The pullback in AI momentum is not the start of a bear market, and non-AI sectors may actually face a more favorable environment, according to research from CITIC SEC (ASX: 06030). The core point is that the AI industry, at least in the short term, does not broadly lift other sectors of the economy the way past industrial trends did. Instead, it has generated considerable negative externalities for other economic sectors, including surging materials and resource prices, soaring interest rates, and crowding out of employment.
This also means the old experience that "the end of an industrial trend rally equals a turn from bull to bear" may no longer apply, and many non-AI sectors have already entered a downtrend. At present, whether it is the second-derivative valuation correction in AI sectors or the valuation contraction in non-AI sectors, the main valuation-shrinking process is expected to be largely completed this year. Looking ahead, after deep adjustments, the probability is higher that divergence will gradually converge and market breadth will expand. The only issue is that the clearing of capital flows needs time, and at this stage a steady-return strategy represented by PB-ROE may be more advantageous.
The pullback in the AI trend is not the start of a bear market
A decline in an industrial trend does not equal a turn from bull to bear. In the A-share market, there has long been a fairly widespread view that "a decline in an industrial trend equals a turn from bull to bear." The core reason is that past industrial trends, such as industry and finance in 2006-2007, real estate and infrastructure in 2009, mobile internet in 2013-2015, and the energy transition in 2019-2021, broadly lifted other sectors of the economy. By contrast, the progress of the AI industry in this cycle, at least in the short term, has produced considerable crowding-out effects on other economic sectors, including causing sharp increases in materials and resource prices, driving tight supply-demand conditions in the bond market and a sharp surge in long-term bond yields, and to some extent crowding out employment. The growth of the wealth effect has been extremely unbalanced, the degree of market divergence is historically rare, and there has even been a so-called opposition between the "silicon-based world" and the "carbon-based world." This also means the old experience that "the end of an industrial trend rally equals a turn from bull to bear" may not apply to this AI-hardware-led rally. Many non-AI sectors have already entered a downtrend, and AI has merely been the last to "catch down."
AI has created negative externalities for other sectors in the short term, and if AI momentum fades, some non-AI sectors may instead see a turnaround
At least three types of negative externalities have already become evident. First, sharp price increases in basic daily necessities such as electronic components and electricity have raised production and living costs and suppressed demand in other areas, with consumer electronics being a typical example. Second, investment and construction of computing infrastructure has pushed private-sector debt financing to enormous levels, driving up global risk-free interest rates and dealing a blow to all demand that is sensitive to higher rates. Third, employment is being crowded out, especially for young people. For example, according to the latest September data from Stanford and the ADP employment database, employment for the 22-25 age group in the occupational group with the highest share of automated AI use fell 10.9% compared with November 2022, while the lowest group rose 7.8% over the same period. These factors have triggered sharp divergence in the real economy and markets. Conversely, if the pace of AI investment slows in the future, non-AI sectors may not face such a negative macroeconomic environment, such as a sharp decline in risk-free rates. In fact, based on current U.S. Treasury pricing, the direction of the 10-year U.S. Treasury yield is highly correlated with the share price trend of North American AI hardware stocks. If there is a sustained decline in the future, it could at least boost demand in overseas non-AI sectors.
The valuation contraction process is basically complete this year, and market breadth is expected to gradually expand
The second-derivative correction in AI and the valuation contraction in non-AI are both expected to be basically completed this year. AI has completed the first wave of "second-derivative" correction. Similar to the new energy rally adjustment from the end of 2021 to the end of April 2022, this round of AI also saw valuations revised down in advance when industry prosperity remained intact but growth expectations weakened. From the perspective of valuation contraction, the valuation compression in this round of AI hardware has already been comparable to the first round of adjustment in new energy in 2022. According to Wind data, since July 1, the aggregate PE (TTM) of the electronics and communications sectors has compressed from 74.0 times to 42.4 times, a decline of 42.8%; during the first round of new energy adjustment from December 17, 2021, to April 26, 2022, the market-cap-weighted PE of the power equipment and new energy sector compressed from 51.2 times to 27.8 times, a decline of 45.7%. From a global comparison perspective, A-share non-AI sectors are in a global valuation trough. We used baskets of AI and non-AI leaders in the Chinese, U.S., Japanese, and South Korean markets, calculated PE (TTM) using the aggregate method, and compared them across markets. As of October 22, A-shares were the only market where the valuation premium of AI relative to non-AI was still above its historical median: the AI/non-AI PE ratio was 4.4 times, at the 82nd percentile since 2016, while this ratio in the U.S., Japan, and South Korea had fallen to the 1st, 17th, and 24th percentiles, respectively. The PE of U.S. and Japanese non-AI non-financial leaders was 33.5 times and 23.0 times, at the 98th and 95th percentiles; A-share non-AI non-financial leaders were only 15.6 times, at the 21st percentile historically, while consumer, manufacturing, and pharmaceutical leaders were only at the 12th, 1st, and 3rd percentiles, respectively. Of course, low valuations stem from weak domestic demand and the market's potential concerns about external demand in non-AI areas, such as trade friction, high overseas interest rates, and high oil prices, and there is some rationale for that. But from another angle, the pessimism that has led to low valuations largely stems from the huge disconnect between this year's reality and market expectations at the start of the year. Investors could hardly have anticipated that in the first year of the "15th Five-Year Plan," the broad fiscal deficit would unexpectedly turn to a sharp negative growth, that the market would shift from expectations of Federal Reserve rate cuts to expectations of rate hikes, that the 10-year U.S. Treasury yield would hit a new high since 2007, that China-EU trade friction would suddenly intensify, and that a U.S.-Iran war would break out and last nearly a year. This gap is the reason many non-AI sectors saw sharp valuation contraction this year despite strong earnings growth. At present, the market has at least fully recognized these negative macro factors. Whether in overseas interest rates or domestic policy, any marginal change that exceeds expectations may help alter the current A-share valuation framework, which is more pessimistic than that of global markets.
K-shaped divergence is gradually converging, and market breadth is expected to gradually expand
On the AI side, anxiety about the growth of commercialization space will continue to constrain the absolute annual scale of AI investment, while deepening the market's vigilance toward a future "first-derivative" inflection point. It will be difficult for AI to again siphon funds away from non-AI sectors. At the same time, considering that mainstream institutional funds are still overweight TMT sectors, the process of position rebalancing during a market rebound will also constrain the elasticity of AI sectors. We have seen that during this week's process of the market stabilizing after declines, the first AI new technology to rebound was the robotics sector, where positions have been cleared quite thoroughly and institutional holdings are low, rather than sectors related to computing infrastructure. For non-AI sectors, almost all current negative factors have already been recognized by the market, even if they may not be fully priced in. After a round of systemic adjustment, low expectations themselves are a favorable condition. Take the overseas expansion sector as an example. China-EU trade friction is already on the table, and future annual rule revision negotiations and games may become normalized, but companies will ultimately adapt with strong competitive advantages. As long as the pace of RMB appreciation slows next year, exchange losses will decline year on year, bringing a recovery in net profit margins. If Democrats control Congress after the U.S. midterm elections, the Middle East war situation may change positively, and the resulting oil price impact may weaken marginally. A slowdown in AI investment growth would bring a decline in long-term overseas risk-free rates, weakening the suppression of demand in non-AI industries. Because these negative factors have already been systematically recognized by investors this year, any subsequent marginal improvement will also boost non-AI sectors themselves.
The clearing of capital flows needs time, and a steady-return strategy represented by PB-ROE may be more advantageous
The process of rapid deterioration in capital flows is likely nearing its end. With the signal of significantly increased ETF volume this week, active buying has begun to cautiously flow into the market, and the negative loop of margin deleveraging has been interrupted. As of October 21, the balance of margin financing in the two markets had fallen to 2.45 trillion yuan, a cumulative decrease of 560 billion yuan from the high point of the year, or 18.6%, returning to the level of late November last year. The cooling at the trading level is more obvious. The proportion of margin purchases in the turnover of the two markets, based on a 5-day moving average, averaged 9.86% in June, fell to 8.99% in July, was 8.66% in September, and further dropped to 8.01% since October. The reading on October 16 was 7.84%, the lowest since September 2024 and below the roughly 9% level before the "9.24" rally. The positions of active private funds have also returned to low levels. Surveys of CITIC SEC channels show that sample active private fund positions have continued to decline since the week of August 7, falling to 67.5% on October 16, the lowest since March 2024 and 8.8 percentage points below the historical median of 76.3%. Although capital flow indicators have fallen to relatively low levels, complete clearing and the start of sustained improvement still require time. At this stage, momentum and prosperity strategies will remain weak, market rotation will accelerate, and low-valuation quality strategies represented by PB-ROE will be more advantageous. At present, two types of industries fit this framework next year. One is varieties with high dividend yields and steady profit growth, such as white goods, textiles and apparel, food seasoning, and thermal power. The other is varieties with potential supply-clearing logic, such as oil shipping, express logistics, leading real estate developers, some lithium battery materials, and midstream petrochemical segments.
Risk factors include intensified friction in China-U.S. technology, trade, and financial areas; domestic policy intensity, implementation effects, or economic recovery falling short of expectations; domestic and overseas macroeconomic liquidity tightening more than expected; further escalation of conflicts in Russia-Ukraine, the Middle East, and other regions; and China's real estate inventory digestion falling short of expectations.
Comments