The recent sharp decline in AI-related stocks is more a matter of capital positioning than a deterioration of the industry's fundamentals, according to market analysis. This correction stems from a clash between public fund managers and absolute-return investors amid high market crowding, compounded by the ongoing contest for influence and profit distribution within the global AI supply chain.
Industry fundamentals, however, remain solid. Recent earnings from Broadcom have confirmed that capital expenditure on computing power has not cooled, while the profit pool is beginning to shift downstream toward software and platform layers. The pricing focus is expected to transition from "price increases" to "volume growth," with domestic computing power, PCB, and optical communications potentially outperforming core leading stocks.
Last week, the Shanghai Composite Index fell 0.56%, the CSI 300 dropped 1.33%, and the ChiNext Index declined 4.03%, while the Hang Seng Index edged up 0.26%. Value stocks outperformed growth stocks, with large-cap shares leading gains. Average daily trading volume across the A-share market was 1.9643 trillion yuan last week, a decline from the previous week.
Although the market has stabilized since September, persistently shrinking trading volume has become a notable issue. The market structure continues to show disorderly rotation among low-position sectors, with a barbell strategy favoring dividend-yielding defensive plays and the GZ2000 index proving advantageous. Overall, the market action remains frustratingly choppy.
In such an environment, patience with the market structure is warranted. Historical data shows that the A-share market has a clear main investment theme only about 60% of the time each year. For roughly 40% of the time, there is no clear direction, and about 15% of that period is marked by both an absence of consensus themes and persistently shrinking volume, similar to the current situation.
On the other hand, there is little reason to lose confidence in the broader market. Based on mid-year reports, the fundamental support for A-shares remains robust. The earnings growth rates for the full A-share market and the non-financial, non-petrochemical sectors are projected at 16.82% and 16.22% for the first half of 2026, respectively, a significant improvement from the 7.63% and 13.05% seen in the first quarter of 2026. The non-financial, non-petrochemical sectors have maintained double-digit year-on-year earnings growth for two consecutive reporting periods, largely driven by the structural prosperity of technology and overseas expansion. The combined earnings share of technology and overseas expansion has risen to nearly 50% in the latest reports. Historically, when this proportion exceeds 50%, a new earnings cycle tends to emerge.
From both industry tracking and valuation perspectives, a systematic turnaround for the AI technology sector is anticipated to become clearer in the fourth quarter. When stock prices are at elevated levels, they become naturally sensitive to negative news, a dynamic that was particularly evident during the AI selloff in July. Conversely, during downturns, positive news tends to be overlooked.
Indeed, there has been positive news on the AI industry front recently. Broadcom and Snowflake both reported earnings on September 2. Broadcom demonstrated that upstream capital expenditure on computing power has not significantly cooled, while Snowflake further confirmed that AI commercialization is beginning to shift from infrastructure construction to the enterprise data layer. This suggests the industry profit pool is diffusing downstream rather than signaling an end to upstream prosperity.
At the U.S. stock level, since late June, the "Magnificent Seven" index has risen approximately 10.8%, with Nvidia up 18.2% and approaching its previous highs. This indicates that core computing power assets have not been abandoned. However, the Philadelphia Semiconductor Index has fallen about 14.4%, showing that the market is not exiting AI or technology but rather rotating from broad semiconductor beta toward fundamentally solid core hardware leaders and software/platform layers. The real pressure is on second- and third-tier semiconductor and hardware chain companies that rose alongside AI capital expenditure but lack strong earnings support.
Given this context, the essence of the recent AI technology correction should be judged from the perspective of capital positioning rather than industry logic damage. Externally, this is an intense conflict between the U.S. and non-U.S. nations over the voice and profit distribution in the AI industry chain. Internally, it is a confrontation between absolute-return funds and public fund managers within the highly crowded AI space. As long as the industry logic remains intact, after a quarter of digestion, the probability of AI making a comeback in the fourth quarter is high.
From a macro-strategy perspective, the current pricing direction of the A-share market is jointly determined by domestic earnings diffusion (K-shaped divergence) and overseas liquidity factors (U.S. Treasury yields). The intersection of these two variables creates four macro scenarios.
Scenario One: K-shaped divergence narrows coupled with rising U.S. Treasury yields corresponds to a cyclical sector takeover. This scenario sees improvements in real estate, consumption, and investment data as Q4 stabilization policies take effect, but rising U.S. Treasury yields suppress growth valuations, and the liquidity environment does not support sustained expansion of high-valuation tracks. The market style would shift toward large-cap blue chips, with cyclical sectors succeeding the technology theme. Positioning would target real estate chains, resource products, and major consumer sectors.
Scenario Two: K-shaped divergence narrows coupled with falling U.S. Treasury yields corresponds to a broad bull market. This is the optimal scenario with positive resonance from both domestic and external factors. Domestic earnings see broad recovery while falling U.S. Treasury yields remove overseas liquidity constraints. Fundamentals and liquidity form a double positive, allowing the market to break through the constraints of a single track. Core assets, cyclical sectors, and technology growth would all benefit, providing the foundation for a comprehensive bull market.
Scenario Three: K-shaped divergence persists coupled with rising U.S. Treasury yields calls for a barbell strategy. Domestic demand recovery falls short, earnings diffusion is hindered, and rising U.S. Treasury yields pressure growth sectors. Under these dual constraints, the market struggles to form a sustained theme, leading to index fluctuations and rapid sector rotation. A barbell strategy is recommended: allocate one end to high-dividend defensive positions such as banks and coal, and the other end to small caps and precious metals for flexible gains, balancing risk amid high volatility.
Scenario Four: K-shaped divergence persists coupled with falling U.S. Treasury yields corresponds to a technology "M-top" pattern. Domestic fundamentals show weak recovery, loose liquidity fails to transmit to real earnings, but falling U.S. Treasury yields release valuation space for growth. The strategy would be to seize technology rebound opportunities.
Currently, the convergence of the domestic K-shaped economy cannot be confirmed, while U.S. Treasury yields remain elevated. The next move depends on the U.S. Federal Reserve's mid-September meeting. Although August non-farm payroll data significantly exceeded expectations, it did not resolve the question of whether inflation necessitates further rate hikes. August CPI data is due on September 11, and its marginal pricing power deserves close attention.
Finally, it bears repeating that until the domestic macro K-shaped divergence narrows, there can be no bull market in the capital markets without technology. The eventual ending tends toward an "M" rather than an "A" shape. It is also worth remembering that after every major selloff or high-to-low rotation, changes occur-some stocks never recover, and the most persistently rebounding stocks are not necessarily those that fell the most.
The decision of whether the main theme shifts is not determined by the magnitude of gains or losses but by whether the industry prosperity of the old theme can be validated across periods (60%-70% of the time leads to a complete change in the main theme, while 30%-40% of the time the same theme persists). Even if the main theme remains unchanged, core stocks will undergo changes. This implies that the subsequent AI pricing focus will shift from "price increases" (e.g., storage, optical fiber, electronic cloth) to "volume growth" (e.g., hardware volume growth sectors: domestic computing power, PCB, optical communications). Second- and third-tier names may outperform core leaders during this phase.
It is worth noting that during the gap period before a new main theme forms, small- and micro-cap indices such as the CSI 1000 and the GZ2000 have historically shown a higher win rate relative to large caps.