Following a period of adjustment, we have conducted a comprehensive review of the current positions, challenges, and conditions required for the next stage of advancement across four key sectors: the North American AI supply chain, the domestic Chinese AI supply chain, non-AI export-oriented sectors, and domestic demand-focused plays. For the North American AI chain, the evolution of industrial logic carries greater weight; a valuation shift from cyclical to growth stocks requires the emergence of a broader downstream commercial closed-loop. For the domestic AI chain, sentiment and liquidity factors are more influential; a market reacceleration awaits the conclusion of liquidity disruptions, a decline in volatility, and opportunities at the emotional trough. For non-AI sectors, the core focus remains overseas demand, with the trade environment being paramount. The market underestimates the potential demand shocks from China-Europe trade negotiations and significant currency depreciations in emerging markets, with the Federal Reserve's monetary policy environment being secondary. For domestic demand plays, weak domestic demand is a consensus among investors. The foundation for a new round of incremental policies akin to the "September 24th" measures is in place. As defensive plays during the adjustment, expectation-driven trading for the domestic demand sector has already commenced. Despite various imperfections and issues, we believe many concerns have already been recognized and priced in by the market, as reflected in the sustained weakness of non-tech sectors since Q2 and the recent pronounced "catch-up decline" in tech sectors. We believe a new market upswing is in the making, awaiting shifts in various narratives to unlock medium-term upside potential. Currently, we are closely monitoring the end-of-month guidance from North American cloud providers, market sentiment and volume-price indicators, the Federal Reserve's policy orientation after the weakening of the AI inflation narrative, and expectations for the late-month Politburo meeting. At this stage, we anticipate three convergences in the market: first, a convergence in the excess returns of upstream AI hardware and price-increase plays relative to downstream cloud service giants; second, a repair of the discount for non-AI industrial stocks relative to their overseas peers; and third, a convergence in valuations between tech and non-tech sectors.

North American AI Chain: Shifting from Cyclical to Growth Stock Valuations Requires a Broader Downstream Commercial Closed-Loop

The primary issue for the North American chain is not its prosperity or the AI industry narrative, but rather that the commercial cycle remains confined between hardware manufacturers, frontier model companies, and cloud providers. The dominant narrative is cost reduction and efficiency improvement for enterprises rather than revenue growth. As companies shift from "Tokenmaxxing" to "Token Optimizing," market concerns about the sustainability of AI capital expenditures will persist, making it difficult for investors to view AI hardware companies outside a cyclical stock perspective (where stock prices peak ahead of the cycle) and instead adopt a growth stock valuation framework (featuring a higher central valuation). Based on current market consensus expectations for the AI capital expenditure trajectory, external explicit commercialized revenue may need to reach a scale of approximately $3.2 trillion to cover the 15% Weighted Average Cost of Capital (WACC) for infrastructure investment. This figure excludes cloud service revenue from Cloud Service Providers (CSPs). If all AI-related revenue across the full spectrum is included, the required scale is estimated to be around $4.95 trillion. Currently, the clearest explicit payment market is the subscription fees from frontier AI model companies like Anthropic, currently at a scale of hundreds of billions of dollars. To fill the remaining commercial revenue gap of approximately $3 trillion, it remains unclear which market entities and payment forms will bear the burden. If this commercialization primarily relies on enterprises using AI to replace or save human capital expenditures, its addressable space still requires cautious evaluation. According to the Stack Overflow 2026 Developer Survey and various national industry data, there are approximately 30 million software developers globally, with an average annual salary of about $60,000 to $70,000, corresponding to a total compensation pool of roughly $2 trillion. This scale represents only about 5% to 6% of the global white-collar compensation pool. This is not to say that work in other industries does not use AI or generate substitution, but rather that scenarios like finance, law, administration, and customer service consume far fewer tokens than software development, and usage can be significantly reduced after targeted optimization. New revenue ceilings require new commercial scenarios and narratives to further open up, similar to the product explosion seen with Coding Agent this past February. Regarding this, it is difficult for us to make forward-looking judgments; we can only patiently wait. At this stage, the North American chain may oscillate around previous high-volume trading zones rather than independently advancing to the next level.

Domestic AI Chain: Awaiting the End of Liquidity Disruptions, Declining Volatility, and an Emotional Trough

Since June, the domestic semiconductor index has generally outperformed indices from South Korea, Japan, and the North American Philadelphia Semiconductor Index. In June alone, the A-share semiconductor materials and equipment index achieved an excess return of 50.7% relative to the Philadelphia Semiconductor Index. From early June until the close this Friday, the A-share semiconductor materials and equipment index still posted a positive return of 6%, while related indices in Japan, the US, and South Korea were all negative. Trading in the North American AI chain has consistently focused on industrial progress, commercialization, profit realization, and valuation frameworks, essentially pricing the "shovel sellers." In contrast, the domestic AI chain is fundamentally still in the stage of building and shoring up weaknesses for the "shovel sellers," focusing on whether domestic models' capabilities are close to international frontrunners, whether domestic models can smoothly adapt to domestic computing power, and the degree of autonomy and control over domestic computing power processes. It is not about business models, actual performance, or valuations, but rather marginal changes, industrial narratives, and the sentiment of trend-following capital. Therefore, it requires continuous event catalysts and incremental fund inflows, and stock price fluctuations can deviate far from industrial progress in the short term. Conversely, once a trend-driven rally reverses, it is also difficult to use industrial logic to judge the extent of sector pullbacks. We analyzed the performance of 3,400 active equity products (ordinary stock type + partial equity hybrid type) on trading days when the semiconductor materials and equipment index rose more than 5%. On July 9th, 926 products saw single-day net value increases exceeding 4.5%. The semiconductor materials and equipment ETF experienced two waves of historically high subscriptions at the end of June and on July 10th, both occurring at the initial stage of market peak adjustments. Based on the historical holder structure of these products, we estimate these massive subscriptions following the initial adjustment from highs came largely from retail investors rather than institutions. All of this suggests that the fund repositioning process is difficult to conclude in the short term, and trend-following capital is unlikely to quickly return under pressure. Therefore, for the domestic AI chain, we prefer to wait for a more extreme point in staged sentiment and position swapping as an entry point for positioning.

Non-AI Sectors: Core Focus on Overseas Demand; Trade Environment is Paramount, Fed's Monetary Environment Secondary

Since last year, several core logics have supported non-AI external demand: China transforming its share advantage in dominant industries into global profits through anti-internal competition and supply control, robust industrialization demand in non-US markets, certain products with monopolistic power creating domestic-foreign price differentials through export controls to enhance profit margins, and the Federal Reserve entering a rate-cutting cycle. However, standing in the current year, these factors have, in fact, been challenged one by one. First, China-Europe trade, as the most critical link for Chinese manufacturing to realize global profits, faces an uncertain outlook. While it has not escalated into a trade war, new developments such as the Industrial Acceleration Act, the new version of the Foreign Investment Review Regulation, the new Cybersecurity Law, customs reforms (i.e., low-value parcel tariffs), the Foreign Subsidies Regulation, and the Chinese EV price commitment mechanism have all been proposed or are advancing this year, directly targeting China. Second, emerging market countries have generally faced capital outflows and currency depreciation this year, with shortages and price increases in energy and chemical products further crowding out consumption. In recent years, the highlights of China's export data were in regional structure, whereas this year, market discussions have universally shifted to product structure (AI vs. non-AI). Third, the marginal effects of certain export controls have begun to weaken. Last year, price differentials created by export controls were a novel change for investors, showing the market the possibility of influencing prices through supply control. However, this year investors are beginning to worry that the impact of such control policies is gradually shifting from short-term supply constraints to medium- to long-term supply chain restructuring and demand adjustments, as evidenced by rare earth price adjustments since Q2. Fourth, AI inflation and price pressures triggered by Middle East conflicts rapidly prompted a Federal Reserve pivot, and upward revisions to interest rate expectations further weighed on non-AI sectors. For prosperous non-AI sectors (representative ones like innovative drugs, lithium batteries, non-ferrous metals, chemicals), what has been weak is the stock price, not the performance, reflecting the market's long-term anxiety rather than immediate profit realization, leading to sustained valuation pressure.

Domestic Demand Plays: Weak Domestic Demand is a Capital Consensus; Foundation for a New Round of Incremental Policies Similar to "9.24" is in Place; Expectation-Driven Trading Has Already Commenced

Judging by apparent GDP growth, Q2 recorded 4.3%, already below the lower bound of the annual target. Based on past experience (such as before "9.24"), this signals a new round of policy impetus. However, the current macro environment differs from that before "9.24" in several aspects: first, prices are still in a moderate recovery trend, allowing nominal GDP growth to be maintained; second, industries like petrochemicals have been affected by Middle East conflicts impacting operating rates, and even after the first opening of the Strait of Hormuz, China's net crude oil imports have not recovered (while other East Asian countries showed significant inventory replenishment). Subsequent inventory replenishment could lead to spontaneous repair in related data indicators. Third, indicators related to new growth drivers are extremely strong, with the effectiveness of structural transformation overshadowing aggregate pressure. Fourth, measures since Q2 to standardize "invoice-based economies" and strengthen tax collection have created additional temporary pressures, which are not endogenous economic factors. Many of the above factors can also be classified as temporary, allowing for further observation or correction. This implies uncertainty over whether the slowdown in real GDP growth is sufficient to drive the introduction of policies exceeding expectations. However, the intensity of counter-cyclical adjustment policies in the second half of the year is expected to strengthen, at least in terms of expectation guidance and tone, and the accelerated advancement of the "six networks" is also expected to speed up broad fiscal expenditure. From this perspective, the nature of the domestic consumption sector's performance is closer to a trade betting on policy expectations rather than an allocation based on fundamental facts.

Some Concerns Already Recognized or Priced In; New Market Upswing in the Making; Awaiting Narrative Shifts to Unlock Medium-Term Potential

Despite various imperfections and issues, we believe many concerns have already been recognized and priced in by the market, as reflected in the sustained weakness of non-tech sectors since Q2 and the recent pronounced "catch-up decline" in tech sectors. Based on our tracked capital sentiment indicators, the market is also nearing the end of a panic sell-off: 1) Broad-based ETFs have shifted to large net subscriptions, with large funds injecting liquidity into the market again. On July 17th, net inflows into broad-based ETFs reached 64.8 billion yuan, the highest since 2025 second only to the 103.2 billion yuan on April 8th last year. 2) Our constructed investor sentiment indicator has fallen to its lowest level since the "September 24th rally" in 2024, with market sentiment temperature hitting bottom. 3) Implied volatility for stock index options has risen sharply. On July 17th, the implied volatility for CSI 1000 and CSI 300 stock index options reached historical high levels at the 94.6% and 94.0% percentiles respectively since 2023, the highest since 2025 second only to levels during the equivalent tariff shock, indicating that panic sentiment in market liquidity has been released. Recently, some non-AI sectors that fell earliest, with larger declines, yet possess prosperity and institutional holdings have shown signs of stabilization, such as innovative drugs and base metals. The price spread between the Sci-Tech Innovation Chip ETF and the Communication ETF is converging. The rolling 20-day return difference between the Sci-Tech Innovation Chip ETF and the Communication ETF rapidly declined from a peak of 40.8% on July 9th to 15.7% on July 17th, indicating that capital within the tech sector has completed a round of rotation from high to low and re-entered a steady state. After all, mainstream institutional capital prefers to hold core North American chain plays heavily, while possibly controlling positions in high-volatility domestic chains in the short term. This also means that after a round of catch-up declines in tech stocks, the A-share market has actually entered a new phase of brewing a fresh upswing. Although factors and narratives constraining upside potential persist in the short term, these factors are unlikely to affect the direction of the medium-term trend. With non-AI sectors bottoming out and AI sectors experiencing catch-up declines, structural differentiation convergence is the more likely path in such a market structure, and there is no need for pessimism towards the index.

Key Variables Requiring Attention at Present:

Capex Guidance from North American Cloud Providers at Month-End

Market Sentiment and Volume-Price Indicators

Federal Reserve's Policy Orientation and Expectations for the Politburo Meeting

Synthesizing our above analysis, breaking through this period of uncertainty essentially hinges on several key variables: 1) In the North American AI field, we are concerned about cloud providers' capital expenditure guidance in the short term. If the guidance is sound, the core segments of the North American chain remain a "resting" place for tech capital, but driving the sector to the next level still requires unexpected leaps in model capability or the emergence of new commercial cycle scenarios, which can only be awaited patiently. 2) In the domestic AI field, we are more concerned with sentiment and volume-price indicators, watching to see if sufficiently strong industrial catalysts can drive trend-following capital to re-enter and actively price assets after volatility subsides. 3) For non-AI sectors, three external variables are crucial: China-Europe trade negotiations, the Federal Reserve's monetary policy orientation, and exchange rates in emerging market countries. These three are key to unlocking the external demand narrative, which in turn is key to broadening the A-share market rally. Returning to the domestic front, the Politburo meeting expected at the end of the month is also critical. Current market narrative pressures mostly originate externally, which also implies that if market performance relies solely on external demand (whether AI or non-AI), resilience will be severely insufficient. This is also a core reason for the vast valuation gap between non-AI sectors in the Chinese and US markets (leading to widening overall market performance differences).

Consolidation and Brewing

Currently, we have the following four judgments. First, the index is in a phase transitioning from consolidation within a medium-term upswing (characterized by tech sector catch-up declines) to the brewing stage of a new upswing (requiring the initiation of new sectors with valuation upside potential), with short-term sell-offs nearing completion. Second, the North American AI chain may serve as a short-term safe haven within the tech sector, potentially seeing a rebound around the late-July guidance from North American CSPs. However, advancing to the next level requires a new leap in AI model/product capability and expansion of commercial monetization space. The key is breaking the cyclical stock valuation framework for hardware companies and achieving a new system-level uplift for both hardware and applications. Third, the domestic AI chain is highly dependent on catalysts and the intensity of trend-following capital. Currently, major catalysts have been largely realized, while the intensity of trend-following capital is unlikely to return quickly after experiencing sharp pullbacks. The relatively healthy margin financing situation domestically suggests there is currently no entry point characterized by a筹码出清-style washout. Convergence in valuations between domestic and North American plays is a more likely path. Fourth, the non-AI chain is characterized by rotational recovery, starting from innovative drugs and non-bank financials, transitioning to industrial chains like non-ferrous metals, chemicals, and lithium batteries, accompanied by policy expectation trading in the domestic demand chain. In terms of allocation, focus on convergence across three dimensions: 1) Convergence in excess returns of upstream AI hardware and price-increase plays relative to downstream cloud service giants; 2) Repair of the discount for non-AI industrial stocks relative to their overseas peers; 3) Convergence between tech and non-tech sectors.

Risk Factors

Intensification of friction between China and the US in technology, trade, and finance; North American CSP guidance falling short of expectations; Domestic policy intensity, implementation effectiveness, or economic recovery falling short of expectations; Unanticipated tightening of domestic and international macro liquidity; Further escalation of regional conflicts such as Russia-Ukraine and the Middle East; Slower-than-expected digestion of China's real estate inventory.