Summary of Key Points
In the second quarter, there was a significant shift in the stock holdings of public funds: AI-related hard technologies made their debut on the top ten lists of actively managed equity funds for the first time, while leading companies in consumer and internet sectors (such as Tencent and Moutai) saw substantial reductions in their holdings. However, these AI leaders experienced a 10%-20% decline in the third quarter, causing those who bought them at high prices to suffer temporary losses. Institutions view this as a "healthy correction" rather than the bursting of a bubble, indicating that the overall trend of the AI industry remains unchanged. Recently, funds have begun to buy into these stocks at lower prices, suggesting that the adjustment may be coming to an end, and AI is likely to remain the main focus of the market in the second half of the year.
I. AI Hard Technologies Take the Lead; Consumer and Internet Stocks Retreat
By the end of the second quarter, the top ten holdings of actively managed equity funds were all core companies in the AI computing power chain: Zhongji Xuchuang (optical modules), Xin Yisheng (optical modules), Dongshan Precision (PCBs), Cambricon (AI chips), and Beifang Huachuang (semiconductor equipment). Among them, Zhongji Xuchuang became the largest holding in public funds, with a market value of 166.2 billion yuan, held by 1,816 funds, and its stock price increased by 123% during the quarter; Xin Yisheng ranked second with a market value of 136.6 billion yuan and a 92% increase in stock price.
In contrast, once-popular consumer and internet leaders withdrew significantly from the top ten list: Tencent dropped from a prominent position in the first quarter to 28th place, and Moutai fell to 30th place. Among the top 50 holdings, only Moutai, Midea, and Zijin Mining remained from traditional industries. The largest reductions were seen in Tencent (-18.7 billion yuan), Moutai (-14.6 billion yuan), and CATL (-17.5 billion yuan).
China Merchants Securities described this change as a "K-shaped divergence": while the AI computing power sector was favored by funds and saw growth, traditional industries such as consumer goods and new energy faced indiscriminate reductions in holdings. The rationale behind this is clear: the trend of the AI industry is strong (for example, there is a surge in demand for optical modules, and there is a shift towards domestic chip production), while the long-term earnings prospects for traditional industries have been downgraded.
II. AI Leaders Decline in the Third Quarter; Investors Who Bought at High Prices Are Temporarily Losing Money but Have Seen a Rebound
AI stocks rose too rapidly in the second quarter, leading to a slowdown in the third quarter. As of July 30, five AI leaders, including Zhongji Xuchuang and Xin Yisheng, had an average decline of 18.6%, with Cambricon experiencing the steepest drop (21.7%). Investors who bought these stocks at high prices in the second quarter have suffered temporary losses (for example, their holdings decreased by an average of 25.8% from the peak to the bottom on July 21).
However, the situation has improved recently: as of July 23, these stocks rebounded by 5.5%-12.8%, with Zhongji Xuchuang and Beifang Huachuang rebounding by more than 12%, indicating that some funds have started to buy in at lower prices.
III. Institutions See This as a "Healthy Reset," Not the Bursting of a Bubble—Three Key Reasons
Why aren't institutions panicking?
1. Leverage Reduction and Risk Release: JPMorgan Chase believes that this correction is a result of leverage reduction. The amount of funds used for margin trading has decreased, with the balance dropping from 3 trillion yuan to 2.72 trillion yuan, and the proportion of financing transactions in the IT industry falling from 12% to 8%-9%. The most leveraged parts of the market have already exited, unlike during the 2000 U.S. internet bubble when leverage ratios reached as high as 230% (currently, cloud service providers have leverage ratios around 40%).
2. Insufficient Production Capacity for AI Hardware: There will be a shortage of production capacity for AI chips and optical modules by the end of 2027, forcing companies to invest in expansion, indicating continued demand.
3. Unchanged Industry Trends: Institutions like CICC and UBS believe that the overall direction of the AI industry has not changed; the rapid growth in recent times simply required a period of adjustment.
IV. Underlying Financial Trends Indicate an Adjustment is Nearly Over
Financial signals suggest that the adjustment may be coming to an end:
- ETF Inflows into Technology Stocks: Since July, there have been significant net inflows of funds into technology and hardware-related ETFs, with 142 billion yuan flowing in on July 17 and 20 alone. On July 21, semiconductor ETFs even hit daily limits, indicating that investors are betting against the trend.
- Leverage Reduction Completed: The balance of margin trading funds has been declining for 13 consecutive days and has recently stabilized, reducing the pressure from leveraged investments.
V. AI Remains the Main Focus for the Second Half of the Year—Institutional Consensus is Clear
Many institutions agree that AI will continue to be the core driver of the market in the second half of the year:
- UBS: The technology sector is expected to see strong profit growth, with ongoing capital inflows. The space for public funds to overweight AI has not yet reached its peak (historically, it takes about three years for public funds to significantly increase their allocation to a sector).
- CITIC Securities: After the electronics sector adjusts, sub-sectors such as semiconductor equipment, AI components (optical modules, PCBs), and value-added chains (storage chips) have solid fundamentals and room for recovery. Global demand for AI will support high growth in sectors like PCBs, storage, and AI power supplies.
In summary, the recent decline in AI stocks is a temporary pause, but the long-term trend remains positive. The focus for the second half of the year should still be on AI-related hard technologies.