A report from CICC indicates that AI capital expenditure is expanding rapidly, with large cloud providers' capex-to-revenue ratio potentially rising from 12% in 2023 to over 40% by 2027. This is projected to create approximately $3.5 trillion in external financing needs over the next five years, primarily absorbed by investment-grade bonds and private capital.

The core challenge lies in AI applications needing to generate roughly $1 trillion in annual revenue to cover debt costs, with a required profit margin of around 50% and a depreciation period of about five years. The report estimates that between 2027 and 2032, large cloud providers will enter a peak period for bond maturities, with an annual maturity size of approximately $28 billion. This represents a 60% increase compared to the 2024-2026 period, escalating refinancing pressure.

The Shift from Cash Flow to Debt

The capital expenditure-to-revenue ratio for large cloud providers has climbed from approximately 12% in 2023 to about 23% in 2025. According to market consensus, this ratio could further rise to over 40% by 2027, approaching or even exceeding the historical peaks of capital expenditure cycles in the internet and energy sectors.

Under the pressure of AI capital spending, cloud providers may shift their AI investment from operating cash flow to greater reliance on external financing. Based on market consensus estimates for operating cash flow and capital expenditure, CICC calculates that AI capital spending could create about $3.5 trillion in external financing demand over the next five years.

Where Does the $3.5 Trillion Come From?

In a baseline scenario, the report suggests the $3.5 trillion in external AI financing will be sourced from public equity ($0.4 trillion), investment-grade bonds ($1.5 trillion), leveraged finance ($0.3 trillion), asset-backed securities ($0.3 trillion), and private capital ($1.1 trillion).

Investment-grade bonds and private capital serve as the cornerstone of this financing. The former relies on cloud providers' balance sheet expansion and cash flow debt-servicing ability. The latter can meet the financing needs of higher-risk, larger-scale projects and can reduce cloud providers' capital outlay through off-balance-sheet financing structures, offering more flexible ways to bridge funding gaps.

The Trillion-Dollar Question: How to Repay AI Debt?

To meet debt servicing requirements and shareholder returns, assuming a ROIC of 10%, CICC estimates AI applications must ultimately generate approximately $1 trillion in sustainable annual revenue. Achieving this revenue scale by 2030 implies that AI application revenue needs to nearly double each year for the next five years.

Beyond revenue scale, profit margins and asset lifespan are critical. With an EBITDA profit margin of around 50% in the mature phase and an infrastructure depreciation period of about five years, investment is expected to generate positive returns. However, if profit margins fall below 20%, covering capital costs becomes difficult even with a longer asset life.

Can Refinancing Replace Cash Flow?

Large cloud providers issue long-term bonds with maturities typically ranging from 10 to 30 years, and infrastructure fund investment horizons also often exceed ten years. After data centers become operational, construction financing can be replaced by project bonds and securitization. Nevertheless, refinancing can only buy time, not replace cash flow. If project utilization, profit margins, and asset lifespan remain persistently below expectations, continuous rollover financing will instead increase leverage and financing costs.

The report calculates that large cloud providers will face a peak bond maturity period between 2027 and 2032, with annual maturities of about $28 billion—a 60% increase from the 2024-2026 period, intensifying refinancing pressure.

Opportunities and Risks for Financial Institutions

Banks can generate revenue from equity and debt underwriting, trading, M&A advisory, and project financing, while private capital and insurance institutions gain new long-term assets. In the second quarter of 2026, the six largest U.S. banks saw non-interest income rise 32% year-over-year. By June 2026, bank loans to non-bank financial institutions had increased about 25% year-over-year, with AI financing being a significant contributor.

Bank revenue is often recognized during the financing and construction phases, while credit risk only materializes when projects become operational and require refinancing. This creates a pattern of "revenue upfront, risk deferred."

Is AI a Financial "Bubble"?

Currently, AI investment is primarily driven by large cloud providers with strong cash flows. Equity and subordinated capital can also absorb losses before bank senior loans. Therefore, even if some projects underperform, risks are more likely to first manifest as valuation adjustments, a slowdown in capital expenditure, and localized credit losses, rather than immediately triggering a systemic financial crisis.

However, the rapid growth of external financing, particularly private capital and leveraged finance, creates a financial spillover effect. If commercial growth persistently lags behind capital spending, risks will gradually shift from valuation corrections to credit quality deterioration. This deterioration could then be transmitted through the financial system via revolving financing, private credit, and securitization channels.

The scale of financing itself is not the problem. The true test is whether a cash flow cycle can be established before financing costs rise and asset values depreciate.