained. Since AI investments are funded by cash generated from core operations, pressure to cut capital expenditures may increase if the profitability of existing businesses deteriorates.
The growth rate of AI-related revenue is also crucial. The market is reacting more sensitively to whether the growth rate is accelerating than to the revenue size itself. The explanation is that the growth momentum in AI cloud revenue, which gained traction in the first quarter, must continue into the second quarter to alleviate concerns about falling AI model prices and slowing profitability.
Finally, attention should be paid to the capital expenditure plans of Alphabet and Microsoft. While both companies continue to compete in AI, their ratio of capital expenditures to operating cash flow is lower than that of Oracle, Amazon, and Meta. If these two companies—which have been relatively cautious about investment—raise their capital expenditure outlook, expectations that the AI infrastructure investment cycle will continue could strengthen.
Researcher Kim stated, “If Alphabet and Microsoft demonstrate a proactive stance toward AI investment, investment sentiment in the domestic semiconductor sector could recover rapidly,” adding, “Conversely, if they adopt a more conservative stance due to investment burdens or regulatory concerns, fears that the AI investment cycle has peaked could intensify.”