Source: China Galaxy Securities Co., Ltd.

How large is the current bubble level in US stocks: (1) From the perspective of current short-cycle indicators, the valuation of the US technology bubble has not reached extreme levels, earnings growth has largely digested valuation expansion, and the magnitude of PE expansion is far smaller than the price increase. (2) The funds occupied by the stock market have entered a high range, exceeding the level during the 2000 internet bubble. (3) The Buffett Indicator trend shows that the upward speed of total market capitalization has been too fast, entering a relatively high-risk range. (4) The Shiller CAPE for the Nasdaq market is in an overvalued range, but has not entered an extreme zone. (5) US stocks are currently significantly ahead of stock markets in other countries such as Germany, and the rise in the index ratio comes more from earnings premium. Industry perspective: Analyzing current signals related to the AI bubble: (1) Upstream computing hardware and infrastructure: The upstream is currently more about profit realization. Corporate capital expenditure will rise to a certain extent in 2026, revenue growth exceeds Capex growth, and overall capital intensity is trending downward. Revenue growth is mainly driven by demand and prices, rather than piled-up capital investment. (2) Midstream cloud platforms and model layer: The risk of nonlinear rupture in the supply-demand relationship with the upstream is not yet obvious, but there are hidden bubble concerns. There is a scissors gap between the cloud business growth rate of major North American companies and the inventory growth rate of upstream chip manufacturers, and the scissors gap continues to widen. The pace of stocking at the chip end has systematically exceeded the revenue growth rate at the cloud end, and the midstream of the AI industry chain is bearing an increasingly large exposure to inventory and capacity risk. The pull of capital expenditure on revenue is weakening, meaning that the monetization capability of AI is currently below expectations. (3) Downstream applications: At present, most AI downstream application scenarios remain between "pilot" and "partial scale-up." The main risk comes from multiple mismatches in the value realization chain. What stage is the US stock AI bubble currently in: The US stock AI bubble is currently in a complex stage of partial high valuation superimposed on earnings expectations about to peak, and high dependence on debt-financed expansion. The AI industry is in the "late stage of frenzy," and 2027 is regarded as a key dangerous node. (1) At the macro level, as the 10-year US Treasury yield breaks through 5%, the growth model of the AI industry, which is highly dependent on debt financing, is facing severe pressure. (2) At the valuation level, the bubble has been significantly compressed over the past year. The valuation adjustment of AI was completed alongside upward revisions to earnings expectations, and the market has not ignored fundamentals. (3) At the industry level, AI technology giants are using increasingly depleted free cash flow to fill AI capital expenditure, while the huge funding gap is filled by issuing debt. The exponential rise in leveraged loans and high-yield bonds has spread AI debt risk from the technology industry to the entire US financial system. 2027 will be a key node for testing the financial health of US AI technology giants. If CDS spreads and corporate bond credit spreads remain at high levels, the bursting of the US stock AI bubble will begin first with default events in the credit market. (4) At the theoretical level, in the current environment, discussions on AI regulation are heating up, but a systematic institutional reconstruction has not yet formed; financial capital is still dominant, and debt financing and off-balance-sheet tools are still expanding. AI is in the late stage of frenzy, but has not yet reached a turning point. Embrace volatility and rational allocation strategies: In the complex macro environment where the latter half of the AI frenzy intersects with high interest rates, finding companies and assets that are not afraid of bubbles and can transcend cycles is not only a defensive strategy, but also the core of obtaining excess returns over the long term. The key lies in finding companies with strong free cash flow that can substantially improve ROIC without relying on continuous high capital expenditure, and in positioning in barrier assets where value shifts to non-model factors, such as upstream hard technology that controls absolute computing power or patents, platform giants that monopolize data channels and key workflows, and utility companies that provide efficient energy solutions. Risk warning: Risk that US domestic policy intensity and effects fall short of expectations; risk of repeated interest rate hike expectations; risk of unstable market sentiment.