Shenwan Hongyuan Group: The AI chain rebound may continue until late September, focusing on exploring new segments and targets for computing power inflation.

Shenwan Hongyuan Group released a research report stating that after a sharp drop and rebound, disturbances in the AI industry have re-emerged, compounded by concerns over elevated U.S. Treasury yields, leading to a secondary bottom-seeking market. The short to medium-term judgment remains unchanged, with the rebound potentially continuing until late September, supported by steady and long-term policies. The bank believes the September rebound will persist, with structural selection focused on identifying directions that can significantly reverse the extent of declines during the sharp rebound. Internally within the AI sector, there is increased attention to non-institutional heavyweight directions, with domestic computing power chains and AI small-cap stocks providing mid-term cover for the short-term. Additionally, among global computing power chains, the downturn inflection point has appeared relatively late, and there are new catalytic directions in the short term that can be easily realized with high elasticity, such as storage and PCBs. These industrial chains also need to explore new sub-segments of computing power inflation and new targets. The main points from Shenwan Hongyuan Group are as follows: 1. The short-term market trend aligns with the bank's four-step deduction for the AI industrial chain regaining strength. After a sharp drop and rebound, disturbances in the AI industry have re-emerged, compounded by concerns over elevated U.S. Treasury yields, leading to a secondary bottom-seeking market. The short to medium-term judgment remains unchanged, with the rebound potentially extending to late September, supported by steady and long-term policies, where optimistic expectations in the long, medium, and short term may be concentrated and fermented, which could be the peak of this rebound phase. Funds are not quickly returning to the state experienced at the end of June, and achieving new highs in the AI chain requires fundamental expectations to surpass those from the end of June. Relative to experiences since 2025, the level of industrial catalysts needed for this AI chain rally to restart is higher, and the time required is likely longer. There is still potential for new adjustment waves if September does not see new highs. Consequently, the market may expect the tech sector adjustment wave to lengthen, transitioning from monthly to quarterly levels. The short-term market trend basically aligns with the bank's deductions, reiterating the AI industrial chains four-step approach to regaining strength: 1. A sharp drop and rebound occurred, restoring the effectiveness of fundamental research. 2. Floating profits and losses have returned to the breakeven line, naturally increasing resistance to the sharp drop rebound. This week's short-term disturbances related to large models, compounded by concerns over high U.S. Treasury yields, have triggered a secondary bottom-seeking market. However, after the adjustment, the force of low-configuration is showing, and the rebound extending to late September remains highly probable. The stable and long-term policies are supporting, and optimistic expectations across long, medium, and short terms may converge and ferment, likely marking the peak of this rebound. 3. The 2026 Q3 reports are likely to digest valuations, with high growth performance reinvigorating support for stock prices, while overseas computing power chains are expected to exhibit differentiated trends. For leading firms with high certainty in performance, a return to an upward trend may be possible. 4. The overall condition for the AI chain to achieve new highs is the realization of substantial industrial catalysts, which will solidify a new consensus trend in the AI industry. The peak of the AI chain's sharp rebound in September is unlikely to break the previous highs. The return of funds to the June end state is slow, and new highs in the AI chain require fundamental expectations to exceed those from the end of June. Each mid-term adjustment phase in the AI market since 2025, where floating profits shift from positive to negative, confirms new industrial trends within a quarter. This round's adjustment is at a higher level (possibly corresponding to high penetration rates of AI Coding), requiring a higher level of industrial catalysts and likely longer waiting times for the market to reinstate. It is possible that a new consensus on the AI industry trend may not be formed by late September, preventing the market from breaking through previous highs and triggering new adjustment waves. During this time, the market may expect the tech sector adjustment wave to lengthen, moving from a monthly to a quarterly scale. Relative to experiences since 2025, the level of industrial catalysts required to restart this AI chain rally is higher, and the time required is likely longer. If September does not see new highs, it will lead to additional adjustment waves. Consequently, the market may expect the tech sector adjustment waves to extend, transitioning from a monthly to a quarterly level. 2. The high U.S. Treasury yields are driven by both medium-term and short-term factors. The rise in U.S. Treasury yields is not a new change. In the short term, the stock market is focused on pricing that centers around tech adjustment waves, exacerbated by concerns over high interest rates and increasing the amplitude of adjustments, which tends to reflect small wave characteristics. The bank summarizes the medium-term issues: the Federal Reserve's loose monetary policy must take place against the backdrop of weakening support from AI capital expenditure on the economy. The investment clock has returned to a favorable quadrant, which requires a longer time and basic expectations to experience fluctuations. The market's reasonable inference suggests that before the arrival of a new round of tech trends, there may be a phase where risks concentrate and are released. Discussing the impact of high U.S. Treasury yields: First of all, the rise in U.S. Treasury yields is driven by both medium-term and short-term factors. The medium-term factors include the recurring conflicts between the U.S. and Iran, the spread of resource nationalism, which pushes up mid-term inflation, and ongoing concerns about passive interest rate hikes. The weakening of fiscal discipline in the U.S., with short-term borrowing to cover long-term debt, accumulates medium-term risks. Trump has continuously politicized key aspects of the dollar cycle. Short-term factors include increased issuance pressure of U.S. debt, diversion of bond financing needs in the AI industry, and reallocating reserve assets (such as gold). U.S. Treasury yields have risen for some time and are not a new change. Suddenly, the market's focus has shifted to the tech industry's disturbances, compounded by concerns over elevated U.S. Treasury yields, intensifying short-term adjustment amplitudes. The increase in attention itself reflects small wave characteristics. Medium-term worries focus on the idea that short-term factors will not persist forever, and certainly, there will still be a phase with gradual adjustment. The medium-term issues behind rising U.S. Treasury yields are summarized by the bank as follows: financing in the U.S. AI industrial chain is high, with a lack of direct and effective means for managing potential credit risks. There are numerous price tool constraints, while the existing policy direction is cautious in the use of quantitative tools. This week, the U.S. Treasury's disguised quantitative tightening has had some effect, but more effective measures should incorporate QE or even more groundbreaking targeted monetary easing tools. At the same time, the Federal Reserve will require many prerequisites to transition from interest rate hike expectations to actual rate cuts, including the weakening support from AI capital expenditure on the overall economy, corresponding to a tightening of computing power inflation, indicating the onset of computing power deflation. Therefore, before a favorable investment clock quadrant appears and before a beneficial environment for the diffusion of AI applications emerges, basic expectations for fundamentals and liquidity may still experience disturbances. Thus, prior to the arrival of a new round of tech trends, a phase of concentrated risk release may be a reasonable inference. In the short term, high U.S. Treasury yields may trigger a one-time shock in risk appetite. However, in the new adjustment wave by the end of September, the medium-term problems behind high U.S. Treasury yields might still serve as significant disturbance factors. 3. With the continuation of the September rebound, the structural focus is on identifying directions capable of significantly reversing declines during the sharp rebound. Internally within the AI chain, there is greater attention to non-institutional heavyweight directions, with domestic computing power chains and AI small-cap stocks providing mid-term cover for the short-term. Moreover, in the global computing power chain, the downturn inflection point has emerged relatively late, and there are new catalytic directions in the short term that can easily realize high elasticity, such as storage and PCBs. These industrial chains also need to explore new sub-segments of computing power inflation and new targets. After September, the market may anticipate that tech sector adjustment waves will lengthen. Consequently, time for non-tech directions to outperform has lengthened; adjustment phases have increased, while high-dividend assets have extended periods for absolute returns. In non-tech tracks, innovations in drugs and contract research organizations (CROs) signify directions where funds have generated a positive cycle. Consensus around precious metals is quickly forming and is expected to represent a positive fund cycle. Attention should also focus on opportunities in industrial metals and basic chemicals. In the medium term, as adjustment waves increase, high-dividend assets may repeatedly see opportunities. It is worth noting that the weight of the CSI 800 Index public fund holdings in Q2 2026 may provide an effective method for identifying high-dividend assets, with a focus on banks, non-bank financials, food and beverages, and public utilities. Risk Warning: Overseas economic recession exceeds expectations.