Guojin Strategy: The market rebound may gradually enter a plateau phase, and the elasticity of physical assets will also become apparent.

1. The unveiling moment is gradually passing, and the market may be entering a plateau phase. The rapid decline of AI assets globally in July was caused by multiple factors, including heightened concerns about internal industry fundamental contradictions, geopolitical risks between the U.S. and Iran, and rising expectations of liquidity tightening. Since August, however, with new signals of negotiation between the U.S. and Iran and a weakening of the Federal Reserve's tightening expectations following the post-World Cup U.S. employment and inflation data, the performance of technology assets has gradually improved. However, risks related to contradictions within the industry chain and excessively high fundamental expectations still exist. The phenomenon of "earnings reports widely exceeding expectations, yet stock prices plummeting" has occurred repeatedly. The recent rebound in technology assets is largely driven by structural factors or temporary easing of market concerns about specific segments of the industry chain, rather than a reversal of market expectations leading to a new trend for the entire sector. As the upward momentum driven by liquidity weakens, the overall rebound of technology assets may also enter a bottleneck period. Higher interest rates are more favorable for technology stocks, and a decrease in interest rates may drive a rebalancing of market styles in the U.S. equity market. Similarly, in the A-share market, the style of the market continues to exhibit a rebalancing trend, with non-AI sectors like healthcare, real estate, and food and beverage also recording good gains. Within technology assets, there is a divergence, with communication showing strong performance while electronics have begun to show signs of stagflation. It is worth mentioning that the current cycle of the AI industry is dominated by the United States. Among the driving factors of the recent rebound in the U.S. AI industry chain, whether it is the commercialization of AI applications, the capital expansion and preliminary commercialization pathways of cloud vendors, or the production capacity expansion of American optical module manufacturers and reports of U.S. government import restrictions on Chinese optical module manufacturers, they all indicate that the U.S. government and enterprises are attempting to redistribute profits and recirculate funds within the AI industry chain. This includes large-scale supply agreements between Samsung, SK Hynix, and U.S. tech giants, as well as TSMC's increased investment in the U.S., illustrating their deepening ties to the U.S. AI industry chain. In contrast, due to geopolitical constraints, China and the U.S. are already significantly decoupled at multiple stages of the AI chain, and the closely connected overseas computing power chain is suppressed by U.S. government import policies. This has led to the narrative of the recent rebound of U.S. tech stocks not being easily reflected or transmitted to the A-share market. In fact, alongside the increasing "K-shaped divergence" on the export side, under the dominance of AI-related exports, exporters' willingness to settle foreign exchange has declined, and the scale of overseas assets is expanding again. 2. Matching debt and income: Time seems not to be on the side of AI assets. As AI capital expenditures gradually shift from self-generated cash flow to external debt financing, the market's focus is shifting from "AI investment expansion" to "verification of AI investment returns." From the downstream demand perspective, since the rapid proliferation of AI agents represented by Claude Code and OpenClaw since the second half of 2025, enterprises' willingness to pay for AI products has only begun to significantly rise. As of now, the core demand from enterprises for AI products is still primarily in the calling of APIs (currently accounting for over 50% of AI product expenditure), which means that the narrative supporting the debt expansion of the AI industry chain mainly derives from the nonlinear growth of the Token market scale. The ideal commercialization path is: penetration and diffusion of downstream agentsgrowth in Token revenue (market scale)continuation of a supply-demand imbalance in computing powerfurther increase in capital expenditure by cloud vendors and large model vendorsresulting in reasonable large-scale debt expansion. Therefore, if the expansion rate of the Token market scale lags behind that of debt, signs of rising leverage ratios in the industry chain will appear, and market concerns about debt pressure will intensify. Since 2026, the leverage ratio indicators have exhibited an overall downward trend, and a clear commercialization path has provided a favorable environment for the rise of technology markets. However, since mid to late July, we have observed a renewed increase in leverage ratios. If this trend continues, market concerns about debt pressure will continuously deepen, constraining cloud vendors' further debt expansion capabilities. Large model vendors and cloud vendors themselves will also need to reassess the matching between capital expenditure and returns. On the funding supply side, non-bank financial institutions (such as VC/PE) are important marginal funding sources for AI cloud computing capital expenditures. Recently, commercial banks are no longer increasing credit support for them, which may itself signal a warning from conservative investors about the financing debt pressures in the AI industry chain. 3. The bottom of physical assets is beginning to emerge. For physical or HALO assets, AI investment is still at relatively high levels, and the actual rate of decline in interest rates is limited. The U.S. economy is only gradually receding from a high peak, but it is far from needing to cut interest rates. The Strait of Hormuz is still closed, and shipping volumes are recovering slowly. A comprehensive rebound in demand from non-U.S. economies, especially emerging markets, may still need to wait. However, compared to AI needing to prove its forthcoming new round of expansion, physical assets are, in a sense, on time's side; the bottom position has gradually become clear. For commodities, the current actual interest rates are beginning to decline and converge with inflation expectations, and their equal-weighted prices have started to bottom out and rise. In the future, as interest rates and the dollar weaken, the performance gap between the U.S. economy and non-U.S. economies, as well as the split between technology assets and HALO assets, will begin to close. Referencing the rapid rise of AI leverage ratios to high levels from Q4 2025 to Q1 2026, we could see a pattern emerge again with declining real yields on U.S. Treasuries, with HALO assets showing greater elasticity compared to the overall market. 4. During this tug-of-war period in the market, greater attention needs to be paid to changes outside of "attention assets." As the phased easing brought about by the denominator side approaches its end, and with no systemic major catalyst emerging on the numerator side to pull in different directions, the AI assets currently capturing the market's attention and non-AI related physical assets have both entered a plateau phase. However, looking ahead, the advantage of AI-related assets is their continued high prosperity, with unfavorable factors accumulating. Still, based on past experiences, market investors are more willing to pay an option premium for stock prices, meaning that non-linear changes are likely to reemerge. However, the option premium itself will erode as time goes on. For the broader demand side of physical assets, considering the friendly face of time erosion, the bottom of fundamentals and asset prices is being confirmed. If the ratio of new debt issued in the AI industry chain to revenue continues to rise, this will significantly constrain the financing capabilities and capital expansions of cloud vendors and large model vendors. Alongside the convergence of fundamentals between the U.S. economy and non-U.S. economies, scenarios resembling the rise of AI leverage ratios at the end of 2025 and the beginning of 2026, with declining real yields on U.S. Treasuries, will reappear, and the elasticity of physical assets will also manifest. Regarding specific allocation recommendations, we suggest: First, the rebound of commodities has been established; the combination of energy and metals is shifting from an overall downward trend since March to an overall upward phase. The U.S.-Iran conflict only determines internal distribution relationships, and considering the short-term escalation of conflict, the current resource ranking is as follows: Energy (coal, oil and petrochemicals), non-ferrous metals (gold, copper, aluminum). Second, the dividend style benefits from the switch of absolute returners; high-dividend, low-volatility, and stable cash flow dividend assets remain the core direction for the reflow of absolute return funds. Third, the resonance between southern country narratives and Chinese manufacturing is evolving, with attention on areas such as construction machinery, power grid equipment, and refining. Risk Warning Domestic economic recovery is weaker than expected: If subsequent domestic economic data unexpectedly weakens, then the assumptions regarding the capital market's recovery driven by improving fundamentals mentioned in this article become inapplicable. Overseas economy significantly declines: If the overseas economy unexpectedly declines, the global manufacturing recovery may pause, and demand for physical assets will also slow down. This article is compiled from the WeChat public account Yilin Strategy Research, edited by GMTEight: Chen Yufeng.