Last week, South Korea's stock market experienced severe volatility, prompting investors to reassess whether AI-related stocks were overheated. This correction was driven not only by Korean retail investors' concentrated purchases of semiconductor blue-chips via margin financing and 2x leverage ETFs—amplifying selling pressure during the retreat of leveraged funds—but also by a reevaluation of valuations after rapid P/E expansion amid rising interest rates and concerns over Chinese memory competition.
1. Leverage Concentration in Korean Stocks and Policy Tightening Triggered Selling Pressure
The recent sharp decline in Korean stocks was initially triggered by the retreat of leveraged capital. Over the past period, Korean retail investors used margin trading and 2x leverage ETFs to heavily buy large semiconductor stocks such as Samsung Electronics and SK Hynix, fueling the upward momentum in Korean tech stocks. This year, although leveraged/reverse ETFs account for only about 7% of total Korean ETF assets, they contributed approximately 20.6% of net ETF inflows over the past three months. Notably, leveraged ETFs tied to Samsung Electronics and SK Hynix accounted for about 95% of year-to-date net inflows into leveraged/reverse ETFs.
The nature of leveraged capital is that it amplifies buying during rallies and selling during downturns. Recently, South Korea’s Financial Services Commission raised the minimum margin requirement for single-stock leveraged ETFs and tightened trading rules for related ETFs/ETNs, raising market concerns that highly leveraged positions would need to deleverage or reduce exposure. As capital shifted from chasing gains to deleveraging, selling pressure concentrated on previously high-performing, crowded holdings like Samsung Electronics and SK Hynix, further triggering profit-taking and risk-control selling. Since these two companies are key indicators in the global AI and semiconductor supply chain, Korea’s early sell-off also cooled risk appetite for tech stocks in other markets.
Source: Bloomberg, compiled by Nicheman Buy Fund. Data as of July 16, 2026. Scope includes Korean leveraged and inverse ETFs. Asset size and fund flows are in USD. The denominator for percentages is the sample of Korean ETFs listed by Bloomberg, not total retail holdings in Korea.
2. Model Efficiency and Rising Chinese Competition Increase Memory Uncertainty
From a valuation perspective, Korea’s index has a trailing P/E of about 18.7x and a forward P/E of about 7.1x, a gap of 163.6%, significantly higher than other major markets. Assuming constant stock prices, this implies the market expects next year’s EPS to be about 2.64 times that of the past year. In other words, Korea’s seemingly low forward P/E does not simply indicate cheapness, but is predicated on the assumption of substantial earnings growth. Given Korea’s high correlation with the memory industry, and the significant weight of Samsung Electronics and SK Hynix in the Korean market, any market skepticism about memory prices, gross margins, or demand growth trajectory can quickly prompt a reevaluation of Korea’s valuation.
Uncertainty in memory demand stems first from the AI model architecture itself. Chinese AI firm Moonshot AI recently launched Kimi Linear, which claims to reduce temporary memory usage and improve computational efficiency when processing long texts and complex tasks. Google Research has also proposed new model architectures to improve AI efficiency in handling ultra-long data. These developments do not mean AI demand is disappearing, but rather that the amount of high-end memory required per unit of AI service remains uncertain, making market assessments of memory sector profitability more cautious.
The supply side is also changing. Chinese memory manufacturer ChangXin Memory Technologies (CXMT) is advancing a major IPO. If the raised funds are used for capacity expansion and technological upgrades, it could enhance China’s domestic memory self-sufficiency. Even though CXMT still lags behind Korean giants in high-end AI memory, any increase in China’s self-sufficiency in general and some server memory could affect the demand momentum for Samsung Electronics and SK Hynix from the Chinese market. For Korean stocks, the risk is not necessarily a reversal in AI demand, but rather that memory demand, pricing, and margins may prove more volatile than previously expected.
Source: Bloomberg, compiled by Nicheman Buy Fund. Data as of July 16, 2026. The gap between trailing and forward P/E is calculated as (trailing P/E ÷ forward P/E) – 1. U.S. tech uses Nasdaq 100, U.S. uses S&P 500, Taiwan uses Taiwan Weighted Index, Korea uses Korea Composite Stock Price Index, Japan uses Nikkei 225, Developed Markets use MSCI World Index, Emerging Markets use MSCI Emerging Markets Index.
3. Capital Expenditure Continues to Expand, Long-Term AI Growth Theme Remains Intact
However, uncertainty in memory demand does not mean AI capital expenditure will decrease. Current plans from major tech companies show AI infrastructure investment remains high, with hyperscale cloud providers expected to spend a combined $710 billion in capital expenditures in 2026. TSMC recently raised its full-year USD revenue growth forecast and increased its capex guidance, indicating that AI computing demand continues to drive expansion in advanced process nodes and advanced packaging capacity.
What the market is reassessing is likely the allocation of AI capex across different supply chain segments, not the disappearance of the overall investment logic. If model efficiency improves or general memory supply increases, the market may revise downward price and margin expectations for certain memory products. However, cloud providers still need to continuously invest in advanced processes, advanced packaging, high-speed transmission, thermal management, power management, and server assembly to expand overall computing capacity. Taiwan’s supply chain participates across wafer fabrication, advanced packaging, servers, and key components, with benefits more diversified. If AI investment shifts from chasing memory price hikes to emphasizing computing performance, system integration, and data center construction, Taiwan’s hardware supply chain could still receive relatively strong support.
Source: Bloomberg, compiled by Nicheman Buy Fund. Data period: January 1, 2015 – July 22, 2026. 2026 estimates based on latest company guidance or median guidance. Unit: billion USD.
Nicheman Investment Strategy
The recent correction in tech stocks primarily reflects the retreat of leveraged funds in Korea, valuation reevaluation, and cooling market risk appetite—a short-term liquidity fluctuation, not a reversal in long-term AI demand. With hyperscale cloud providers maintaining high capex and TSMC continuing to expand advanced process and packaging capacity, AI computing infrastructure remains a key direction for the tech industry. Even if memory industry pricing and competitive dynamics change, investors bullish on long-term AI development may consider diversified exposure via Taiwan equity funds, avoiding concentrated bets on single memory companies or leveraged products.
Nicheman Selected Funds
Selected Funds
A49038 Neuberger Berman Taiwan 5G Equity Fund Subscribe
A36004 Allianz Taiwan Technology Fund Subscribe
A05016 Yuanta New Mainstream Fund Subscribe
Nicheman Buy Fund operates independently
This information is for reference only. Nicheman Buy Fund has made every effort to provide accurate opinions and information from reliable sources, but cannot guarantee the completeness of such data.
Under the latest Financial Consumer Protection Act regulations, to provide better investment services, our company implements an 'Investor Attribute Assessment' procedure. By completing the 'Investment Risk Suitability' assessment on the Nicheman Buy Fund website, we will apply product suitability rules based on your profile and request you to update it at least annually. Failure to complete the assessment may affect your transactions.
Fund investments are intended for long-term purposes and should not expect high returns in the short term. Fund prices and returns may fluctuate, and there is no guarantee that the full investment amount can be recovered. Investors should understand and judge independently before subscribing.
For funds denominated in foreign currencies, investors bear the exchange rate risk when converting distributions or redemption proceeds back to New Taiwan Dollars. If the NTD exchange rate strengthens against the foreign currency at conversion compared to the original investment date, investors may incur exchange losses.
Each fund employs mechanisms such as 'Fair Pricing Adjustment' (or 'Anti-Dilution'), short-term trading rules, and 'Price Adjustment Mechanism' (or 'Swing Pricing'). For details, please refer to each fund’s prospectus. Investors must comply with the relevant trading rules and restrictions stated in the fund prospectus.
Each fund has been approved or recognized by the Financial Supervisory Commission, but this does not imply zero risk. Past performance of the fund manager does not guarantee minimum investment returns; the fund manager, besides exercising due care, is not responsible for the fund’s profits or losses, nor does it guarantee minimum returns. Investors should carefully read the investor information sheet/fund prospectus before subscription. Investors should understand and judge independently before subscribing.
Fees borne by each fund (including distribution fees) are disclosed in the fund prospectus and investor information sheet. Investors can check them at the Public Information Observation Station or Fund Information Observation Station.
Due to non-investment-grade bond funds having credit ratings below investment grade or no credit rating, and being sensitive to interest rate changes
FACT BOX
- Source: PR Times
- Category: News
- Organizations: Google