Zhi Tong Cai Jing·2025-11-19 22:53Core Viewpoint - The report from Shenwan Hongyuan indicates that under a neutral assumption, the risks associated with AI soft constraints by 2026 are manageable, with a focus on whether the penetration rate of B-end enterprises can improve and whether cash flow remains stable. Companies in the US stock market with relatively stable expansion are worth monitoring. However, hard constraints, particularly regarding power supply for computing infrastructure, will become more pronounced, necessitating higher verification requirements for ROI [1][2]. Group 1: Market Trends and AI Investment - Since 2023, the current AI market in the US has been ongoing for three years, primarily focusing on valuation increases and widespread performance in AI infrastructure and applications [2]. - The AI-related industries (technology, communication) are expected to contribute limited valuation in 2025, with increased reliance on debt financing amid unclear ROI for AI investments [2]. - Concerns about an "AI bubble" stem from the gap between the time required for new technologies to generate economic scale effects and the optimistic expectations of capital market returns [3]. Group 2: Constraints and Adjustments - Hard constraints often lead to periodic adjustments in the AI sector, with the "Buy the Dip" strategy still showing effectiveness in the US stock market. Adjustments have been triggered by macro liquidity tightening and concerns over computing power, algorithms, and electricity [4]. - The report outlines that the AI sector has faced multiple adjustments since 2023, with each decline exceeding 10% and lasting over a month, influenced by factors such as liquidity tightening and supply shocks [4]. Group 3: ROI and Financial Metrics - The report emphasizes that the ROI for AI investments is sensitive to GPU depreciation, with rising debt financing costs posing tail risks. The total AI investment commitment projected by Trump for 2025 is approximately $3.8 trillion, aimed to be completed by 2028 [6][7]. - Current AI penetration in US enterprises is around 10%, with higher rates in information-intensive sectors, expected to rise to over 30% in the next six months [6]. - The report highlights that the profitability of AI applications varies, with B-end applications showing higher margins compared to C-end applications, which generally have negative margins [6]. Group 4: Debt and Liquidity Risks - The overall debt pressure on AI infrastructure is manageable, but significant differentiation among companies is expected by the second half of 2025. Long-term debt levels related to AI hardware have been gradually increasing, with a decline in the free cash flow to debt ratio [7][8]. - The report warns that in a tightening liquidity environment, the probability of debt risks increases, particularly for companies with lower asset quality [8]. - The current static PE ratio for the US stock market is 28x, with historical data indicating a low success rate for holding stocks at this valuation over three years [10].