CITIC Securities: The reversal of the rate-hike narrative, together with changes in liquidity, places emphasis on the repair of some non-AI sectors.

The K-shaped divergence has intensified due to narratives, and it is normal to see recent fluctuations and convergence as narratives shift. This week, two sets of narratives underwent marginal changes: one relates to the global monetary environment assumptions, as the market begins to more comprehensively assess the Federal Reserve's policy stance rather than presupposing a hawkish rate hike path, leading to a reversal in the tightening and strong dollar narrative, which has helped repair negative sentiment in non-AI sectors; the other involves AI industry trends, as a piece of news about META sparked significant controversy, reflecting the market's current lack of tolerance for any negative AI industry information and highlighting the need for more diverse commercial monetization forms downstream to support more aggressive upstream investment expectations. Additionally, we anticipate a substantial easing of outflows from broad-based ETFs in the A-share market, which represents the most important marginal change in market liquidity. The correction of the rate hike narrative, combined with marginal changes in market liquidity, could catalyze a recovery in some non-AI sectors with solid earnings.

Narratives Behind K-Shaped Divergence Shift: Market Volatility Is Expected

Over the past few weeks, we have repeatedly explained how the global K-shaped divergence has been reinforced by narratives. First, the strong dollar and rate hike expectations have a greater impact on non-AI sectors, exacerbating the K-shaped divergence driven by differences in business conditions. Second, weak domestic demand data in China has deepened investors' bias against non-AI sectors, leading to a narrative of "carbon-based" versus "silicon-based" opposition rarely discussed overseas. Non-AI sectors in A-shares have significantly underperformed their overseas counterparts, even for companies with global exposure, particularly in cyclical manufacturing sectors with some business momentum and innovative drug sectors with upward industrial trends. Third, in addition to the spillover effects from the North American AI chain, A-share tech stocks have also priced in a valuation premium for domestic supply chain self-sufficiency. Since June, they have notably outperformed overseas AI sectors (the Philadelphia Semiconductor Index, Korea KOSPI50, and MSCI Global Semiconductor Products and Equipment have posted returns of -1.6%, -1.9%, and -2.2%, respectively, while A-share semiconductor materials and equipment, CSI Semiconductor, and STAR Chip have returned +43.2%, +32.9%, and +20.7%). This has made the divergence in A-shares more pronounced than in overseas markets. Setting aside narrative and emotional opposition, the market should have a more diversified pricing system and inclusive aesthetic perspective, as also discussed in the Shanghai Securities News article on July 2, "Beyond the 'Tech Herding' Narrative: Capital Market Inclusivity Requires Diverse Pricing." When a market is temporarily dominated by narratives and emotions, it is normal to experience some volatility due to narrative disruptions. The long-term bull logic of the market and the main theme of technology will not change as a result.

Marginal Changes in Two Sets of Narratives: Monetary Assumptions and Industrial Trends

1) Monetary Assumptions: The market begins to more comprehensively assess the Federal Reserve's policy stance rather than presupposing a hawkish rate hike path. Waller recently stated, "Hawkish stance unchanged, marginal rhetoric turning dovish," indicating that U.S. inflation risks have eased over the past four weeks and that AI-driven supply expansion could profoundly change the way the economy operates. However, whether it will bring inflation or deflation should be judged by central banks based on data. The market narrative that AI infrastructure squeezes energy and resources, leading to inflation in the carbon-based world and triggering rate hikes, currently lacks sufficient data validation. Waller previously mentioned that compared to core PCE, he focuses more on the trimmed mean PCE, which excludes abnormal price disturbances. The year-over-year trimmed mean PCE for May was 2.4%, significantly lower than core PCE, indicating that trend inflation pressures remain relatively mild. The subsequent release of June non-farm payroll data further confirmed labor market cooling, prompting the market to reprice the Fed's policy path, with U.S. Treasury yields and the dollar index falling in tandem. Since February, the market began repricing hawkish expectations, leading to a rebound in the dollar index and intensified K-shaped divergence across markets. AI hardware demand, given its high business momentum and strong order certainty, has relatively limited sensitivity to monetary tightening. In contrast, non-AI sectors are more suppressed by high interest rates, a strong dollar, and risk appetite. Recent marginal changes in the dollar and gold trends show that extreme concerns about liquidity tightening in the market have eased somewhat. As the strong dollar narrative cools and overseas liquidity expectations improve marginally, the valuations of previously suppressed non-AI sectors are expected to gradually recover, and the market style may shift from extreme AI trading toward a more balanced structure.

2) Industrial Trends: META's actions have sparked significant controversy, reflecting the market's lack of tolerance for negative AI industry information. We believe META's shift toward IaaS itself has little substantive impact on AI computing investment; in fact, opening up IaaS business may even be aimed at maintaining CAPEX more sustainably. However, the capital market has given META's stock a positive short-term pricing, and the greater concern lies in CSP vendors learning from each other. In this regard, the risk is actually controllable, as Google, Amazon, and Microsoft already have IaaS businesses and do not need to imitate META. Moreover, current stock prices are not the main short-term contradiction for these companies. The more critical issue is that the current market is so overheated that it is extremely sensitive to any negative AI information, which in itself implies that the market divergence has been too severe and needs some moderation. Recently, controversy over token costs has also heated up among North American AI application vendors. Palantir's CEO raised questions about the business models of leading AI labs, sparking heated discussions. This means that even within the AI sector, imbalances in profit distribution along the value chain are prompting industry reflection. Hardware companies represented by storage and optical communications have significantly outperformed the Mag7 this year, and as of six months of cumulative excess returns, they have reached 199%. Balancing and rationalizing profit distribution may be crucial for the next phase of market re-acceleration.

Important Marginal Change in Fund Flows: Redemption from Broad-Based ETFs Likely to Slow Significantly in H2

Over the past three weeks, the four major CSI 300 ETFs saw net redemptions of RMB 42.2 billion, RMB 90.4 billion, and RMB 58.5 billion, respectively, while no significant net redemptions were observed in other major broad-based ETFs. We expect selling pressure on broad-based ETFs to ease significantly in H2. On one hand, based on actual holdings at end-2025, as of July 3, 2026, large funds' holdings of CSI 300 ETFs amount to approximately RMB 20–30 billion in remaining scale. If the current redemption pace continues, this could end within a week, with limited potential selling remaining. On the other hand, according to market consensus forecasts, the 2026E net profit growth for all A-shares and all A-shares ex-financial and oil & petrochemical are approximately 9% and 13.5%, respectively. An index rise of around 10% for the full year could be considered consistent with a steady bull rhythm (the Shanghai Composite Index rose only 3.2% in H1), reducing the necessity for cooling measures in H2. In H1, broad-based ETFs saw cumulative net redemptions of RMB 1.8577 trillion, while industry/theme ETFs saw cumulative net subscriptions of RMB 253.1 billion, including RMB 88.9 billion in tech ETFs. For non-AI sectors lacking active fund inflows, this effectively amounts to a rare net reduction market. Conversely, if ETF net outflows ease in H2, the funding pressure on non-AI sectors in A-shares will also be significantly alleviated.

Reversal of Rate Hike Narrative and Marginal Changes in Market Liquidity: Focus on Recovery of Some Non-AI Sectors

In H1, the strong dollar and Fed rate hike narrative had a stronger suppressive effect on non-AI sectors, thereby intensifying K-shaped divergence. If this narrative reverses, it could in turn prompt convergence of the K-shaped divergence. Looking ahead to H2, net outflows from broad-based ETFs are expected to slow, and the funding suppression on non-AI sectors in the earlier period could ease significantly, benefiting the recovery of some non-AI sectors with business momentum. In terms of allocation, we continue to recommend a structure of AI + Energy & Chemicals. On the AI side, our preference leans toward strong supply constraints and low valuations, with a focus on downstream parts of the value chain such as cloud vendors, storage, gas turbines, and diesel generator sets. On the Energy & Chemicals side, in New Energy & Electrical, we favor sectors like electrolyte and additives, separators, etc., for earnings delivery. In Chemicals, the decline in oil price center and volatility brings demand for restocking and production starts, and the peaking of macro liquidity expectations could be a subsequent inflection point. Currently, we prefer varieties with larger cost reduction potential, relatively rigid demand, and low valuations, such as refrigerants, phosphorus chemicals, spandex, dyes, and large refining. In Metals, we recommend computing metals with some AI exposure but temporarily suppressed valuations due to the rate hike narrative, such as tin and copper. Additionally, we continue to recommend increasing allocation to low-valuation securities brokers, as liquidity suppression and other drawbacks may gradually fade in H2, with H1 earnings previews also serving as catalysts. Meanwhile, we remain positive on the recovery of innovative drugs, which have a sustained upward industrial trend and thoroughly cleared holdings.

Source: CITIC Securities Research

Risk Disclaimer and Terms

        Market risk exists; investment requires caution. This article does not constitute personal investment advice and does not account for the specific investment objectives, financial situations, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular circumstances. Investment decisions based on this article are at your own risk.