When about 60% of institutional holdings are related to AI

Since October, market volatility has increased, but the success rate of market timing has not been high. The underlying reason is that the structure of incremental capital is changing, with prudent absolute return funds continuously entering the market, reducing the effectiveness of traditional aggressive timing strategies. The truly important variables at present remain the stability of the environment for corporate expansion abroad and AI, which involves China-US relations and the progress of AI infrastructure construction. Currently, not only the TMT sector but also the rise in non-ferrous metals, chemicals, and new energy is directly or indirectly influenced by the AI narrative, and the combined institutional holdings in these sectors now exceed 60%. In this case, the idea for portfolio reallocation is not to deliberately avoid the AI narrative, but to choose stocks with an upward trend from the bottom in ROE as much as possible. The AI narrative merely influences the slope of market movements, not the trend.

Since October, market volatility has increased but timing success rate is not high

Since October, the market has experienced two rounds of emotional fluctuations. The first occurred at the beginning of October when Trump issued a new round of tariff threats, causing active funds to quickly reduce positions and a rapid decrease in the proportion of transactions by speculative capital. The market’s daily turnover dropped from 2.5 trillion yuan to 1.7 trillion yuan within two weeks. Subsequently, after China-US engagement and before the leaders’ meeting, the market broke through 4,000 points. The second round happened after the China-US leaders’ meeting. Many active funds believed that the long-term outlook for China-US relations is highly uncertain, so every new agreement or meeting could be a temporary emotional peak. Coupled with the market’s high level approaching year-end, it’s an opportunity to realize returns, leading to a second round of reduction in positions by active funds. However, in reality, the number of stocks hitting new highs has actually increased. As of November 6, the number of stocks reaching a 12-month high was 232, with increases in sectors such as power equipment, public utilities, machinery, transportation, and steel, while TMT saw a clear decrease. By comparison, on September 30, the number of stocks reaching a 12-month high in the two markets was 216, lower than the 232 on November 6. In terms of stocks reaching a new high in the past month, there were 384 on September 30 compared to 680 on November 6. From this perspective, there are still many structural opportunities in the market, and the success rate of short-term market timing is not high. The differing experiences of various investors are mainly due to portfolio structure.

The continuous entry of prudent absolute return funds is reducing the effectiveness of traditional aggressive timing strategies

1) The amount of capital entering the market through prudent income-oriented products continues to increase, while traditional active funds lack sensitivity to the behaviors of these investors. "No chance when the main theme cools down" has been a basic principle for active timing, a pattern that was effective from 2022 to 2024, but its effectiveness has clearly declined this year as a broader investor base continuously increases allocations to equities. With deposit rates, bank wealth management yields, and money fund rates all declining, the low-risk preference groups are facing an "asset shortage" problem, forcing them to extend duration (e.g. longer-term deposits) or choose new investment directions. These groups may indirectly participate in the equity market through insurance, fixed income plus, bank wealth management, or annuities at relatively low proportions. For example, insurance companies’ original premium income reached 5.21 trillion yuan in the first nine months this year (up 8.8% YoY). If 30% of new premiums are invested in A-shares as required by the "Implementation Plan to Promote the Entry of Long-term Capital into the Market," theoretically this could bring an incremental 1.56 trillion yuan. For comparison, active public funds newly issued about 109.5 billion yuan in the first nine months, passive products about 327 billion yuan, private equity new filings 343.1 billion yuan, and cumulative net outflows from stock ETFs totaled 33.9 billion yuan—significantly smaller than the theoretical amount from insurance. In addition, forms like fixed income plus, bank wealth management, and annuities allocate very little to equities but have large total capital, and differ greatly in information and strategy from traditional active funds; however, even slight fluctuations in their equity positions are now having an increasingly significant impact on short-term market fluctuations. These product types, which emphasize drawdown control and net value stability, match well with prudent income client needs, and their future growth ceiling is likely much higher than traditional subjective long-only products.

2) Passive tools represented by ETFs are exhibiting some countercyclical fluctuation features, to some extent smoothing out index-level volatility. The countercyclical nature of ETF subscription and redemption is also dampening short-term market timing effectiveness. A counterintuitive fact: excluding the approximately 200 billion yuan of broad-based ETF purchases by Central Huijin in April, cumulative net outflows from broad-based ETFs this year have been about 510 billion yuan (with net inflows of sector/theme ETFs at 308.3 billion yuan), which is a stark contrast to the feeling of a structural bull market in previous months. Looking at the funds flow in existing stock ETFs since 2018, comparing three-month cumulative net purchases to the three-month rolling return of the Wind All A index, ETF net purchases are basically negatively correlated with market movement—subscriptions rise when the market falls, and redemptions rise when the market climbs: a “buy-on-dip, sell-on-rise” countercyclical pattern. In just October, there were two trading days with net ETF inflows over 20 billion yuan (27 billion on October 10, 22.7 billion on October 31), both occurring during market corrections. Funds that exited during exuberance (June–August, net ETF outflow of around 118 billion yuan) are gradually returning during cooling periods (September–October, net ETF inflow of 95.1 billion yuan).

The truly important variables for market trend are China-US relations and the progress of AI infrastructure construction

1) The share of overseas business of listed companies keeps rising, and fundamentals are becoming less tied to domestic economic cycles. Taking the Shanghai Composite Index as an example, the non-financial sector’s consumption and real estate chain market cap is now less than 13%. The sample stocks’ overseas revenue share in non-financial sectors is now close to 20%. It’s clear that the A-share market is now more influenced by global fundamentals, Chinese companies' overseas expansions, and China-US relations. Our overall view regarding China-US relations is that before the US midterm elections, there may be frequent confrontations, but the conflict is controllable. In other words, before November next year, the chance of China-US relations causing volatility like in April this year is very low and can be temporarily ignored.

2) The sustainability of AI infrastructure is now crucial not only for the US market but also for A-shares. Public funds’ allocation to TMT has exceeded 40%. In reality, the recent rise in electric equipment, new energy, non-ferrous metals, and chemicals is also related to global AI infrastructure construction—such as the recent performance of power equipment, inverters, energy storage, energy metals, and phosphorus chemicals. If we roughly consider the beta drivers of non-ferrous, chemicals, and new energy sectors as coming from the AI industry trend, then institutional investors’ exposure to AI is expected to reach 60% of allocations by Q3 2025 (i.e., counting TMT + new energy + non-ferrous + chemicals holdings). This single beta-driven exposure would exceed the allocation to core assets by public funds in 2021. The market is increasingly discussing and concerning itself with the sustainability of North American AI infrastructure. For example, recently the CDS spreads of North American big tech have significantly expanded. From September to November, the CDS (Credit Default Swap) of Oracle, Google, and Microsoft all rose by over 30%, while the quote for North American investment-grade CDS dropped by 3%, showing that credit traders are betting on deteriorating repayment capability of these companies. Essentially, the market is questioning whether AI can yield enough commercial returns to support ever-increasing data center construction costs, especially with power shortages looming. Therefore, now the greatest influence factor for A-shares is actually in sync with US equities: the commercialization progress and sustainability of AI infrastructure investment; only when this expectation fluctuates enough will market timing become meaningful. At present, however, there’s no clear answer, and industrial changes need close watching.

If TMT, non-ferrous, chemicals, and new energy are all influenced by AI beta, how should allocation be adjusted?

We believe that in future portfolio adjustments, the basic principle should be to focus on sectors with independent logic beyond the AI narrative, whose ROE is in long-term cyclical uptrend from a low base. From this angle, appropriately increasing allocation to chemicals, non-ferrous metals, and new energy—sectors that have been quiet for long and whose industry prosperity and margins are at historical lows—would be a better choice. We believe that the spillover effect of the AI narrative only affects the slope of sector performance in phases; if the AI narrative fluctuates in the future, it may indeed cause volatility in these sectors, but as long as ROE continues to rise from the bottom, share price volatility remains short-term and will not threaten principal safety (compared to sectors with historically high ROE). So, the core issue is not to worry about why these sectors are rising, but to consider whether, absent the AI narrative, they still have medium-term independent upside potential; and to evaluate whether current valuations, assuming margins recover, offer sufficient investment potential. We’ve discussed this in "A-share Strategy Focus 20250817—Domestic Supply Reforms, Overseas Demand Fuels Profits," “A-share Strategy Focus 20251012—Opportunities in Traditional Manufacturing,” and “Strategy Focus 20251026—Market Calms, New Clues for Allocation.” The key is the repricing of China’s manufacturing sector. If you just care whether these sectors are trading higher because of an AI story in the short term, it’s hard to allocate ahead of time and you’re reduced to explaining after the fact. Lastly, the consumer sector, as a relatively independent variety, now has a sufficiently low market cap share (currently only 7.5% in non-financial samples of the Shanghai Composite, near the low of 2015), but the timing and scale of new catalysts are worth close tracking.

Authors: Qiu Xiang, Liu Chuntong. Source: CITIC Securities, original title: "Strategy Focus｜When Around 60% of Institutional Holdings Are Related to AI"

Risk Warning and DisclaimerThe market carries risks, and investment should be made with caution. This article does not constitute personal investment advice and does not account for the specific investment objectives, financial situation, or individual needs of any particular user. Users should consider whether any views, opinions, or conclusions contained herein are suitable for their specific circumstances. Investment decisions based on this article are at your own risk.