In a recent research report, CITIC SEC (HK: 06030) highlighted that during the latter phase of an industrial supercycle, a rally in non-institutional stocks typically follows the peak of institutional favorites. Currently, the AI narrative, the earnings cycle position, and the global monetary environment are likely to constrain institutional stocks for several reasons: 1) While AI computing power investment remains robust, market expectations for the commercialization potential of frontier model developers are being recalibrated; 2) Although full-A non-financial earnings may continue to improve quarter-over-quarter in Q3 2026, the peak year-over-year growth rate could emerge in Q4 of this year; 3) The Federal Reserve's firm stance on controlling inflation is set to create a tighter macro liquidity atmosphere at least for the remainder of the year. Viewed through an institutional lens, these factors undoubtedly cap the upside for market rallies. However, from the perspective of short-term sentiment and the chip cycle, coupled with the catalyst of Q3 earnings reports, the market environment is conducive for active capital to target new technologies and themes. The firm advises investors to actively seize the final offensive window within the year.

On the allocation front, given the high-interest-rate environment, the market's K-shaped divergence is likely to widen again during this final year-end offensive window, with AI-related assets re-emerging as the preferred choice. Within the technology sector, two key directions are highlighted: first, new optical communication technologies, PCB, and advanced packaging that benefit from increased manufacturing complexity; and second, wafer fabrication and gas turbines with clear volume growth logic. Non-institutional heavyweight stocks may offer greater upward elasticity, while North America-linked chains could perform relatively better in the near term. Outside of technology, the focus remains on energy and chemical companies and leading brokerages with overseas expansion potential.

Where to Begin

Historical data suggests that the first price peak of an industrial supercycle often coincides with a peak in relative outperformance of institutional stocks. Following a pullback in their excess returns, the index may form a double-top structure, where non-institutional heavyweight small caps often drive the second peak. The second top is typically a result of limited rebounds among institutional stocks and new highs among non-institutional ones. Notable examples include 2009, Q4 2015, and Q2 2022, where non-institutional small caps led the formation of a second top. After the sharp market pullback in Q3 2015, divergence became pronounced: from September 15, 2015, to year-end, heavily held institutional portfolios rose 54.1%, significantly underperforming low-holding portfolios (73.3%) and low-holding small-cap portfolios (83.2%). Similarly, after the new energy rally peaked at the end of 2021, divergence between institutional and non-institutional small caps emerged during the pullback and failed to converge in the subsequent uptrend.

The commonality across these second rallies is a trade executed from a chip and sentiment bottom while the industrial narrative heat and earnings trend remain unchanged. In these second-phase rallies, a considerable number of stocks surpass their first-phase highs. For instance, during the Q3 2009 rebound, 58.6% of high-holding, 73.3% of low-holding, and 76.5% of low-holding small-cap stocks exceeded their previous highs. Moreover, non-institutional small caps consistently demonstrate stronger rebound magnitudes compared to institutional stocks across all rallies.

Constraints on Institutional Stocks

The current AI narrative, earnings cycle position, and global monetary environment are likely to curtail institutional stock performance. First, while AI computing power investment shows no signs of slowing, expectations for the commercialization space of frontier model developers are shifting. The supply-demand imbalance for computing power remains intact. The recent "frontier model slowdown" discussions in North America may be better interpreted as follows: under the existing business models of frontier model developers, further capability advancements might not yield a more imaginable commercial payment market, and they are currently the largest buyers of computing power. The distribution of US enterprise AI payments reveals an unhealthy market structure dominated by developer-led agents. According to enterprise spending data platform Ramp, disclosed in early September, OpenAI and Anthropic generate about 80% of enterprise revenue from their top 1% of clients. These clients spend 10.7 times more than the top decile, and 576 times more than the median enterprise, with this concentration not improving despite the growth in paying firms. More concerning, per-capita AI spending by the top 1% of enterprises fell 9.7% month-over-month in August, suggesting the payment ceiling for the heaviest users may have been reached. Whether due to hit ceilings on marginal payment capacity or price declines from competitive landscape shifts, the conclusion points to future market growth driven by broader enterprise spending rather than further increases at the pyramid's apex. Integrating AI into broader business processes requires substantial efforts from enterprise service providers. The ongoing AI adoption progress among North American SaaS companies supports this trend, with software stocks notably outperforming hardware recently—IGV has gained 35.7% relative to SOXX since June. However, "enterprise services + AI" integration is a relatively slow-moving variable compared to Anthropic's rapid ARR surge in May, offering limited near-term catalysts for market optimism. If total computing power investment lacks further upside imagination, with consensus expecting a noticeable growth slowdown by 2028, opportunities may be limited to product iteration rather than a broad rise in the AI sector's ceiling.

Second, full-A non-financial earnings are expected to continue improving quarter-over-quarter in Q3 2026, with the year-over-year growth peak potentially occurring in Q4 2026. For a market reliant on sustained upward earnings momentum to support trend-following capital, the trajectory of overall earnings growth is crucial for how institutional stocks are perceived. We project that full-A non-financial earnings will improve sequentially through Q3 2026, driven primarily by cyclical and technology sectors. The peak of this earnings growth uptrend may hit Q4 2026, after which growth could gradually decelerate in the following year amid high bases in technology and slower commodity price increases. This outlook contributes to institutional investors' hesitancy to adopt overly optimistic stances for the coming year. Nevertheless, variables capable of shifting this consensus remain. Since Q2, investor anxiety over non-AI overseas expansion and domestic demand has deepened—overseas expansion faces headwinds from complex trade relations and RMB appreciation-driven exchange losses, while domestic demand contends with subsidy tapering and broad fiscal contraction. Over a full year, these factors may evolve, and at least some negative expectations are already priced into equities. Yet within the limited Q4 time frame, the market is more likely to treat "continued sequential improvement with a potential peak in annual growth" as the base-case pricing scenario.

Third, the Federal Reserve's hawkish inflation stance will create a tighter macro liquidity atmosphere, at least for the year-end period. The Fed's September rate hike was largely anticipated; as of September 18, the implied probability of a further hike this year reached 90.1%, with two additional rapid hikes expected by Q1 next year. The market currently prices a cumulative 100bps of hikes for the current cycle (Q3 2026–Q2 2027). Therefore, even if the Fed raises rates again this year-end, it should not be viewed as a risk event. The lack of breadth in economic growth limits the Fed's room for trend-driven hikes next year. According to the US Bureau of Labor Statistics, average private-sector wages (seasonally adjusted) rose 3.1% year-over-year in August, the lowest since 2022, with inflation-adjusted real wage growth at -0.3%. At such wage growth levels, an inflation expectation spiral is unlikely. In fact, the 10-year TIPS implied inflation rate stands at 2.33%, notably below April and May levels and roughly matching pre-conflict readings. As long as the Fed maintains its inflation focus, using precautionary or symbolic hikes to manage expectations, the necessity for trend-driven hikes will diminish once energy costs from the Middle East conflict and consumer electronics price pressures ease. Consequently, under this hiking path, investors may treat hikes as short-term disruptions or tactical events, affecting year-end rebounds and re-ratings of institutional stocks, but without excessive pricing of higher discount rates on equity valuations—thus not becoming a sustained market driver.

Short-Term Sentiment and Catalyst for Active Capital

The market has undergone a thorough sentiment cooling, with widely discussed risks now largely priced in. Since September, concerns over high oil prices, Fed hikes, and negative AI narratives have been factored in. Some overseas-linked blue chips even experienced abrupt sell-offs earlier this week, resembling forced selling by institutions facing liability-side pressures. Trading sentiment indicators have yet to show significant recovery. Channel surveys from CITIC Securities (SH: 600030) show that as of September 11, active private fund positions stood at 73.0%, slightly up from the previous week yet below the 76.4% historical median, sitting at approximately the 29th percentile since end-2020. Our constructed investor sentiment indicator remains at low levels.

This sentiment positioning provides a foundation for short-term offensive moves. With the Mid-Autumn Festival and National Day holidays approaching, even if trading volumes modestly increase this Friday, investors are unlikely to aggressively add positions before the holidays, more probably deferring the offensive window until around the Q3 earnings season. In this constrained time frame, taking an offensive stance based on short-term sentiment cycles requires sector-specific catalysts and consensus on limited capital, primarily concentrated in new AI technologies and themes. Given the potential for intense competition among institutional heavyweight stocks, active capital may lean toward non-institutional holdings.

Seizing the Final Offensive Window

When an industrial trend has yet to cool, with only far-future narrative ceilings being touched and priced to some degree, it is highly unlikely that a sharp correction signals the end. A secondary attack or even new highs are probable. Historical experience suggests that the second offensive phase is primarily driven by non-institutional stocks aligned with the industrial trend. Considering the near-term industrial narrative heat and capital recognition, the offensive rally around October's Q3 earnings reporting is more likely concentrated in non-institutional heavyweight tech stocks, especially those involving new technologies and themes. From a positioning and sentiment perspective, September's market has already priced in numerous risk events—Fed hikes, AI slowdowns, and high oil price narratives. The foundation for sentiment to warm in October and launch an offensive move readily exists, potentially even front-running optimistic expectations for next year through an early valuation switch. This may well be the market's final offensive window for the year.