The logic of capital is undergoing a significant shift. While AI assets remain in a high-growth phase, the mismatch between mounting debt and revenue generation is eroding their time premium. In contrast, physical assets have likely found their bottom and are poised for a rebound. As US Treasury yields decline, the elasticity of commodities and other physical assets is set to be fully unleashed. The current top picks are energy metals, high-dividend assets, and export-oriented manufacturing.
Where the Rally Stands
Since August 2026, global risk assets have staged a recovery, with previously battered tech stocks leading the gains. However, the risks of industrial chain contradictions and overly optimistic fundamental expectations persist. The phenomenon of "earnings beating expectations, yet stocks falling sharply" is occurring repeatedly. Rallies driven by fundamental improvements have been largely structural or have only temporarily alleviated concerns about specific segments of the chain, rather than shifting the overall market narrative towards a new sustained uptrend. As the momentum from a liquidity-driven rally weakens, the broad tech rebound of this cycle may be hitting a bottleneck. As we have previously discussed, rising interest rates can be a signal of accelerated AI investment, while falling rates may drive a rebalancing of market styles in the US.
Domestically, the market's dilemma is clear: compared to the US stock market, which is nearing new highs, China's A-share tech stocks have experienced more significant declines, a relatively weaker rebound, and a slower reduction in volatility. In contrast, South Korean stocks, which also fell sharply, have shown stronger recovery elasticity. It is important to note that the AI industrial cycle is inherently led by the US, with the foundational capital inflow being global funds returning to USD-denominated assets. The recent rebound in the US AI chain, whether through the commercialization of AI applications, the initial closing of the loop between cloud providers' capital expenditure and revenue, or US government restrictions on Chinese optical module imports, all point to an attempt by the US government and companies to reallocate profits and guide capital flows back to the US. The massive supply agreements between Samsung, SK Hynix, and US tech giants, along with TSMC's increased investment in the US, further demonstrate this deep integration. The trade surplus generated by South Korea from this AI wave has flowed back to the US through financial channels. In contrast, China's A-share market faces two constraints. First, there is significant decoupling between the US and China in several segments of the AI supply chain, with limited direct mapping for cloud providers and AI applications. Second, China faces hurdles in earning US dollars: the overseas computing power chain, where it is closely linked, is suppressed by US import policies, and even when profits are realized, an increasing portion of export proceeds is being left un-hedged, similar to the phenomenon in South Korea where high-profit realizations do not lead to capital repatriation, thereby limiting the liquidity elasticity of the A-share market. In this context, the divergence in performance between US and Chinese tech stocks has an intrinsic logic.
AI's Self-Proof Phase: The Race Between Debt and Revenue
As AI capital expenditure increasingly shifts from internal cash flow to external debt financing, market focus is moving from "AI investment expansion" to "verification of AI investment returns." On the demand side, corporate willingness to pay for AI products truly took off from the second half of 2025 with the rapid adoption of AI agents like Claude Code and OpenClaw. However, the core demand for AI products remains API calls, accounting for over 50% of AI product spending. This means the narrative supporting the AI industry's debt expansion is fundamentally based on non-linear growth in the token market. The ideal path is: downstream agent penetration drives more frequent API calls, leading to token revenue growth, which sustains the supply-demand imbalance for computing power, allowing cloud providers and model companies to increase capital expenditure, thus justifying large-scale debt issuance. If token market expansion slows down relative to debt, leading to a rise in the industry's leverage ratio, market concerns about debt pressure will intensify.
From the second quarter of 2026, the leverage ratio showed a downward trend, and the clarity of the commercialization path provided a tailwind for tech stocks. However, since mid-to-late July, model companies' "price-for-volume" strategy has not significantly expanded the total token market, and the leverage ratio has started to rise again. On the supply side of capital, conservative investors, such as banks, are showing signs of caution regarding the debt pressure on the AI financing chain, gradually tightening credit directed towards venture capital and private equity. Looking ahead, without a new, non-linear advancement in AI products or technology to drive a fresh round of rapid revenue expansion, the mounting debt burden will compress the industry's ability and space for further large-scale financing expansion, thereby reducing the support for real interest rates. The current AI stock prices embed the value of a call option, anticipating a non-linear change, but the passage of time itself is a drain on this value.
Physical Assets: Confirmation of a Bottom and the Waiting Game
AI investment remains at a relatively high level, and the decline in real interest rates has been limited. The 10-year real yield has started to turn, but the 30-year yield continues to fluctuate at high levels. The recovery of traffic through the Strait of Hormuz is slow, and a broad recovery in demand from non-US economies, especially emerging markets, may still require patience. However, the bottom is gradually becoming clear. For commodities, the real yield is beginning to decline and converge with inflation expectations, and their equal-weighted prices have started to bottom out. A further performance rebound is still pending, which is the core reason for the current pullback. In the future, if we see a repeat of the scenario from late 2025 to early 2026, where a rapid rise in AI leverage leads to a fall in US real yields, a true upward trend for HALO assets (Hard Assets, Liquid Assets, and Other) will emerge.
Navigating the Period of Uncertainty
The market is still in a period of uncertainty. Our recommendations are as follows. First, the bottom for commodities is confirmed. The commodity complex, comprising energy and metals, is transitioning from a broad decline since March to an overall upward phase. The US-Iran conflict only determines the internal distribution of gains. The current order of preference is: energy (coal, oil & petrochemicals) and base metals (gold, copper, aluminum). Second, the high-dividend style benefits from the rotation of absolute return investors. Dividend assets, characterized by high yields, low volatility, and stable cash flow, remain the core target for capital seeking absolute returns. Third, the story of the Global South is converging with Chinese manufacturing strength. Focus on areas like engineering machinery, grid equipment, and oil refining.
Risk Warning
Domestic economic recovery may fall short of expectations, and the overseas economy could experience a significant downturn.
Comments