Shen Wan Hongyuan: The tipping point of the US AI “money-burning model” may depend on the fulfillment of profit expectations

The Zhitong Finance App learned that Shen Wan Hongyuan released a research report saying that as free cash flows from leading cloud vendors turn negative one after another, market concerns about the sustainability of AI capital expenditure are increasing day by day. The market's pricing logic for AI is shifting from “capital expenditure expansion” to “financing and cash flow constraints.” Looking back, whether profit expectations can be fulfilled at the micro level and whether the Federal Reserve will raise interest rates at the macro level are the keys to whether AI's “money-burning model” can continue. The bank believes that at a micro level, compared to the Internet bubble period, the current financial indicators of leading US technology companies are steady, and the increase in profit is far greater than the increase in supermarket values; at the macro level, comprehensive employment is weak, inflation is cooling, and economic momentum is slowing down, and fundamentals do not yet support the Fed's immediate interest rate hike.

Shen Wan Hongyuan's main views are as follows:

Since the end of 2025, as free cash flows from leading cloud vendors have turned negative one after another, market concerns about the sustainability of AI capital expenditure have increased day by day. As corporate leverage ratios increase and liquidity becomes tight, where is the tipping point of the “money-burning model” of AI capital expenditure?

1. Hot Thoughts: What are the boundaries of the “money-burning model” of American AI?

(1) AI financing: Shifting from cash flow to external financing, corporate bonds, private equity credit, etc. may be the main financing channels

The market's pricing logic for AI is shifting from “capital expenditure expansion” to “financing and cash flow constraints.” Take the financial report for the second quarter of 2026 as an example: when tech giants raised capital expenditure guidelines, but free cash flow was significantly compressed, and at the same time, when the market questioned the ability to cash out cloud and computing power assets, the market instead penalized “capital expenses”, and stock prices fell in response (Google (GOOGL.US), META (META.US)).

As the scale of AI capital expenditure increases, internal cash flow is stretched, and dependence on external financing increases. As of August 11, the market expects the capital expenditure of the five large cloud vendors to be 769.2 billion US dollars in 2026, up 40.9 billion US dollars from before March; their capital expenditure may exceed operating cash flow, and the scale of AI investment is expected to continue to rise over the next five years (BIS forecast).

The importance of AI in various financing channels is increasing day by day. In the future, corporate bonds, private equity credit, etc. may become an important financing tool for AI infrastructure. In the first half of the year, AI issuance accounted for 34%, 40%, 85%, and 58% of investment-grade corporate bonds, high-yield corporate bonds, VC, and IPO+ issuances, respectively. According to FSB forecasts, AI infrastructure spending in 2025-28 may reach 2.9 trillion US dollars, mostly supported by external financing.

(2) What is the pressure on AI financing? Risk indicators such as credit spreads and private equity credit default rates have released “early warning” signals

Corporate bonds are the main financing tool for large cloud vendors. The volume of issuance has led to widening interest spreads, and stock prices have shown an inverse relationship with CDS. Since the beginning of the year, corporate bonds issued by the five major cloud vendors have exceeded 200 billion US dollars (investment grade), and the share of AI issuance in 10Y+ corporate bonds has reached 40%; interest spreads on investment-grade technology credit bonds have widened, mainly reflecting increased supply and uncertain returns.

The scale of private equity credit continues to expand, and it has become an important off-balance sheet financing channel for large cloud vendors. By the third quarter of 2025, the global private equity credit scale had exceeded $2 trillion (North America accounted for 72%) and may grow to 4.5 trillion US dollars in 2030; as of the second quarter report of 2026, the off-balance sheet debt of the five major cloud vendors had reached 2.6 trillion US dollars; the private equity default rate increased, and the share of PIK that was forced to amend agreements in BDC increased, and the decline in software service stocks in the early period released an “warning” signal.

The bank has built a complete AI “financing risk” monitoring system that can be tracked at high frequency. The first is “stock leverage”. Although leading cloud vendors expect the debt-to-equity ratio to continue to improve, the expected net debt/EBITDA continues to decline, indicating strong profits; the second is “external financing,” which covers high-frequency indicators such as credit spreads, equity financing amounts, BDC stock prices, and data center ABS issuance amounts.

(3) The tipping point of AI's “money-burning model”? Fulfilling profit expectations is the key; the Fed's interest rate hike may be an important catalyst

How high is the return on AI capital expenditure to be “economical”? The bank models capital expenditure and the annualized revenue of the big model, and estimates that the big model and cloud computing links will each obtain part of the operating profit retained in the big model revenue link. The bank expects that if ARR of 350 billion US dollars is added in 2027, that is, the ARR level reaches 500 to 600 billion US dollars, it can support more than 1.4 trillion US dollars in capital expenditure.

The bank has built a high-frequency monitoring system for AI's “profit prospects”. The tipping point of America's AI “money-burning model” may depend on the fulfillment of profit expectations. Therefore, the bank has built an AI “profit expectation” monitoring system, which includes profit index expectations for ARR, leading cloud vendors, and the technology sector. Over the past year, the above indicators have all been in an upward or upward trend.

Looking back, whether profit expectations can be fulfilled at the micro level and whether the Federal Reserve will raise interest rates at the macro level are the keys to whether AI's “money-burning model” can continue. At the micro level, compared to the Internet bubble period, the current financial indicators of leading US technology companies are steady, and the increase in profit is far greater than the increase in supermarket values; at the macro level, comprehensive employment is weak, inflation is cooling, and economic momentum is slowing down. Fundamentals still do not support the Fed's immediate interest rate hike.

Risk Alerts

The oil price center moved higher than expected; Walsh's policy stance was “hawkish”; the US economic slowdown exceeded expectations.