CITIC Securities: Trends within the technology sector are diverging, and non-AI sectors present broad opportunities for structural recovery

Main text

English text is being prepared.

The full text is as follows

A-share market | Adjustment front-loaded, turning point to emerge: 2027 investment strategy

English text is being prepared.

▍Market characteristics: K-shaped convergence, expansion of market breadth.

1) This year, major global stock markets have all experienced historically rare structural divergence. Since 2026, market breadth in global stock markets has generally narrowed, with clear divergence between AI and non-AI sectors. In the four markets of China, the US, Japan, and South Korea, the excess return of core AI portfolios relative to core non-AI portfolios reached about 30% during the year (as of October 22), and at one point reached as high as 150%. The market's trajectory deviated markedly from early-year expectations in two respects. First, it was difficult to foresee that, as AI investment expanded sharply, AI hardware would in the short term generate a large volume of negative externalities for other economic sectors, causing prices of production materials for various consumer goods to soar, siphoning funds from the bond market and causing overseas government bond yields to surge, ultimately making the extreme degree of market divergence far exceed expectations. Second, it was difficult to foresee that sectors such as overseas expansion and resources, which constitute market breadth and scope, would encounter various valuation-suppressing factors this year: market expectations shifted from Fed rate cuts at the start of the year to rate hikes, China-Europe trade frictions heated up rapidly, and the pace of RMB appreciation greatly exceeded expectations and caused huge foreign exchange losses. Interwoven with these factors, the market's optimistic expectations at the start of the year for high-prosperity sectors such as overseas expansion and resources turned into the current pessimistic expectations, and dynamic valuations were substantially compressed. Under such a huge expectation gap, the market rarely experienced broad declines in a year of rapid profit growth. Investors currently hold broadly cautious attitudes toward both AI and non-AI, very different from the start of the year.

2) Looking ahead to 2027, the AI industry trend is still upward, but the momentum in stocks may fade. The probability that AI hardware becomes the market's core theme for a fifth consecutive year is declining. Based on the current market's neutral expectations for the AI capex path, we estimate that by 2030, end-market AI applications will need to generate US$2.75 trillion in external direct-charge revenue to achieve a reasonable and sustainable level of capital returns across the entire industry chain; relying on frontier model makers alone would be far from sufficient. If no new scenarios capable of creating enormous commercialization space emerge (such as physical AI), then the existing model applications extended from Agents will either gradually become embedded as part of corporate IT budgets and gradually integrate into corporate business processes, or become a cost for consumer platforms to increase user retention time and secure traffic entry points (for example, the computing power consumed by AI short dramas). These are more like cost items than revenue-generating items. This means that AI investment may still continue, but the investment intensity in any single year may decline, gradually evolving into a purer process of infrastructure construction. From the perspective of the elasticity of the stocks themselves, assuming global AI capex reaches US$1.6 trillion in 2030, we estimate that, based on a 2030 PE of 15x, the market capitalization of A-share + Hong Kong AI hardware industry chain companies would be about 51% higher than current levels; overall compound returns are no longer attractive, and Alpha selection is needed. In addition, institutional holdings in AI remain relatively high, which will constrain the sector's ability to continue attracting incremental funds.

3) Expectations for non-AI are extremely low, but potential marginal improvement factors are increasing. Several main factors suppressing overseas non-AI demand may not get worse in 2027. First, the high interest rate situation overseas may ease. We expect that as North American AI investment slows next year and the year after, the scale of private-sector debt financing will peak and decline, weakening the crowding-out effect on government public-sector financing, improving supply and demand in the government bond market, and driving long-end rates lower. Second, the pace of RMB appreciation may slow, and exchange losses may decrease year on year. We estimate that when the USD/CNY exchange rate is in the 6.35-6.65 range, the impact of exchange losses on the net profit growth of all-A non-financial companies is 0.5%-6.2%, while in the first half of this year it reached 7.3%. Third, if the Democratic Party smoothly wins the U.S. midterm elections, the Middle East situation and oil prices may also improve, easing inflation and rate-hike anxiety and also easing currency depreciation pressure in non-U.S. countries. As for non-market factors, especially the escalating China-Europe trade frictions, we believe that in the long run they cannot stop the process of outstanding Chinese manufacturing companies going global; and factors such as export controls on key technologies and equipment and cuts in export tax rebates, although putting short-term pressure on the industry, are crucial for clearing out low-quality companies and stabilizing the supply-demand landscape and niche markets, and in the long run are very favorable for companies that operate compliantly and have already made full preparations for overseas expansion. For domestic demand, policy is an economic reaction variable that is difficult to predict; when the overall economy weakens and confidence among micro entities is sufficiently low, we can often still see policy intensity increase. This year, apart from the special circumstances of 2021 and 2023, is a year in which the fiscal deficit contracted year on year, which is very rare in history; tax revenue grew faster than expected while expenditure progress consistently fell short of expectations, and perhaps next year some adjustments will be made to certain micro-level implementation processes. Considering that more than half of this year's export growth was contributed by price increases in AI-related goods (volumes were basically flat), nominal export growth will face greater pressure to slow next year, which may also provide some reason for domestic demand policy to step up. Of course, the most favorable aspect of the non-AI sector is that expectations are sufficiently low, especially for the overseas expansion sector. For example, the valuation percentiles of home appliances, consumer electronics, textiles and apparel, commercial vehicles, and medical devices are basically close to historical extremes, and marginal better-than-expected improvement in macro factors may bring sufficient return elasticity.

▍Profit trends and valuation: growth may slow in 2027, but the main process of valuation contraction has already been completed this year.

1) We expect the profit growth of all-A non-financial companies to slow somewhat in 2027. Entering 2027, the profit base of the technology and cyclical sectors will rise markedly; even if prosperity remains high, their contribution to overall profit growth will weaken. At the same time, easing exchange rate pressure for overseas expansion companies, improvement in overseas demand, and stronger central fiscal support for domestic demand plus the disappearance of the impact of fading consumption subsidies are expected to make growth sources more diversified. We expect all-A net profit attributable to parent companies to grow 15.5% and 9.5% year on year in 2026 and 2027, respectively, and all-A non-financial net profit attributable to parent companies to grow 20.1% and 14.7%. Full-year profits will still grow, but the incremental structure will gradually shift from concentrated contribution by a few high-prosperity industries to joint support from continued technology growth, repair in overseas expansion, and reduced losses in domestic demand. Specifically: 1) Cyclical sectors: nonferrous metals digest a high base, while energy and chemicals provide profit support; net profit attributable to parent companies is expected to grow 45.5% and 9.8% year on year in 2026 and 2027, respectively; 2) Technology sector: the industry chain continues high growth, but growth slows under the influence of a high base; profit is expected to grow 101.9% and 43.0% year on year this year and next, respectively; 3) Overseas expansion sector: easing exchange rate pressure and improved overseas demand drive a full-year profit recovery; profit is expected to grow -0.5% and 8.8% year on year this year and next, respectively; 4) Domestic demand sector: stronger central fiscal support is the core variable for profits to stop falling and rebound; profit is expected to grow -14.4% and 8.8% year on year this year and next, respectively; 5) Financial sector: non-bank profit growth returns to normal, while banks remain stable; profit is expected to grow 11.5% and 4.6% year on year this year and next, respectively.

2) However, the valuation contraction process for most sectors has essentially been completed this year. As of October 23, 2026, the technology sector has completed the first wave of valuation de-rating after the "second derivative" downward revision. The 15 most core AI stocks have adjusted for 77 trading days since July, falling as much as 44.5% at the peak; the 200 stocks heavily held by institutions fell as much as 35.8%, which already exceeds the decline of core stocks during the collapse of the new energy crowded trade from December 2021 to the end of April 2022, when (December 17, 2021 to April 26, 2022) the 15 core new energy stocks fell 40.9% and the 200 institutionally heavily held stocks fell 33.7%. From a global comparison perspective, the valuations of non-AI leaders in the US and Japan are at historical highs, while A-share non-AI leaders are below the median. We used baskets of AI and non-AI leaders across the four markets of China, the US, Japan, and South Korea, calculated PE (TTM) on an aggregate basis, and conducted cross-market comparisons. As of October 22, 2026, the A-share market is the only one where the AI valuation premium relative to non-AI remains above its historical center: the AI/non-AI PE multiple is 4.4x, at the 82nd percentile since 2016; the same multiples for the US, Japan, and South Korea have fallen to the 1st, 17th, and 24th percentiles, respectively. The PE of non-AI non-financial leaders in the US and Japan is 33.5x and 23.0x, respectively, at the 98th and 95th percentiles; the PE of A-share non-AI non-financial leaders is only 15.6x, at the 21st historical percentile, with consumer, manufacturing, and pharmaceutical leaders at only the 12th, 1st, and 3rd historical percentiles, respectively.

3) From the perspective of shareholder returns, the A-share market has made substantial progress compared with the past. Although various types of equity financing in the technology sector reached RMB 246.5 billion this year (already 2.2 times that of 2025), cash returns and share buybacks for the market as a whole, especially for non-technology sectors, continued to hit record highs. This year, as of October 23 (the same below), against the backdrop of a concentration of technology company listings, A-share dividends and buybacks combined were 3.4 times equity financing (5.1 times in 2025 after excluding the special treasury bond capital injection into major state-owned banks, and 10.6 times in 2024 due to the temporary tightening of IPOs and refinancing). By comparison, the US stock market is rapidly transforming from its past status as a net shareholder-return market into a financing market. Federal Reserve Z.1 data show that net equity issuance by the US non-financial corporate sector (seasonally adjusted annualized) was negative in all four quarters of 2025, turned to +$181.4 billion in Q1 2026, and rose further to +$607.8 billion in Q2, the highest single-quarter level since 1990. If (cash dividends + buybacks - equity financing) ÷ average total market capitalization is used to measure the net cash return to shareholders, it can be seen that the net cash return rate of the S&P 500 has fallen from 4.6% in 2016 to 2.5% in recent years, and for US AI leaders it is only 0.81%. The direction in the A-share market is exactly the opposite: equity financing has decreased while dividends and buybacks have increased, and the net cash return rate for all A-shares rose from 1.5% in 2016 to 2.5% in 2024; it was 2.1% in 2025 and 1.4% so far this year.

▍Liquidity and style: Clearing in liquidity conditions takes time. Stable-return strategies represented by PB-ROE are more advantageous than prosperity strategies, and there may be a major shift in style.

1) The liquidity ecosystem needs time to recover, and investors are returning to reasonable expected returns. In terms of market liquidity, we expect the retail incremental process to slow markedly. After a new round of severe drawdowns, actively managed equity funds have once again fallen into the cycle of "larger redemptions as NAV rebounds." Sector ETFs, private quantitative funds, and long-term capital will constitute the main incremental sources of market liquidity in the future, and the influence of "long money" in the market is continuously rising. However, at the micro level, under the multiple influences of quarterly and annual assessments, unstable liability sides, solvency constraints, and drawdown stop-loss mechanisms, some formal "long money" such as annuities has also exhibited some short-term behavior. The key to repairing the liquidity ecosystem is not merely an increase in the absolute scale of a certain type of institution, but the re-matching of liability-side stability, assessment horizons, and trading behavior. Before this process is completed, the market may still focus more on short-term capital behavior and be reluctant to easily price for the long term. After experiencing this year's extremely high abnormal return elasticity and severe market adjustment, investors are expected to gradually return to reasonable return expectations, using moderate position levels and stable-return holdings to get through the stage of liquidity clearing and restoration of investor confidence.

2) Stable-return strategies will clearly prevail, and supply logic will surpass demand logic as the main factor. Judging from the characteristics of the A-share market, when demand clues are unclear, momentum and prosperity strategies enter a weak phase, and rotation accelerates markedly, stable-return strategies represented by PB-ROE will clearly prevail. Of course, static high-ROE or high-dividend-yield strategies have not been excellent in historical performance, and the stable-return industries differ considerably across stages, requiring subjective judgment to identify varieties whose profitability or dividend capacity has room for improvement in the future. Backtest results show that a PB-ROE strategy incorporating distress-reversal elements clearly outperforms a simple PB-ROE strategy and is generally relatively stable across different market environments, especially when momentum strategies fail. We believe that two types of industries currently meet this framework for 2027. One is varieties with high dividend yields and steady profit growth next year, such as white goods, textiles and apparel, food seasoning, and thermal power; the other is varieties with supply-clearing potential logic, such as oil shipping, express delivery, leading property developers, some lithium battery materials, and midstream segments of petrochemicals.

▍Industry allocation: Diversify appropriately and respond around four threads.

English text is being prepared.

▍The market adjustment has already occurred earlier this year, and structural turning points are increasing next year.

After a broad decline there is no bear market, or at least it is a structural market. As of October 23, 2026, 4,284 stocks among all A-shares recorded negative returns, accounting for 76.8%, a relatively rare broad-decline year; historically, after the share reform, A-shares have rarely seen such a high proportion of declines in two consecutive years (2017 counts as one, but that year traditional-industry white-horse leaders delivered very considerable excess returns for institutions). Excessively high expectations at the start of the year and the frequent surprises this year (the US-Iran war, China-EU trade frictions, high interest rates, year-on-year contraction of fiscal deficits, etc.) led to market performance that, in hindsight, was not entirely satisfactory. But looking ahead to 2027, at a point when expectations are generally cautious, we believe the breadth of the market's performance will instead widen next year; the valuation contraction process for most sectors has basically been completed this year, AI investment may slow down, but its crowding-out effect on other non-AI industries will decline, K-shaped convergence will occur, and market performance will become more balanced. Improvement in the capital ecosystem takes time; steady-return strategies represented by PB-ROE are likely to clearly outperform prosperity strategies and should be actively pursued. In terms of industries, it is recommended to allocate around four threads: steady returns, outstanding manufacturing going global, domestic demand recovery, and new AI technology themes.

▍Risk factors:

Intensified frictions between China and the US in technology, trade, and finance; domestic policy intensity, implementation effects, or economic recovery falling short of expectations; overseas and domestic macro liquidity tightening more than expected; further escalation of conflicts in regions such as Russia-Ukraine and the Middle East; and China's real estate inventory digestion falling short of expectations.

(Source: Jiemian News) Middle of the article Embedded advertisement