돈經濟

Between the structural strength of AI demand and cyclical fears

On October 30, 2026 (local time), Intel's stock price plummeted by approximately 29% in a single day, causing its market capitalization to evaporate by about $140 billion (approximately 187 trillion Korean Won). This was the largest drop in 52 years since a 29.9% decline during the oil shock in July 1974, and it even exceeded 30% during intraday trading. The Philadelphia Semiconductor Index fell by over 7%, and major semiconductor stocks such as Nvidia, AMD, and Qualcomm also showed weakness, leading to the disappearance of approximately 860 trillion Korean Won in market capitalization from the US semiconductor sector in a single day. This shock was interpreted as a signal that the market's confidence in the AI investment boom was beginning to crack, going beyond mere underperformance by an individual company.

Intel's third-quarter performance itself was not bad. Revenue increased by 21% year-over-year to $16.5 billion, and adjusted earnings per share exceeded market expectations. In particular, data center and AI segment revenue surged by 78%, demonstrating that demand for AI server CPUs was indeed alive. However, investors' attention was focused not on the already announced results, but on the fourth-quarter guidance. Intel's projected fourth-quarter revenue of $15 billion to $16 billion was significantly below market expectations of approximately $17.1 billion. The adjusted EPS forecast of $0.25 also fell short of the estimated $0.43. The gross margin was also expected to be lower than the previous quarter, at around 38%.

There were two direct causes for the stock price crash. First, the slump in the PC market. As AI data centers absorbed a significant portion of the global memory supply, PC memory prices surged, leading to higher laptop and desktop prices and a contraction in demand. Intel predicted that PC shipments would decrease by 12% in 2026 and 10% in 2027, and that recovery would be difficult until memory supply and prices normalized. Second, the explosion at Fab 28 in Kiryat Gat, Israel. This accident, which killed two workers, halted operations for nearly two weeks, causing disruptions in 10nm server chip shipments and an estimated revenue loss of approximately $600 million over two quarters. The stock, which had surged nearly 150% year-to-date on expectations of a recovery in AI server CPU demand and news of investments from the US government and Nvidia, collapsed all at once due to these two negative factors.

This incident is in line with the shock and continuity of Alphabet (Google's parent company). Despite a 25% increase in total revenue in the third quarter, Alphabet's search and advertising segment growth fell short of market expectations, and above all, its free cash flow (FCF) recorded a negative for two consecutive quarters, causing its stock price to plummet by 12.6%. The evaporation of approximately $500 billion in market capitalization in a single day was the second-largest single-day decline in Wall Street history. Amazon also recorded a large deficit in its FCF over the past 12 months, and Microsoft issued a large bond to invest in AI. The market has now begun to ask, "When will the massive AI investment be recovered with sufficient cash flow?" rather than "How fast are big tech companies growing?" Intel's plunge became an opportunity for this doubt to spread throughout the entire semiconductor supply chain.

In the Korean market, fear of Samjeonnix (Samsung Electronics and SK Hynix) immediately grew. Already in October, due to foreign selling and the theory of AI peaking, stock prices had adjusted by over 40% and 50% respectively compared to their June highs, and the Intel shock was highly likely to put additional downward pressure on Monday's market. On November 2nd, in the Nextrade pre-market, Samsung Electronics and SK Hynix both plunged by over 4%, reflecting these concerns. Immediately after the opening of the regular session, the KOSPI started at 5637.77, down 1.62%, falling to the 5600 level, with Samsung Electronics down nearly 3% and SK Hynix showing weakness in the 1% range. Amid continued net selling by individuals and foreigners, doubts about the sustainability and profitability of AI investment appeared to be a burden on the domestic semiconductor industry.

However, the intraday trend was different. A large net purchase by institutional investors (approximately 1.7 trillion won) flowed in, rapidly narrowing the decline, and the KOSPI turned upward in the afternoon. At market close, the KOSPI finished trading at 5859.08, up 2.24% from the previous day. Samsung Electronics closed at 229,000 won, up 1.78%, and SK Hynix closed at 1,472,000 won, up 3.81%, fully recovering from the sharp pre-market decline and performing well. Related stocks such as SK Square also showed strong performance, and the entire semiconductor sector closed with a 2% increase. This is interpreted as a result of the undervalued appeal based on earnings coming into play after significant adjustments on October 30.

Samsung Electronics recorded an operating profit of 107.4 trillion won in the third quarter, becoming the first Korean company to usher in an era of 100 trillion won in quarterly profit. The semiconductor (DS) division alone accounted for 106.5 trillion won, effectively driving the entire profit. SK Hynix achieved an operating profit of approximately 83.3 trillion won with an operating profit margin of 80%, demonstrating record-breaking profitability. The combined operating profit of the semiconductor divisions of both companies reached approximately 190 trillion won. This is the result of a simultaneous surge in memory prices and a skyrocketing demand for high-bandwidth memory (HBM) for AI servers.

More importantly, there is visibility into future demand. SK Hynix stated that customer requests for HBM demand exceed available supply by more than double, and including volumes still under negotiation, DRAM and HBM are virtually sold out until 2029. Samsung Electronics also confirmed that its HBM production for next year is already exhausted, and it has signed long-term supply agreements (LTAs) with 10 major customers. SK Hynix has signed LTAs with 14 customers. The industry and securities firms are even predicting that the memory shortage could continue until 2030. RBC and others have projected that HBM prices could rise an additional 80-100% by 2027, and there is a possibility that the DRAM upcycle could extend beyond the current 14 quarters to 2027. As memory production lines focus on HBM, the supply of general DRAM and NAND has also become tight.

Stock prices show an extreme divergence. Despite record-high performance, sold-out situations, and long-term contracts, stock prices have been suppressed by the "AI peak theory" and "cycle fear," pushing the PER down to around 3x. This is due to the market's learned behavior that the memory industry is a typical cyclical industry. In past boom periods, preemptive selling based on "oversupply is coming soon" was repeated, and this time, the fear was amplified by factors such as the deterioration of big tech FCF, the technological pursuit by Chinese companies (e.g., CXMT), and Elon Musk's statement that "China will flood the market with an enormous amount of DRAM." Global asset allocation shifts, with foreign capital moving towards the attractiveness of US Treasury bonds and a strong dollar, also made Samsung Electronics and SK Hynix, which have abundant liquidity, targets for selling. In one week in October, foreigners net sold over 11 trillion won in KOSPI, with over 78% of that concentrated in Samsung Electronics and SK Hynix.

The future trend of the semiconductor industry, especially the Korean memory sector, is the coexistence of short-term volatility expansion and mid-to-long-term structural growth.It can be summarized as follows.

In the short term (from order receipt to several months), the ripple effect of the Intel and Alphabet shocks remains, making it highly likely that Samsung Electronics and SK Hynix will experience further declines or sideways movement. If big tech companies adjust the pace of their AI investments or concerns about ROI spread further, memory order forecasts may be conservatively readjusted. We must also be wary of the phenomenon where a contraction in PC demand acts as a paradox for rising memory prices (price increase → demand decrease). Since stock prices are dominated by sentiment rather than performance in this phase, volatility will be high for several weeks after Monday. Leverage liquidation and forced selling could also amplify the decline.

However, from a mid-to-long-term perspective (2027-2030), growth is dominant. The expansion of AI infrastructure is evolving beyond simple chatbots to inference, agents, multimodal, and physical AI, exponentially increasing memory intensity. Data center server replacement cycles and hyperscalers' large-scale capital expenditures are occurring simultaneously, and the company's confidence that supply will remain tight until 2028-2029 is not mere marketing. Korean companies possess both technological leadership in HBM (SK Hynix's market share advantage) and economies of scale, with demand somewhat fixed by long-term contracts. China's pursuit is a risk, but there is still a significant technology and yield gap in advanced HBM and high-capacity server DRAM. Although Micron is increasing capital expenditures, it will be difficult to quickly reshape the HBM market that Samsung Electronics and SK Hynix have dominated.

Ultimately, this Intel crash is not a sign of an 'AI bubble collapse,' but rather a readjustment where reality catches up after expectations have run too far ahead.Even though Intel itself showed growth in AI server CPUs, being hampered by its core PC business and supply constraints demonstrates a complex phase where AI demand and traditional cycles coexist. Korean semiconductor stock prices have already undergone significant adjustments and entered an undervalued territory. If evidence accumulates that big tech's AI investments are translating into actual profits, or if the memory shortage lasts longer than expected, it could trigger a rebound. While brokerage target prices vary widely based on upward revisions to earnings forecasts, current stock prices do not adequately reflect either scenario.

However, the risks are clear. If the ROI on AI investments is slower than expected, or if the global interest rate and exchange rate environment deteriorates, capital outflow could recur. The possibility of oversupply materializing after 2028 cannot be ruled out. China's increased mass production and technological pursuit could also exert downward pressure on prices. Therefore, rather than a full sell-off due to short-term fear, an approach of managing portfolio weight based on valuation relative to earnings and long-term demand visibility seems reasonable. While the semiconductor industry as a whole is on a growth trajectory, significant volatility and selectivity are expected before the fruits of that growth are evenly reflected in stock prices. Intel's worst day in 52 years showed that even in the age of AI, the stability of core business and supply capability remain key variables determining stock prices.