CITIC SEC: Actively seize the final offensive window for A-shares within the year.

CITIC SEC released a research report stating that in the later stages of an industrial supercycle, after institutional favorites peak, there is usually a round of new-high rallies in non-institutional stocks. The current AI narrative, the position of the earnings cycle, and the global monetary environment are likely to constrain institutional favorites: 1) AI computing power investment has not slowed, but the market's expectations for the commercialization space of frontier model companies are adjusting; 2) All-A non-financial earnings may continue to rise quarter-on-quarter in 2026Q3, but the peak in year-on-year growth may appear in Q4 this year; 3) The Federal Reserve has shown a tough stance on inflation control, which will create a relatively tight macro liquidity atmosphere at least within the year. If viewed from an institutional perspective and mindset, these factors will undoubtedly constrain the height of the market. However, from the perspective of short-term sentiment and the chip cycle, combined with the catalyst of Q3 earnings, the market has the soil for active capital to attack new technologies and new themes. We recommend actively seizing the final offensive window within the year. In terms of allocation, considering the high interest rate environment, during the final offensive window within the year, the market's K-shaped divergence may re-expand, with AI regaining dominance. In the technology sector, we recommend focusing on two directions: first, new optical communication technologies, PCB, advanced packaging, etc. that benefit from increased manufacturing complexity; second, wafer manufacturing, gas turbines, etc. with clear volume growth logic, among which non-institutionally heavily held stocks may have greater upward elasticity, and the North American chain may be relatively dominant for some time to come; in the non-technology sector, continue to focus on energy and chemicals and leading brokerages with overseas expansion potential. The main views of CITIC SEC are as follows: Later stages of an industrial supercycle After institutional favorites peak, there is usually a round of new-high rallies in non-institutional stocks Review shows that the first stock price peak of an industrial supercycle is often accompanied by the relative excess return peak of institutional favorites, and after the excess returns of institutional favorites decline, the index may form a double-top structure, at which point non-institutionally heavily held small caps often become the core DRIVE of the second top. The second top of the double-top structure is often the result of limited rebound in institutional favorites and new highs in non-institutional stocks. 2009, 2015Q4, and 2022Q2 are all typical cases where non-institutional small caps hit new highs and drove the sector as a whole to form a second top. After the rapid pullback in Q3 2015, the divergence was extremely obvious. From September 15, 2015 to the end of 2015, the cumulative gain of the high-institutional-holding portfolio was 54.1%, significantly weaker than the low-holding portfolio (73.3%) and the low-holding small-cap portfolio (83.2%); after the new energy rally peaked at the end of 2021, the divergence between institutional favorites and non-institutional small caps had already begun during the pullback and did not converge in the second wave of gains. The common feature of the second wave is a round of trading at the bottom of the chip and sentiment cycle against the backdrop of unchanged industrial narrative heat and earnings trends. Our review found that in the second wave, a considerable number of stocks are expected to break through the highs of the first wave. For example, in the rebound of 2009Q3, the proportion of stocks breaking previous highs in the high-holding, low-holding, and low-holding small-cap portfolios was 58.6%, 73.3%, and 76.5%, respectively. In addition, in each round of rebound, the rebound magnitude of non-institutional small caps was significantly stronger than that of institutional favorites. The current AI narrative, the position of the earnings cycle, and the global monetary environment are likely to constrain institutional favorites 1) AI computing power investment has not slowed, but the market's expectations for the commercialization space of frontier model companies are adjusting. From the supply-demand perspective, the fact that computing power is scarce still holds. The recent hot discussion in North America about "frontier model slowdown" is, in our view, more reasonably interpreted as: under the existing business models of frontier model companies, continuing to push forward frontier model capabilities may not a more imaginative commercial payment market, and they are precisely the largest buyers of computing power at present. In fact, the current distribution of AI payments by U.S. enterprises shows that the paid market structure dominated by frontier model companies' Agents is not healthy. According to data disclosed by Ramp, an enterprise spending data platform, in early September, about 80% of OpenAI's and Anthropic's enterprise revenue comes from their top 1% of customers. The spending intensity of these customers is 10.7 times that of enterprises in the top 10th percentile and 576 times that of median enterprises, and this concentration has not improved with the expansion of the number of paying enterprises. A bigger hidden danger is that in August, per capita AI spending by the top 1% of enterprises fell 9.7% month-on-month from July. In other words, the ceiling of willingness to pay among the heaviest users seems to have been reached. Whether it is marginal payment capacity hitting the ceiling or price declines caused by changes in the competitive landscape, the conclusion is the same: future market space lies in broader enterprise payment growth, not in further raising the payment level of the very top users. Enterprises integrating AI more broadly into business processes requires substantial efforts from enterprise service providers. Recent AI adoption progress among North American SaaS companies is confirming this trend, and North American software stocks have recently clearly outperformed hardware stocks. Since June, IGV's excess return relative to SOXX has reached 35.7%. But the adoption of "enterprise services + AI" is a relatively slow variable. Compared with the rapid surge in Anthropic's ARR from February to May, its short-term catalyst for market optimism is limited. If there is no further imaginative space for the total scale of computing power investment, and the market consensus is that growth will decline significantly in 2028, then what has opportunity is product iteration, not a rise in the ceiling of the entire AI sector. 2) All-A non-financial earnings are expected to continue rising quarter-on-quarter in 2026Q3, with the peak in year-on-year growth likely appearing in 2026Q4. For a long-only market, maintaining a trend of continuously rising earnings growth is crucial for trend capital driven by the prosperity approach. The "holding experience" of institutional favorites in A-shares is usually related to the direction of overall earnings growth. We expect the earnings of the All-A non-financial sector to continue improving quarter-on-quarter in 2026Q3, with cyclical and technology sectors still the main DRIVE. The peak of this earnings growth upcycle may appear in 2026Q4. Next year, against the backdrop of a high base in the technology sector and slower commodity price increases, All-A non-financial earnings growth may begin to decline slowly. This may be the main reason why many institutional investors do not dare to be too optimistic about next year at the current position. Of course, variables that could change this market consensus still exist. Since Q2 this year, investors' anxiety about non-AI overseas expansion and domestic demand has continued to deepen. Overseas expansion is mainly constrained by complex foreign trade relations and exchange losses caused by RMB appreciation, while domestic demand is constrained by the phase-out of subsidies and the year-on-year contraction of broad fiscal spending. But over the next full year, these factors may also change, and at least the potential negative expectations have more or less been reflected in stock pricing. However, within the limited time window of Q4 this year, the market is still more likely to treat "continued quarter-on-quarter improvement, with year-on-year growth possibly peaking" as the baseline pricing scenario. 3) The Federal Reserve has shown a tough stance on inflation control, which will create a relatively tight macro liquidity atmosphere at least within the year. The September Fed rate hike was basically within expectations. As of September 18, the implied probability of another Fed rate hike this year has reached 90.1%, while an additional two rapid hikes have already been priced in for Q1 next year. The market currently expects a cumulative 100bps of hikes in this cycle (2026Q3-2027Q2). Therefore, even if the Fed hikes again within the year, it should not be regarded as a risk event. Insufficient breadth of economic growth limits the Fed's room for trend rate hikes next year. According to U.S. Bureau of Labor Statistics data, the year-on-year increase in average hourly earnings in the U.S. nonfarm private sector in August (seasonally adjusted) was 3.1%, the lowest level since 2022, while after adjusting for inflation, real average wage growth was -0.3%. At such a wage growth level, runaway inflation expectations are a low-probability event. In fact, the current implied inflation rate of 10Y TIPS U.S. Treasuries is 2.33%, significantly lower than the levels in April and May this year and roughly the same as before the U.S.-Iran conflict. As long as the Fed continues to strengthen its emphasis on inflation and uses timely preventive or signaling rate hikes to control inflation expectations, after the pressure from energy and chemical costs brought by the Middle East conflict and consumer electronics price increases eases, the necessity for trend rate hikes will decline substantially. Therefore, we believe that under this rate hike path, investors will regard rate hikes as short-term disturbances or game-like events. This may affect the rebound and revaluation space of institutional favorites within the year, but the market will not overprice the impact of rising denominator-side rates on equity valuations (that is, it will basically not affect the discounting of long-term cash flows), and it will not become a factor driving sustained market adjustment. The position of the short-term sentiment cycle and the catalyst of Q3 earnings provide the soil for active capital to attack new technologies and new themes The current market has already experienced a relatively full cooling of sentiment, and the risk events currently discussed and recognized by the market have all been priced to some extent recently. Since September, the market began pricing concerns about high oil prices and Fed rate hikes. Negative narratives about an AI industry slowdown have also appeared. Some blue-chip stocks related to overseas expansion briefly experienced "flash crashes" in the first half of this week, which looks very much like some institutions being forced to sell due to liability-side pressure. Market trading sentiment indicators still have not rebounded significantly. According to CITIC SEC channel research, as of September 11, the sample active private fund position was 73.0% (a slight increase from the previous week), below the historical median of 76.4%, at about the 29th percentile since the end of 2020. Our constructed investor sentiment indicator also continues to be at a low level. The position of the sentiment cycle provides the basis for a short-term offensive. Considering the upcoming Mid-Autumn and National Day holidays, even if market volume rises slightly this Friday, we believe investors will not aggressively add positions before the holiday and are more likely to postpone the offensive window to around Q3 earnings. In a limited time window, adopting an offensive approach with the mindset of a short-term sentiment cycle requires prosperity catalysts and limited capital consensus. The areas with both characteristics are mainly concentrated in new technologies and new themes around AI. Considering the possible intense gaming in institutionally heavily held stocks, active capital may be more inclined to attack non-institutionally heavily held stocks. Actively seize the final offensive window within the year In an environment where the industrial prosperity trend has not yet cooled, but the long-term narrative ceiling has been touched and priced to some extent, the probability of the market ending directly after a round of severe adjustment is very low. A second offensive or even new highs are likely to appear. However, from historical experience, the main direction of the second offensive is non-institutional stocks in the industrial trend. Considering short-term industrial narrative heat and capital recognition, the offensive around Q3 earnings in October is more likely to be concentrated in non-institutionally heavily held stocks within technology, especially some new technologies and new themes. From the perspective of positions and sentiment, the market concentrated on pricing many risk events in September (Fed rate hikes, AI slowdown, high oil price narrative). The basis for sentiment turning hot in October and launching an offensive fully exists, and it may even price in next year's optimistic expectations in advance (advance valuation switching). This may also be the final offensive window within the year.