South Korean equities extended a sharp technology-led sell-off on Oct. 13 after the KOSPI fell 3.65% to 6,384.25 in the previous session, with pressure centered on semiconductor leaders Samsung Electronics and SK Hynix. The decline followed renewed scrutiny of whether the rapid rise in spending on artificial intelligence infrastructure can translate into revenue and profit quickly enough to support elevated expectations across the chip supply chain.
Samsung Electronics closed Oct. 12 down 4.39% at 250,500 won, while SK Hynix fell 4.52% to 1,605,000 won. Selling carried into the following session, where program trading was briefly halted as losses deepened. The KOSPI has now fallen nearly 32% from its June 22 peak of 9,385.59, a four-month retreat that has turned an AI-driven equity rally into a test of earnings assumptions, financing conditions and retail leverage.
The immediate pressure on Korean semiconductor shares began after Samsung released preliminary third-quarter guidance on Oct. 8. The company projected consolidated sales of roughly 195 trillion won, within a range of 194 trillion to 196 trillion won, and operating profit of about 10.74 trillion won, within a range of 10.73 trillion to 10.75 trillion won.
Those figures pointed to substantial earnings, but the market response showed that traders were focused less on the quarterly headline and more on the pace of future profit growth. Market commentary cited expectations for fourth-quarter operating-profit growth of about 8% quarter-on-quarter, following what had been viewed as strong third-quarter momentum. A slower expansion rate would leave less room for disappointment as manufacturers commit billions of dollars to new memory, packaging and fabrication capacity.
AI revenue expectations reset global technology trading
The Korean decline also tracked weakness in U.S. technology shares after new reporting raised questions about OpenAI’s revenue run rate. By the end of September, OpenAI’s annualized revenue was reported at close to $50 billion, compared with a $70 billion figure that had circulated in recent weeks.
A subsequent Oct. 9 report said OpenAI expected annualized revenue to reach or exceed $70 billion by year-end, driven primarily by enterprise demand. The two figures may reflect different reporting periods and accounting treatment, including whether partner sales are counted on a gross or net basis. Yet the gap was enough to intensify debate over how quickly AI developers can turn heavy infrastructure spending into recurring income.
U.S. semiconductor and cloud-linked shares sold off on Oct. 8 U.S. time. Nvidia closed down 2.94%, Micron lost 4.79%, SanDisk declined 4.90%, and CoreWeave fell 7.77%. Oracle dropped more than 6% during the session before ending down 5.48%, while the Nasdaq Composite finished 1.25% lower.
The reaction exposed a vulnerability in the AI trade: chipmakers, cloud operators and model developers are increasingly valued as parts of one spending cycle. A revision to expectations for software or model-provider revenue can therefore filter rapidly into assumptions about demand for memory chips, advanced packaging, servers and data-center equipment.
Capacity expansion brings higher costs
Samsung and SK Hynix are spending heavily to secure positions in high-bandwidth memory, or HBM, the specialized memory used alongside AI accelerators. Samsung has said it plans to invest more than 110 trillion won, or about $82.6 billion, in capital expenditure and research and development during 2026. SK Hynix is investing $3.87 billion in an advanced-packaging base in Indiana.
Industry figures cited in market materials indicate that up to 25% of global DRAM wafer capacity could be committed to HBM production by the end of 2026. That shift could tighten supply for conventional memory products, but it also increases the industry’s fixed-cost base. New plants, clean-room equipment and packaging lines carry depreciation and amortization expenses that remain even if demand growth cools.
Equipment costs are adding another layer of concern. ASML has notified Samsung and SK Hynix that replacement parts for its extreme ultraviolet and deep ultraviolet lithography tools will rise 10% from January 2027, according to market materials. Broader price increases for complete lithography tools have also been discussed.
For semiconductor manufacturers, higher equipment and maintenance costs can be absorbed when chip pricing rises sharply. They become harder to manage if memory pricing flattens while new capacity enters service. The market’s focus has therefore moved beyond the question of whether AI demand exists to whether the economics of supplying that demand will remain attractive after the industry’s expansion plans are fully reflected in costs.
Foreign selling and leverage intensify the moves
Recent trading tallies showed foreign traders sold nearly 2 trillion won of Korean equities on Oct. 12. Net foreign selling over the preceding month totaled 29.7 trillion won, according to the figures cited in market commentary.
Such flows can have an outsized effect in an index dominated by large technology exporters. Samsung Electronics and SK Hynix account for a major share of Korea’s equity market capitalization, so synchronized selling in the two stocks can quickly pull down the KOSPI and trigger further risk reduction in index-linked strategies.
Retail leverage has added to intraday instability. Single-stock leveraged products are widely used by Korean retail traders, and declining share prices can trigger margin calls that force positions to be sold into falling markets. One estimate cited in market coverage placed losses connected to these products at 54 trillion won.
SK Hynix’s valuation has also reset during the decline. The company recently traded at a forward price-to-earnings ratio of 6.96, according to market figures. A lower multiple can suggest that traders see value in the stock, but it can also indicate that the market expects earnings to weaken or become less predictable after the current memory-cycle upswing.
Large cloud providers are still expected to maintain substantial AI infrastructure budgets. Morgan Stanley and Goldman Sachs project combined capital expenditure by major global cloud companies could reach $1.2 trillion to $1.4 trillion in 2027. The Korean market’s latest correction does not erase those spending plans; it places greater weight on the timing of orders, the durability of AI revenue growth and the cost of building the capacity required to meet demand.
For deeper context on volatility and margin risk, explore our guide on margin trading before leveraging your positions.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
