In 2026, global markets have exhibited highly divergent structural trends. Equities with exposure to artificial intelligence, such as the Korean KOSPI, the Philadelphia Semiconductor Index, and China's STAR and ChiNext boards, have significantly outperformed. In contrast, sectors without AI exposure, like China's consumer staples and the Hang Seng Tech Index, have notably underperformed.
However, this extreme divergence has also led to extreme crowding and even high leverage, which, triggered by recent bubble concerns, has directly resulted in a "deleveraging" turmoil. Market panic is evident, and volatility has risen markedly. Last week, technology stocks in Korea, the US, and China experienced significant adjustments sequentially. The Korean KOSPI fell 9.0% and 6.4% on Monday and Thursday, respectively. The Philadelphia Semiconductor Index dropped nearly 10% for the week, marking its largest weekly decline in nearly a year. Last Friday, China's ChiNext Index and the STAR 50 Index fell 7.2% and 7.1%, respectively, with the ChiNext Index experiencing a weekly pullback of approximately 10.8%.
The recent turbulence has led investors to question whether the current extreme structural market has reached its end and how it will ultimately conclude. History shows similar extreme structural markets, but their final outcomes have varied, broadly falling into three categories: diffusion, collapse, and rotation. So, what stage are we in currently? How can we determine the direction of its future evolution? This article addresses these questions.
Origins of the Current Extreme Structural Market
To judge the inflection point of a structural market, one must first understand how it forms and why it persists. The current extreme structural market is a direct reflection of the divergence in credit cycles across different sectors. Measured by the credit impulse of enterprises and households, the former has continued to rise due to increasing AI investment, becoming the main driver of the economy. The latter, however, has continued to weaken as policy focus shifts towards technology, even falling back to lows seen before the "September 24th" period.
The K-shaped divergence in the Chinese economy intensified in the first half of the year. On one hand, credit expansion in the AI industry chain accelerated. Global AI demand and capital expenditure re-accelerated in the first quarter, and domestic policy support for the AI industry chain increased simultaneously. In the first half, the cumulative year-on-year growth of industrial added value in high-tech industries was 13.3%, and investment in information services grew 15.5% year-on-year. The AI industry also provided significant support for Chinese exports. As of the first quarter of 2026, AI industry chain exports accounted for 20.6% of China's total exports, and the cumulative year-on-year growth of mechanical and electrical product exports in the first half was 24.5%. In GDP terms, the information technology sector contributed over 50% to second-quarter GDP.
On the other hand, traditional domestic demand has weakened further. Weak household income expectations and the fading effect of previous subsidies have led to insufficient willingness among households to increase leverage for consumption and home purchases. In the first half, total retail sales of consumer goods grew only 1.3% year-on-year, while the stock of medium and long-term household loans fell 6.3% year-on-year. Correspondingly, in the credit cycle, the corporate credit impulse is rising, while the household credit impulse has fallen back to pre-"September 24th" levels. This divergence is directly reflected in the market, manifesting as capital chasing the AI industry, while the consumer sector continues to decline, falling back to pre-"September 24th" levels.
However, this extreme divergence is not immutable. Consumption has not always been weak. After the "September 24th" period in 2024, benefiting from fiscal stimulus and policies like trade-in subsidies, the private sector credit impulse rebounded from its bottom, and consumer-related assets rose rapidly. The CSI Consumer Index rose 24.5% in the week of "September 24th," significantly outperforming the broader market indices. However, as subsidies faded and there was no sustained improvement in income expectations or the property market, the household sector credit impulse turned negative again after June 2025. Technology has not always been strong either. In the fourth quarter of last year, the market was once concerned about an AI bubble, leading to a temporary cooling of technology trading in US and domestic markets, with funds rotating to other high-growth areas like non-ferrous metals. But the acceleration in AI capital expenditure, cloud service revenue, and enterprise-paid demand once again strengthened expectations for credit expansion in the AI industry chain. Major cloud providers' cloud service revenue exceeded $100 billion in the first quarter, a 36% year-on-year increase, causing the previously loosening structural trend to reconcentrate on the AI theme.
History shows three similar periods of extreme structural markets, all underpinned by divergence in credit cycles. Although the specific industrial directions differed in these three rounds, they all occurred against a macroeconomic backdrop of flat or weakening overall credit cycles with significant internal structural divergence:
The "Nifty Fifty" from mid-2016 to early 2018: The overall credit cycle weakened, with the household credit impulse strengthening and the corporate credit impulse weakening, reversing later. Monetary compensation for shantytown redevelopment supported household sector leverage, while tightening financial regulation and structural deleveraging suppressed corporate financing willingness. After early 2017, the increment in medium and long-term household loans fell from high levels, and corporate credit marginally recovered. Amid the persistent divergence between corporate and household credit impulses, funds preferred consumer, financial, and manufacturing leaders with higher earnings certainty.
The Semiconductor cycle from 2019 to early 2020: The overall credit cycle was flat, with the household credit impulse declining and the corporate credit impulse recovering. In 2019, policy emphasized "not using real estate as a short-term stimulus for the economy," strengthened targeted support for the real economy, guided credit resources to manufacturing, private enterprises, and small and micro enterprises, with industrial policies favoring import substitution and 5G construction driving semiconductor景气.
The Power Equipment and New Energy cycle from 2021 to 2022: The overall credit cycle declined, with the household credit impulse peaking and falling, and the corporate credit impulse stabilizing and recovering. After pandemic disruptions, policy support led to an overall credit uptick. In 2021, the overall credit cycle weakened, with the household credit impulse falling due to property credit contraction. However, strong exports supported corporate order demand and debt repayment capacity, leading to a stabilization and recovery of the corporate credit impulse. Factors such as green industry policies, technological advancement, export demand, and increased penetration rates made credit resources favor the new energy sector, making it the structural主线.
Identifying the Inflection Point in a Structural Market
Using the excess return of the market theme relative to the broad market index as a reference and taking the peak of excess return as the basis for the structural market inflection point, we analyze the three historical structural markets: the 2017-2018 Nifty Fifty, the 2019-2020 Semiconductor cycle, and the 2020-2022 Power Equipment and New Energy cycle. Historical experience shows that high trading volume and positioning can serve as signals of risk accumulation and market fragility. However, a narrowing relative earnings growth and ROE advantage, accompanied by cash flow pressure and rising interest-bearing debt ratios, are needed to confirm the end of the trend.
Trading Crowding can explain market volatility but is difficult to reliably identify market tops. We use "theme trading volume share / theme free-float market cap share" to measure trading crowding. An indicator above 1 means the theme is receiving more trading resources than its market weight. A持续上升并进入历史高分位 indicates increasingly homogenized trading direction. The stronger the market consensus, the more marginal buyers rely on price momentum. Once negative signals appear, funds may simultaneously reduce positions, amplifying price adjustments.
However, trading crowding has limitations in identifying trend direction: 1) There is no stable absolute historical threshold to predefine what level of crowding represents the final top of a structural market. 2) Multiple阶段性交易拥挤高点 may occur within one structural market cycle; early peaks often correspond only to short-term volatility and do not signify the end of the trend. 3) The final trading crowding peak may not necessarily be higher than previous ones; generally, only after excess returns持续回落 can it be confirmed retrospectively that a阶段性高点 was the "last one." Therefore, trading crowding is difficult to use directly as a leading signal for a structural market top.
In historical samples: 1) For the 2016-mid-2018 Nifty Fifty, trading crowding peaked slightly before relative收益 in late 2017, with price adjustments reflecting trading and valuation pressure. 2) In the 2019-2020 Semiconductor cycle, trading crowding reached阶段性高点 in December 2019 and late January 2020, but excess returns did not peak until February 2020. 3) In the 2021-2022 Power Equipment and New Energy cycle, trading crowding also surged multiple times,明显抬升 one to two weeks before the two peaks in excess returns.
In this context, trading crowding is more suitable for judging fragility and tail risks. High-crowding assets may still achieve excess returns, and overall volatility may not同步升高. When fundamentals are disproven or liquidity contracts, highly concentrated positioning amplifies risks, and leverage further strengthens this mechanism. For judging structural market拐点, crowding and leverage alone are insufficient to determine the medium-term direction but can affect the magnitude of adjustments.
Weakening Performance Indicators appear with a lag. While difficult to time accurately, they can be used to confirm the end of a trend. The basis for the persistence of a structural market lies in the theme's盈利 relative to the broad market index maintaining an advantage. This signal consists of two sets of indicators: 1) Relative earnings and ROE indicators confirm the direction. We use the theme's expected year-on-year earnings growth minus the broad index's expected year-on-year earnings growth, and the theme's expected ROE minus the broad index's expected ROE. When these differences peak and持续下行, the basis for the market paying a valuation premium for the theme begins to erode. 2) Free cash flow / capital expenditure and interest-bearing debt ratio indicators supplement the judgment of the quality of the延续产业趋势. For capital-intensive industries, capital expenditure持续高于 operating cash flow means companies rely on external financing to maintain expansion. If revenue growth cannot cover new investments, risks of future supply-demand imbalance and price competition rise. Technology manufacturing trends typically end their main上升浪 when free cash flow turns negative and deteriorates rapidly.同时, after the market peaks, the interest-bearing debt ratio usually明显上行, indicating the industry's expansion shifts from internally generated cash flow support to debt-driven, increasing financial risk.
Historically, these indicators have a certain lag and can be used to judge whether the industrial logic has been disproven after excess returns retreat. After the Nifty Fifty's excess return peaked in early February 2018, the leading companies' earnings advantage lasted about two weeks before weakening. Subsequently, declining cash flow quality and rising interest-bearing debt ratios in the first quarter further confirmed the trend拐点. After the first peak in excess returns for Power Equipment and New Energy in 2021-2022, earnings remained high, leading to a second top. After the second peak, earnings expectations were持续下修, and excess returns转为持续转弱.
Additionally, external shocks can提前打断 a structural trend, triggering an excess return拐点 directly before盈利 confirmation. The 2019-2020 Semiconductor cycle can be considered a case. During the initial pandemic phase, risk appetite and global liquidity contracted, leading to a significant adjustment in semiconductor excess returns. However, industry demand and import substitution trends did not disappear, and earnings expectations did not明显恶化. Subsequently, as pandemic disruptions subsided, policy easing provided support, and orders gradually recovered, the sector regained excess returns. Therefore, to judge whether an external shock will end a structural trend, one must observe its transmission path. If the shock mainly affects valuation and liquidity, it is closer to a temporary price adjustment. If the shock simultaneously suppresses terminal demand and corporate profitability, the trend may pause or end提前.
Three Possible "Outcomes" for a Structural Market
Based on the above analysis, if the original theme trend has not ended, that is one scenario. If the original theme's excess return has peaked阶段性地, according to historical experience, there may be three different "outcomes": diffusion to the overall market, total collapse, or rotation to a new structure. To judge the direction of evolution, the credit cycle framework can still be used: first, whether the credit expansion of the original theme can延续; second, whether other sectors will获得 credit expansion opportunities.
Diffusion: The original theme's excess return declines, and more industries gain the possibility of credit expansion, usually requiring improvements in both fiscal expenditure and private credit. The core of diffusion is that the credit cycle shifts from局部修复 to broader improvement, manifested as upward修复 in both the广义财政赤字脉冲 and the private sector credit impulse. Improved earnings expectations expand from the original theme to more industries. Under this path, the original theme may not necessarily fall, but its excess returns will gradually narrow, and market breadth will明显扩大. For example, after Covid-19 disruptions减少 in 2020, with policy support, both广义财政赤字 and private sector社融脉冲同步走高, economic momentum recovered. Semiconductors, as the original theme, continued to rise, but consumption, healthcare, cyclical sectors, and new energy also saw earnings revisions upward, and the market gradually shifted from a单一结构 to broad-based上涨.
Collapse: The industrial trend of the original theme is disproven, and other industries are unable to take over,通常对应 the credit cycle turning contractionary. As the weakest of the three paths, collapse requires two conditions同时出现: reversal of the original theme's industrial trend and credit contraction; the overall credit cycle同步下行, with other industries also lacking new demand, policy, or financing support. In this scenario, high crowding and high leverage accelerate capital flight from the theme and transmit to the entire market through risk appetite and liquidity, evolving from a structural adjustment to a普遍下行. The market performance after the Nifty Fifty peak in early 2018 shares certain similar characteristics. Leading companies' relative earnings advantage declined, both广义财政赤字脉冲 and private sector credit脉冲同步向下, and external trade摩擦 increased uncertainty, putting pressure on the overall market.
Rotation: Weak overall credit cycle不支持普涨, and funds concentrate on a few directions with expansion potential. Rotation occurs when overall credit remains偏弱, and the market lacks a comprehensive盈利修复 environment. Funds do not have the conditions for broad diffusion but choose among limited景气 directions. Depending on whether the original theme's credit expansion conditions are broken, rotation can be further divided into two categories:
The first is阶段性调整: The original theme's industrial trend remains, and other industries temporarily outperform. When the original theme is过度集中, valuations are high, or impacted by external liquidity shocks, funds may flow阶段性地 to low-crowding, low-valuation directions. However, these industries do not possess持续 credit expansion logic. As valuation pressure on the theme释放 or new catalysts emerge, funds often return to the original direction that still possesses credit expansion potential. After the first retreat of the Power Equipment and New Energy trend in late 2021, the upward盈利 trend did not end. From April to July 2022, the new energy sector rebounded significantly again, exhibiting characteristics of an intact industrial trend with阶段性调整.
The second is新主线接棒: The original theme's industrial trend weakens, and a new direction with credit expansion potential emerges. If the original theme's盈利预期 or industrial catalysts持续走弱, while another direction gains more policy support or upward盈利 revision expectations, the market theme gradually shifts. The evolution of the Power Equipment and New Energy trend after the second half of 2022 partly reflects this characteristic:盈利预期持续下修, the market began to focus on the sustainability of the sector's credit expansion, the original theme此后持续走弱, weak overall credit was insufficient to推动普涨, funds sought new expansion logic among few directions, and gradually shifted to new themes like "China-specific valuation" and AI/TMT after 2023.
Outlook for the Current Structural Market
Returning to the initial questions: Has the current extreme structural market reached its end? How will it ultimately conclude? Rising trading crowding has amplified market volatility, but this alone cannot判断 the trend has ended. A more effective analysis is to examine the成立条件 for the three paths of diffusion, collapse, and rotation. We believe conditions for全面扩散 and collapse are both insufficient. Rotation is more likely to manifest as阶段性调整 in the original theme awaiting catalysts.
First,全面扩散? The threshold is high. Fiscal节奏 may improve in the third quarter, but total新增 is limited, and the household credit impulse may be难以大幅修复, making it difficult for the K-shaped divergence to收敛 quickly. Diffusion requires overall credit cycle expansion, and the AI technology theme can further带动 multiple sectors修复 through wealth effect等路径. In the current environment, considering constraints on the household sector from income expectations and property returns, the credit impulse is难以自发快速回升 in the short term. The key lies in whether fiscal policy can发挥较强的带动作用.
Regarding fiscal policy: 1) In terms of规模, the annual fiscal总量增量 is limited, still mainly focused on托底. Based on the Government Work Report, the 2026 deficit ratio is arranged at around 4%, with a deficit size of 5.89 trillion yuan, an increase of 230 billion yuan from the previous year;超长期特别国债 of 1.3 trillion yuan, and local government special bonds of 4.4 trillion yuan. Overall, the 2026广义财政赤字率 is 8.1%, and the赤字率拉动总需求 is 7.3%, both slightly lower than 2025. 2) In terms of节奏, the广义财政赤字脉冲 in the third quarter may improve compared to the second quarter. Fiscal融资节奏 in January-June 2026 was偏慢,同比少增 about 1.2 trillion yuan, leaving some room for发力 in the second half. 3) In terms of方向, fiscal policy still emphasizes structural support, with resources偏向 "Two Major Projects," new quality productive forces, and industrial upgrading.
Therefore, the credit cycle may阶段性地向上修复 in the third quarter but does not change the overall震荡格局 and will retain the original structural特征 of strong technology and weak domestic demand. We believe the technology sector is an important落脚点 for structural policies and is expected to continue receiving support. As long as external demand and employment do not明显失速, the necessity for a significant shift in fiscal重心 towards traditional domestic demand to推动全面再杠杆 remains limited.
Second,整体坍塌? This is also unlikely. The credit impulse may阶段性地修复 in the third quarter, and the AI industrial trend has not been系统性地证伪. The condition for collapse is an overall weakening credit cycle, the credit expansion logic of the original theme being disproven, and other industries lacking credit expansion potential. Currently, the AI industry is not in a全面泡沫. On one hand, demand端仍在扩张, with global token调用量, enterprise付费比例, and cloud service revenue保持增长; the industrial trend has not shown系统性转弱. On the other hand, although the ratio of capital expenditure to operating cash flow for major US cloud providers has risen to around 94%, we judge that继续增加投入 will increase reliance on external financing. However, the ROIC of major cloud providers remains higher than WACC, and starting leverage is significantly lower than during the internet bubble,尚未形成普遍的偿债压力. AI remains a key focus area for both China and the US, a确定性较强的扩张方向 in the credit cycle.
Third,新主线接棒, or旧主线阶段性调整? The latter is more likely. With conditions for both diffusion and collapse insufficient, the market may maintain rotation. Currently, the AI theme's industrial trend remains, while other industries are more in低位修复. After market adjustments,回归 AI主线的可能性较大. First, no new theme has emerged that can接棒. Although some consumer and internet sectors recently反弹 from lows, the magnitude and sustainability of their盈利预期改善 are still相对有限, and fiscal support also lacks the impetus for a全面转向消费. Internet and innovative pharmaceuticals face smaller fundamental阻力 compared to consumption.
Second, the logic for AI theme credit expansion remains. Taking optical communications as an example, although trading crowding reached a阶段性高点 in April 2026, relative业绩仍在上行, and the interest-bearing debt ratio走低 in the first quarter, indicating盈利端仍在兑现. Short-term adjustments can反而降低拥挤度,释放估值压力, and improve asset赔率. Subsequently, if业绩, products, or policy catalysts reappear, the likelihood of funds回归原主线 is high.
Therefore, the current structural market is接近阶段性轮动, and the trend is more likely a暂时调整 rather than走向终结, with technology remaining the主线.
Regarding配置策略, our probability and赔率 models show that at the current juncture, the赔率和胜率 for the Hang Seng Tech Index, Korean equities, and the ChiNext board are边际改善. Internet and innovative pharmaceuticals have相对较好的赔率和胜率.
Technology is a typical high-probability, low-赔率 asset. Similar to reducing positions at highs, entering at lows faces timing issues. The difference is that左侧 may承受进一步回撤, while右侧 may少获取一部分反弹收益. Subsequently, focus on three aspects: First, liquidity risk from杠杆交易 and集中踩踏; stabilizing market sentiment is the短期 priority. If policy intervenes, pressure can often be alleviated更快. Second, monetary policy tightening expectations; the late-July FOMC meeting will provide more information on the利率路径. Third, whether the industrial logic can obtain a new round of业绩催化. AI bubble concerns in July-August 2024 and October 2025 were ultimately消化 through subsequent业绩兑现. Most of these key variables will be集中验证 in July.
For high-赔率, low-probability assets like Hang Seng Tech and innovative pharmaceuticals, the biggest constraint remains opportunity cost. Only fundamental improvement can truly提升胜率. Recent increased volatility in the tech sector has somewhat reduced the opportunity cost of holding Hang Seng Tech. Coupled with some fundamental利好 and narrative catalysts, conditions for低位修复 have improved. However, Hang Seng Tech and innovative pharmaceuticals also明显下跌 last Friday. High-赔率 assets relying solely on low valuations也难以独善其身 during risk appetite contraction. Considering港股's high exposure to the consumer大盘 and the large weight of internet leaders, stronger catalysts are still needed to走出底部 and form持续反弹: one type is类似 "September 24th moment" fiscal发力, the other is类似 "DeepSeek moment" technological breakthroughs.
If concerned about the偏慢 or limited magnitude of catalyst兑现 in tech hardware and希望进行适度轮动 or均衡配置, it is more suitable to focus on directions with相对较小的基本面阻力, such as internet and innovative pharmaceuticals with low consumer exposure. Recent events like the new Kimi model, increased腾讯HY3调用量, Apple's国行版 AI model filing, and the US termination of related Hong Kong administrative orders can constitute阶段性催化. But to接棒, successful fundamental兑现 is still required.