The recent wave of global technology stock adjustments, triggered by South Korea's de-leveraging activities, has led to a significant pullback in the A-share computing power sector, with market concerns primarily focused on the depth and duration of this correction. We believe the current shock is primarily a transmission of risk sentiment rather than a liquidity crunch. The main risk in South Korea stems from a triple-layer leverage structure, a scenario absent in the domestic Chinese market. Coupled with China's accommodative macro liquidity environment, declining bond yields, the approaching late-July Politburo meeting, and rising expectations for consumption-boosting policies, a comprehensive assessment suggests the market does not have the conditions for a direct bear turn. Instead, the correction provides a window for bargain-hunting. For sector allocation, short-term focus should be on innovative drugs and high-dividend stocks, while increasing positions in upstream computing power segments should wait until after the South Korean leverage unwinding is cleared.
Primary Risk Source Behind the Recent Market Adjustment
The recent sharp fluctuations in the South Korean market have led to widespread adjustments in global tech stocks, with a notable correction in the A-share computing power sector. Market concerns center on the extent and duration of this pullback. We assess the current adjustment follows a gradient from "memory storage -> upstream computing power -> mid-to-downstream segments," rather than representing a broad-based retreat from AI-related trading. The root risk lies in South Korean de-leveraging: a triple-nested leverage structure involving retail investors (borrowing off-exchange, on-exchange margin financing, and purchasing 2x/3x leveraged ETFs), compounded by the high concentration of memory companies (accounting for roughly 60% of the KOSPI index). This has triggered self-reinforcing forced selling, dragging down South Korean stocks to lead global declines and spreading along the memory supply chain. Upstream computing power segments followed the adjustment but with relatively smaller declines, while mid-to-downstream segments and Hong Kong-listed internet stocks showed relative resilience. The end of this adjustment depends on whether the forced selling is exhausted; a clear market bottom can only be confirmed after leverage scales and passive selling significantly decline.
For the A-share market, this round of impact is more about sentiment transmission than a liquidity shock. China lacks similar structural leverage risks, maintains loose liquidity, and sees declining bond yields. With the late-July Politburo meeting approaching and expectations for pro-consumption policies heating up, the market lacks the conditions for a direct bear market. The correction thus offers a window for accumulating positions at lower levels.
Regarding sector allocation, short-term focus should be on innovative drugs, which have independent industry catalysts, and the defensive high-dividend sector. For the technology direction, before signals confirming the clearing of South Korean leverage emerge, priority should be given to retaining positions in Hong Kong-listed internet stocks, computing power services, and AI applications. Once signals are confirmed, positions can be gradually increased in segments with a price increase logic, such as storage, MLCC, and PCB upstream materials.
Key areas to watch include: innovative drugs, high-dividend stocks, AI applications, and Hong Kong-listed internet stocks.
Key Drivers of the Recent Market Downturn
The recent significant pullback in the A-share computing power sector, influenced by global tech stock adjustments, has raised core market concerns about timing and severity. We observe the global tech stock adjustment exhibits a distinct gradient from "memory storage -> upstream computing power -> tech mid-to-downstream," with clear differences in declines across markets and industry segments.
Firstly, the South Korean market and the global memory sector are the epicenters of this adjustment, with the primary risk stemming from South Korean de-leveraging. South Korean retail investors' leverage features a "triple-nested" structure: 1) off-exchange borrowing (e.g., credit loans for market entry); 2) on-exchange margin financing (further leverage via broker credit trading); and 3) product leverage (borrowed funds concentrated in purchasing 2x/3x leveraged ETFs on stocks like SK Hynix and Samsung Electronics). The multiplicative effect of these three layers means price declines trigger cascading, self-reinforcing forced selling, forming the microstructural root of the recent sharp South Korean market volatility.
SK Hynix and Samsung Electronics have experienced significant volatility since late June. Their combined extremely high weight and trading volume in the KOSPI index means their decline directly dragged South Korean markets to lead losses among major global indices. The volatility then rapidly spread along the memory supply chain, with notable adjustments in U.S. (Micron Technology), Japanese (Kioxia, Advantest, etc.), and A-share memory-related companies. Memory became the deepest declining segment in this global tech sector adjustment.
Secondly, other upstream computing power segments like optical modules, PCBs, servers, and semiconductor equipment/materials have also been affected by risk aversion transmission, but overall declines were smaller than the memory sector. These segments, also in the AI computing power upstream and with significant prior gains and high trading congestion, followed the correction amid sentiment spillover during South Korean market volatility. However, with limited inherent leverage exposure and their underlying order/earnings delivery logic intact, this adjustment primarily reflects passive following under risk aversion contraction, with overall controllable declines.
Thirdly, mid-to-downstream tech companies like Apple and Google faced relatively limited impact, with Hong Kong-listed internet stocks even showing counter-trend gains recently. Earnings drivers for mid-to-downstream companies stem mainly from end-demand and their own product cycles, with lower exposure to memory in their stock prices and holdings structures and weaker links to South Korean leveraged fund flows, thus demonstrating clear resilience in this adjustment. While the Taiwanese market is a key part of the global tech supply chain, its index has low memory weight, leading to relatively stable overall performance; this further confirms that TMT assets without pure computing power exposure faced overall limited impact this round.
Thus, this global tech stock adjustment is not a uniform AI trade unwinding but a diffusion from the South Korean memory epicenter, affecting A-share tech industries and the broader market along supply chains and cross-market risk aversion. The core market pricing contradiction currently is passive selling pressure triggered by a specific market and specific leverage vehicles. Therefore, judging when this adjustment ends requires considering not just index decline depth but, more crucially, when South Korea's passive selling is cleared.
Typical Conclusion Patterns for De-leveraging Episodes
The key to de-leveraging episodes lies in whether passive selling has ended. Reviewing historical典型案例 indicates that the first regulatory intervention typically marks entry into a policy intervention phase but does not equate to de-leveraging completion. A genuine market bottom usually requires several steps: "policy halts disorderly decline -> short-term market rebound -> residual leverage continues unwinding -> price becomes钝化 to negative news."
Case 1: 2015 A-Shares – Comprehensive de-leveraging often involves a second-round decline. The 2015 A-share market is the most典型 case of comprehensive de-leveraging. During the first market crash, the Shanghai Composite Index fell rapidly from its mid-June high, with concentrated liquidations of off-exchange配资 and margin positions. Regulators subsequently intervened by suspending IPOs, restricting减持, state fund入场, and providing liquidity. The market formed a阶段性 low in early July with a significant rebound. However, this rebound did not signal de-leveraging completion. On one hand,受损 leverage funds needed to reduce positions during the rebound; on the other, funds that bought the dip in the first decline could also face renewed stop-losses. Thus, from late July to August, the market experienced a second-round decline, with the Shanghai Composite falling 8.49% on August 24. Only after融资余额 and off-exchange配资明显下降 and forced selling gradually减少 did the market enter a months-long反弹 starting September. The Shanghai Composite fell over 30% from June to early July 2015, yet the post-intervention rebound did not prevent significant adjustments again in late July and late August.
The 2015 A-share case illustrates: 1) The first救市 addresses temporary market liquidity枯竭, not all leverage positions; 2) Policy-driven rebounds often also provide a window for existing leverage to reduce positions; 3) A true bottom usually appears after融资余额明显下降 and forced liquidations near an end; 4) The market often needs to test the previous low to judge if de-leveraging is complete.
Case 2: 2018 "Volmageddon" – The more concentrated the leverage vehicle, the faster the unwinding. In February 2018, the "Volmageddon" event in U.S. markets saw the VIX指数 surge intraday, causing massive losses in inverse volatility products, with some subsequently terminated or liquidated. A key difference from the 2015 A-share case is that leverage was mainly concentrated in a few inverse volatility products. As product net asset values rapidly approached liquidation conditions, positions were强制处理 within a relatively short time. Thus, product-level de-leveraging was very剧烈 but relatively short-lived. U.S. securities regulators later noted that相关 volatility products' concentrated daily rebalancing during market volatility caused significant losses for holders.
The South Korean situation is not完全等同于 the previous two cases. On one hand, South Korean leverage is concentrated in SK Hynix, Samsung Electronics, and related杠杆 ETFs, making the leverage载体 relatively明确. This theoretically makes it easier to monitor and potentially limit new risks via regulatory measures. On the other hand, SK Hynix and Samsung have extremely high weight, trading volume share, and investor concentration in South Korean indices. Sustained declines in these two stocks would not only trigger product被动减仓 but also drag down the entire KOSPI index and global memory sector. Therefore, its risk扩散范围 is明显大于一般的 single-product liquidation events.
South Korean financial regulators announced on July 16 the suspension of new single-stock杠杆 ETF listings, plans to raise the minimum investor cash threshold from 10 million to 30 million won, and strengthened trading unit and investor education requirements. This indicates regulators formally acknowledge the market risks posed by杠杆 products.
However, current measures mainly address new leverage, not directly存量杠杆: suspending new products prevents further risk expansion; raising investor门槛 reduces new散户资金 inflow;但 existing杠杆 ETFs still require daily rebalancing; already-entered融资 and credit资金 won't automatically disappear due to policies; some measures only take effect in early August, meaning存量仓位 may still波动短期.
Therefore, the current stage is closer to the "first policy intervention" phase of the 2015 A-shares case, not yet进入 de-leveraging结束阶段. The South Korean market may see a短期反弹 driven by policy intervention and oversold factors, but post-rebound,观察 of whether存量杠杆 continues减仓 is needed. Only after杠杆 product规模 and passive selling明显下降 can a true market bottom be confirmed.
Three signals to judge the end of South Korean de-leveraging:
1. Policy Signal: Shifting from restricting new leverage to addressing存量 risk. Current South Korean policies focus on suspending new products and raising investor门槛. If后续 further measures like强制降杠杆 for杠杆 products, strengthened liquidity provision mechanisms, market maker承接 arrangements, or more direct market stabilization measures appear, it would signal policy开始处理存量风险.
2. Fund Flow Signal: Significant declines in杠杆 ETF and散户信用资金规模. Substantive contraction in杠杆资金规模 is the core basis for judging if passive selling is接近结束. Measuring South Korean market leverage水位 requires cross-verification from three dimensions: First, product leverage level – tracking asset规模 changes in leveraged ETFs on Samsung and SK Hynix; Second, retail leverage level – whether South Korean individual investors'信用交易余额 is retreating from highs; Third, external funds level – whether foreign investors' net selling规模 in South Korean markets is narrowing.
3. Market Signal: Core memory stocks no longer making new lows on negative news. First, at the core stock level – whether SK Hynix and Samsung continue making new lows facing negatives like earnings revisions or memory price concerns; if股价 becomes钝化 to利空 and不再突破前低, it suggests卖压 is衰竭. Second, cross-market pricing level – whether the premium between South Korean本土股票 and U.S. ADRs converges; narrowing溢价意味着两地定价回归理性. Third, sector relative strength level – whether the memory sector stops持续跑输 other computing power sectors (e.g., optical modules, PCBs); if相对收益企稳, it indicates杠杆抛压 centered on memory is减弱. Fourth, cross-market contagion level – when South Korean markets大跌, whether U.S., Japanese, and A-share memory sectors no longer同步下跌. Only after these four signals are collectively confirmed can one judge de-leveraging is进入尾声.
Synthesizing the above, we believe the nature of this adjustment is risk aversion transmission rather than a shock from liquidity or major macro/fundamental factors. The market lacks conditions for a direct bear turn, and the correction instead provides a window for bargain-hunting.
First, the impact on A-shares is primarily sentiment transmission, not direct冲击 from leverage unwinding. The epicenter is South Korea, rooted in passive liquidations of its retail triple leverage. A-shares lack a similar leverage risk structure and large-scale forced selling pressure; the impact is mainly transmitted via Asia-Pacific market恐慌情绪 and cross-market risk aversion收缩, making its intensity and duration naturally weaker than the震中 market.
Second, domestic macro liquidity conditions are sound and趋向宽松. Domestic monetary conditions remain accommodative, with bond yield中枢持续下移. The decline in无风险收益率 continuously lifts the relative配置价值 of equity assets, providing solid下方支撑 from the macro liquidity layer.
Third, a policy window is临近, with托底预期强化. The late-July Central Politburo meeting is approaching, where稳增长 policies are expected to be further加码, with消费支持政策值得重点期待.发酵 of policy expectations will form an upward对冲 for market risk appetite.
Current Sector Allocation Strategy
1) Innovative Drugs: Relatively independent industry catalysts, with BD transactions maintaining high景气. Innovative drugs have global industry attributes, with relatively low correlation to South Korean memory de-leveraging. Global innovative drug R&D, licensing, and commercialization are primarily driven by Chinese, U.S., and European firms, with sector pricing more dependent on clinical data, overseas licensing, and product放量. South Korean tech leverage unwinding mainly causes indirect扰动 via risk aversion, unlikely to alter innovative drugs' own产业趋势.
BD transactions in innovative drugs延续高景气 in the first half of 2026, with头部药企收入同步兑现. The total BD transaction value for H1 2026 is already接近 2 times the full-year 2024 value and about 73% of the full-year 2025 value, covering热点 like ADC,双抗, and小分子. In Q1 2026, five leading innovative drug companies合计实现营业收入 207 billion yuan, a year-on-year increase of about 23%. The景气 in BD deals and龙头收入增长 mutually印证, indicating the sector still has产业催化 and fundamental support.
2) High-Dividend Sector: Remains an important short-term defensive direction. As market risk appetite declines, high-dividend sectors like banks, coal, utilities, and oil & petrochemicals possess strong波动对冲 attributes. Since July, sectors like coal, banks, and oil & petrochemicals have led gains, showing a trend of资金 switching from high-valuation growth sectors to low-valuation defensive sectors.
Amid the domestic low-interest-rate environment, the relative配置价值 of stable dividend-paying assets remains high. The current 10-year government bond yield is运行 near 1.73%-1.74%, while the CSI红利指数 dividend yield is around 4.87%-5.43%, resulting in an股债息差 exceeding 3.5 percentage points—at an极阔区间 above the 90th percentile of the past decade. Against the backdrop of a存量博弈格局难以改变短期, the relative advantage of high-dividend红利 will remain凸显.
3) Technology: Awaiting a major反弹 after de-leveraging ends. This tech stock adjustment stems more from leverage and risk aversion冲击; the demand, price, and profit logic of the memory industry has not逆转. On the demand side, AI客户 demand for HBM, high-capacity server DRAM, and enterprise SSDs remains旺盛, with clients locking产能 via long-term agreements;短期供给约束尚未明显缓解. On the price side, after前期快速上涨,部分现货涨势趋缓 but报价 remain高位, with contract prices继续受到紧供给支撑. On the profit side, the latest revenue同比 growth for Micron, SK Hynix, and Samsung was 346%, 198%, and 129% respectively,验证 high景气 in龙头业绩. Therefore, the memory industry still has明确的需求与价格支撑; the current股价回调 is more a交易因素扰动 than an产业趋势逆转.
Based on different stages of South Korean de-leveraging, we divide tech allocation into two phases.
Before South Korean de-leveraging ends, tech allocation should shift from high-elasticity上游 to computing power mid-to-downstream and AI applications less affected by it. At this stage,适度控制仓位 in pure memory and upstream hardware strongly linked to South Korean tech stocks is advisable.优先保留 positions in Hong Kong-listed internet stocks and computing power service-related companies, which often can transfer costs and increase profits through price adjustments, offering greater盈利弹性. Additionally,关注 application directions closely结合 with AI is recommended. These sectors' fundamentals depend more on domestic computing power demand, model调用量, user payment, and商业化进展, with lower direct exposure to the South Korean memory leverage chain, potentially exhibiting stronger相对韧性 during de-leveraging.
After South Korean de-leveraging stabilizes, tech仓位 can be gradually increased, with upstream computing power有望 becoming a反弹弹性较大的 direction. When South Korean杠杆 ETF规模下降, SK Hynix不再创新低, and the memory sector's相对收益企稳,适度关注 segments with a涨价逻辑 like MLCC and PCB upstream materials becomes viable. The memory industry still has明确业绩支撑, with Micron, SK Hynix, and Samsung's latest revenue同比增长 at 346%, 198%, and 129%, and Micron's non-GAAP毛利率 at 84.9%, SK Hynix's operating profit margin at 72%. Coupled with旺盛 AI客户 demand for high-end memory like HBM and仍然偏紧 supply capacity, the memory sector有望体现较大的修复弹性 after passive selling ends.
Key risks to monitor include: 1) Domestic demand support policies proving less effective than预期; 2) Escalation of Middle East geopolitical tensions; 3) U.S. stock market volatility exceeding预期.
Comments