CICC: The AI wave may not be over; investment logic is shifting towards "competing in efficiency and returns."

CICC's research report states that the current AI debt risk is more reflected in the changes in financing structure (shifting from internal to external) rather than an excessively large debt scale or a deterioration in short-term debt repayment ability. The recent adjustments in AI assets resemble a logical reassessment and sector rotation influenced by multiple factors. The AI wave may not be over, but the investment logic is transitioning from the first stage of "scaling up" to the second stage of "efficiency and returns": cloud vendors need to demonstrate that capital expenditure can yield sustainable returns, chip and storage companies are facing a normalization of excess profits, while software, large model, and AI application enterprises will focus more on cost control, product efficiency, and commercialization capabilities. In the long run, this will drive the AI investment cycle into a more mature phase, contributing to the sustainable development of AI technology. The main points from CICC are as follows: As AI computing power investments continue to expand, major U.S. tech companies are gradually shifting from a "light asset, high cash flow" model to a "heavy asset, high capital expenditure" model, with rapidly increasing bond financing that has raised market concerns about AI debt sustainability. To assess risk, this paper introduces Minsky's Financial Instability Hypothesis as an analytical framework. Minsky categorizes corporate financing models into three types: hedge financing, speculative financing, and Ponzi financing, judging the stage of the debt cycle and potential risks from the perspective of repayment pressure. The firm found that although the major players in AI infrastructurethe five major cloud computing vendorsare accelerating their bond issuance, from a repayment capability perspective, their operating cash flows adequately cover principal and interest payments, indicating manageable repayment pressure. Among them, Microsoft Corporation (MSFT.US), Alphabet Inc. Class C (GOOGL.US), Meta (META.US), and Amazon.com, Inc. (AMZN.US) maintain operating cash flow coverage multiples that are still at a market-leading level, while Oracle Corporation (ORCL.US) shows a noticeable weakness. However, excluding capital expenditure, the free cash flow of Amazon.com, Inc. and Oracle Corporation has turned negative, and Alphabet Inc. Class C's quarterly free cash flow has also dropped to negative for the first time, indicating that as AI capital expenditure expands, some companies are increasingly reliant on external financing. Furthermore, the overall debt structure of the five cloud vendors is relatively robust. On one hand, their debt maturity is generally longer, with an average remaining bond duration of about 9.3 years, significantly higher than the U.S. corporate bond average of around 5 years; the proportion of short-term debt to total debt averages only 7.2%, far below the S&P 500 companies' average of 27.4%, indicating relatively limited short-term refinancing pressure. On the other hand, in the short term, the average effective interest rate on debt among the five cloud vendors is lower than current market rates, showing lower sensitivity to interest rate increases. Additionally, apart from Oracle Corporation and Amazon.com, Inc., the other three companies hold significant cash reserves, providing strong liquidity buffer capabilities. Overall, the current AI debt risk is more indicative of changes in financing structure (shifting from internal to external) rather than an excessively large debt scale or a deterioration in short-term repayment ability. Based on the Minsky framework, Microsoft Corporation and Meta remain typical examples of hedge financing; Alphabet Inc. Class C and Amazon.com, Inc. are showing signs of evolving from hedge financing to speculative financing, but they are still in the early stages; Oracle Corporation, however, has higher financial vulnerability due to continuously pressured free cash flow and negative net cash reserves. From a macro perspective, the leverage ratios and repayment rates of households and corporate sectors in the United States are at historically low levels, the banking system is well-capitalized, and Financial Institutions, Inc. maintains low leverage. Moreover, this round of AI investments relies more on bond market financing rather than bank credit expansion, with risks primarily absorbed by the capital market and relatively limited spillover effects on the banking system. Thus, the current AI debt risk remains within a controllable range, with a considerable distance from a true "Minsky moment." The firm believes that the recent adjustments in AI assets resemble a logical reassessment and sector rotation influenced by multiple factors. The AI wave may not have ended, but the investment logic is transitioning from the first stage of "scaling up" to the second stage of "efficiency and returns": cloud vendors need to demonstrate that capital expenditure can yield sustainable returns, chip and storage enterprises are facing a normalization of excess profits, while software, large model, and AI application enterprises will increasingly focus on cost control, product efficiency, and commercialization capabilities. In the long term, this will drive the AI investment cycle into a more mature phase, aiding in the sustainable development of AI technology.