Samsung, SK Hynix, and 2x ETF account for 70% of trading volume in South Korean stock market! The culprit behind the sharp drop in AI computing power is not "excessive computing power", but leveraged stampede.

Individual investors in South Korea have become so enthusiastic about leveraged ETFs that these products, as well as the two major semiconductor manufacturers in South Korea that they track - SK Hynix and Samsung Electronics, now account for over 70% of the total market value of the Seoul stock market trading volume of approximately 43 trillion USD. Therefore, some seasoned financial market analysts believe that the recent sharp pullback in the global AI computing power theme, especially the violent fluctuations and pullbacks of the benchmark stock index in the South Korean stock market between single-day surges of over 10% and multiple single-day circuit breakers, is more skewed towards the sell-off of AI semiconductor crowded trades and a structural imbalance in stock market trading under high leverage, rather than a sudden deterioration in AI computing power demand or so-called computing power surplus. This anomaly has sparked criticism of South Korean financial regulatory agencies allowing such high-risk instruments, with one opposition lawmaker even calling for their rapid delisting. Since late May, exchange-traded funds (ETF products) tracking the daily investment returns of Samsung Electronics and SK Hynix, the two major storage chip manufacturers, have been listed on the South Korean stock market, and the volatility of shares of these two giants in the storage chip industry, which account for 50% of the market capitalization, has soared. These high leverage ETF products in the Korean market are designed to attract global individual investor funds into the domestic market. On the South Korean market level, Samsung, SK Hynix, and their leveraged ETFs together accounted for more than 70% of the 43 trillion USD market turnover in South Korea, reaching 84% at the end of June. Index pricing is shifting from "fundamental discovery" to "leverage product rebalancing driving". Therefore, the core of the recent plunge in popular AI computing related technology stocks in South Korea and even globally should not be simply understood as a sudden collapse in demand for HBM, DRAM, AI servers, data center CPUs, etc., but rather a violent collision between AI fundamentals and leverage-driven market structures. Statistical data shows that the CSOP two times leveraged SK Hynix ETF, launched approximately nine months ago, has expanded to around 13 billion USD, becoming the largest such fund globally. On high volatility trading days, this ETF and similar products can account for two-thirds of SK Hynix's total turnover, while SK Hynix's market value is approximately 1.2 trillion USD. This means that a product that was supposed to "track the underlying stock" has instead impacted the liquidity, closing rebalancing, and risk pricing of the underlying stock. The AI computing power bull market sees the most intense position liquidation! From the AI storage frenzy to a technical bear market: South Korea's 43 trillion USD market led by leveraged ETFs These ETF products focused on high leverage AI semiconductor strategies in global stock markets have attracted strong buying sentiment from individual investors, as they provide a high-risk, high-return way to amplify bets on the epic rise of AI semiconductor stocks. Both Samsung Electronics and SK Hynix's stocks have more than doubled in value this year at their peaks. However, as seemingly insignificant questions surrounding massive AI computing infrastructure investments cause violent fluctuations in these AI semiconductor stocks with every minor development in the supply chain, these bets are now backfiring and leading to even greater volatility in leveraged products. According to data compiled by CLSA Securities Korea, on May 26, the day before the launch of these leveraged ETFs in the South Korean market, the turnover of the two storage chip manufacturers accounted for 31% of the entire South Korean market. If the trading volume of these ETFs is added, this proportion reached a staggering 84% at the end of June and 73% on Tuesday, highlighting how the concentration of this market, which has been the best-performing market globally so far this year, has deteriorated. GPUs are responsible for generating intelligence, HBM/DRAM for high-speed data processing, enterprise-level NAND/eSSD for hot data and caching, and HDD for long-term storage of cold/temperate data. Therefore, financial giants like Goldman Sachs Group, Inc. believe that the AI computing power arms race led by cloud computing giants is turning storage chips from cyclical products into scarce strategic assets, and the price increases in DRAM/NAND by 2026 are not the end but possibly the initial stage of a super cycle. Whether it's Alphabet Inc. Class C's massive TPU AI computing power cluster, or NVIDIA Corporation's AI GPU computing power cluster at an immense scale, both rely on HBM storage systems integrated with AI chips, along with the current technology giants accelerating the construction or expansion of AI data centers requiring large-scale purchases of server-grade DDR5 storage and enterprise-grade high-performance SSD/HDD; Samsung Electronics, SK Hynix, and Micron Technology, Inc. are all positioned in the core storage areas of HBM, server high-performance DRAM (including DDR5/LPDDR5X), and high-end data center-level SSD, making them direct beneficiaries in the "AI memory + storage stack" and capturing the "super bonus" of the AI infrastructure wave. As depicted in the above description, Samsung Electronics, SK Hynix, and high leverage ETFs collectively dominate the South Korean stock trading market. The key mechanism is that a single stock leveraged ETF must maintain a daily double exposure: passively increasing holdings when the underlying asset rises and passively reducing holdings when the underlying asset falls. When the trend is upward, this creates a positive feedback loop of "the more it rises, the more you buy"; when the trend reverses, it becomes a mechanical liquidation of "the more it falls, the more you sell". Professional trading desks have begun to put estimating the end-of-day rebalancing size of ETFs on par with analyzing SK Hynix's earnings prospects; banks are hedging exposures through swaps, futures, options, and exotic derivatives, with a significant increase in related hedging costs, with the derivative cost for protecting against a sharp drop in SK Hynix reportedly rising from around 3% in March to over 10%. This explains why the volatility in the South Korean stock market is no longer just driven by company fundamentals but also by the participation of derivatives, market makers, bank balance sheets, and individual investor fund flows - a "microstructure shock". Fidelity International portfolio manager Ian Samson said, "To provide the return on leveraged ETFs, ultimately, when these stocks rise, there needs to be someone buying more of these stocks; when they fall, there needs to be someone selling more of these stocks to maintain a constant leverage ratio." "The active participation of global individual investors in trading around the launch of new leveraged products will only amplify the continuous and record-breaking volatility driven by the booming investment in the South Korean semiconductor industry and the huge, fundamental uncertainty." Even before single-stock ETFs were approved by regulatory authorities, many investors and analysts had warned that any cracks in the AI investment boom could lead to serious repercussions in the South Korean market, given the high dependence on these two storage chip giants - Samsung and SK Hynix, which account for 54% of the benchmark index, KOSPI. While these products were originally intended to shift global individual investor funds from US stocks to the domestic market to curb further weakening of the Korean won, the country's top financial regulatory agency recently expressed regret and mentioned their negative side effects. The Kospi index closed down 5.4% on Wednesday, bringing its decline since the peak in June to 20% and entering a technical bear market. Large fluctuations of over 5% are becoming more common. On Tuesday, as the stock market plummeted by over 8% during the session, the temporary suspension-style circuit breaker was triggered for the sixth time this year, accounting for half of the circuit breaker incidents since 2000. Jongmin Shim, an analyst at CLSA Securities Korea, stated that the concentration of funds on these two super-weighted storage chip stocks and their leveraged products "often exacerbates short-term price fluctuations and, in our view, has dragged down the broader market breadth and bullish sentiment over the past few weeks." "Nevertheless, I would view the current correction as occurring within the context of a bull market, rather than as the beginning of a long-term systemic bear market." The recent sharp pullback in the global AI computing power theme, especially the violent fluctuations and pullbacks of the benchmark stock index in the South Korean stock market between single-day surges of over 10% and multiple single-day circuit breakers, is more skewed towards the sell-off of AI semiconductor crowded trades and a structural imbalance in stock market trading under high leverage, rather than a sudden deterioration in AI computing power demand. It's not a sudden deterioration in AI computing power demand nor the so-called "computing power surplus" triggered by Meta, the parent company of Facebook, selling idle computing power resources. Samsung, SK Hynix, and the single-stock leveraged ETFs tracking them collectively contribute over 70% of the trading volume in the South Korean 43 trillion USD market, with Samsung and SK Hynix themselves accounting for about 54% of the KOSPI index. This means that the market has shifted from "fundamental pricing" to "mechanical amplification structure driven by a few AI storage leaders + leverage products + individual investors chasing highs". KOSPI fell by 5.35% on July 8, entering a technical bear market with a pullback of over 20% since the June high, with Samsung falling by 6.3% and SK Hynix by 5.7%, triggering temporary trading restrictions during the session, resembling a crowded de-leveraging trade rather than an industrial demand cliff. Looking at the actual data from the industry chain, AI computing power demand remains strong: Samsung's second-quarter operating profit is expected to increase nearly 19 times year-over-year, driven by AI-driven storage demand and rising DRAM and NAND prices; Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR also indicated a continued increase in the adoption of AI models across consumer, enterprise, and sovereign applications, driving the continued demand for CoWoS advanced packaging and advanced process AI chips; NVIDIA Corporation's CEO, Huang Renxun, stated that the company has long locked in capacity supply to support very strong growth, but is still in a limited supply situation for AI GPU computing power clusters. The real bottleneck in the AI infrastructure expansion remains in HBM, advanced DRAM, enterprise-grade NAND, advanced packaging and GPU clusters, as well as optical interconnect products, showing that these AI infrastructure resources are the true bottlenecks for the expansion of AI infrastructure. However, when a market's turnover, index weights, and leverage products are concentrated on a few stocks, any marginal news about AI capital expenditure, supply expansion, or regulatory tightening will be magnified by position crowding and rebalancing mechanisms into liquidity stampedes. Therefore, this pullback seems more like a position cleansing and volatility repricing in the AI super cycle rather than a peak in AI computing power demand. Nomura, a renowned Wall Street investment institution, has recently released a research report refuting the "semiconductor peak theory", while Bank of America Corp (BofA) has just released a new research report stating that by 2027, global capital spending on cloud computing and artificial intelligence-related infrastructure will reach 1.5 trillion USD, suggesting that the current summer pullback in AI semiconductor stocks, including storage chips, is a healthy reset trajectory rather than a result of any structural change in artificial intelligence computing power demand. Nomura, a well-known Wall Street investment institution, has issued a research report countering the "semiconductor peak theory". Nomura's key argument against the "semiconductor peak theory" is not simply that AI chip stocks will continue to rise, but that AI cloud infrastructure demand is transitioning from a single point GPU shortage to systemic component mismatch. According to Nomura's research framework, AI server revenue is expected to grow by 78% in 2026 and 76% in 2027, with the global data center projects increasing from 240 to 280, including around 50 gigawatt projects, and a forecast of 32GW of new computing power deployment in 2027 and 23GW by 2028; however, the real bottleneck is shifting from NVIDIA Corporation's AI GPU and Alphabet Inc. Class C's TPU production capacity, Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR's CoWoS advanced packaging to storage chips, wafer-level substrates, AI PCBs, copper-clad laminates (CCL), electronic fabrics, MLCCs, glass substrates/ABF substrates, IC substrates, high-end capacitors, power management chips, and data center optical high-speed interconnect components. Nomura is currently the most bullish large financial institution on the storage chip sector. In a recent research report, Nomura raised the target price of Samsung Electronics from 590,000 Korean won to 670,000 Korean won, and the target price of SK Hynix from 4,000,000 Korean won to 4,700,000 Korean won, both representing potential upside of over 100%, indicating that the soaring trajectory of these two major storage chip giants since 2025 has not yet halted. Nomura's core bullish logic lies in the fact that AI has transformed storage from a traditional PC/mobile cyclical product into a long-term growth asset in data centers: intelligent AI inference requires massive key-value cache (KV Cache), with HBM supply significantly lagging behind demand; it forecasts that global data center capital expenditure will increase from 1.16 trillion USD last year to 6.13 trillion USD by 2030, with memory accounting for a potential increase in data center investments from the current 9% to 23%, meaning that Samsung and SK Hynix are significantly undervalued at around six times their P/E ratio 12 months ago, with a potential valuation adjustment space approaching a valuation system of around 20 times for Taiwan Semiconductor Manufacturing Co., Ltd. Sponsored ADR. Currently, the most bullish target price for US storage chip giant Micron (MU.US) on Wall Street comes from senior analyst Gil Luria of DA Davidson, a prominent asset management institution, who has set a target price of $2,000 per share, maintaining a "buy" rating and significantly raising the target price from $1,500 to $2,000. Based on Micron's stock price of around $975.56, this target price corresponds to a potential upside of about 105%; if the stock price reaches $2,000, the potential market value would be around $2.29 trillion, close to a level of $2.3 trillion.