The AI debt boom: Balancing risk and opportunity

The AI investment boom is generating unprecedented debt issuance, but strong demand still supports the U.S. investment-grade market while issuer concentration and uncertain AI revenues argue for selective exposure.

• Alphabet, Amazon, Meta, Microsoft and Oracle are expected to spend at levels in 2026 and 2027 that dwarf earlier technology megaprojects such as the Manhattan Project, the Apollo program and the internet buildout.
• The five hyperscalers issued $240.7 billion of debt year to date through August 31, including $66 billion in non-dollar currencies; adding SpaceX and Nvidia brings total issuance across currencies to $290.7 billion, versus $321 billion of 10-year U.S. Treasury issuance over the same period.
• Demand has absorbed the supply: U.S. investment-grade corporate spreads ended August at 78 basis points, unchanged from the start of 2026, while annuity sales exceeded $100 billion for the eleventh consecutive quarter and foreign private-sector inflows reached $448 billion over the 12 months to June 15, 2026.
• Hyperscaler and other AI-related spreads, particularly for bonds with maturities of at least 10 years, have widened; Capital Group expects the five major hyperscalers to issue more than $25 billion in the final four months of 2026 and between $100 billion and $200 billion in the U.S. dollar investment-grade market during 2027 and 2028.
• Consensus forecasts suggest that the upfront cost of AI investment will exceed operating cash flow in 2027 and 2028, with cash flow turning positive in 2029, leaving uncertainty over whether future revenue will justify the spending.

Takeaway: Investors should favor liquid bonds from diversified issuers with fungible infrastructure, selected data-center and industrial suppliers, and carefully examined structured vehicles, while remaining cautious about concentration, semiconductor balance-sheet risk and the uncertain conversion of AI investment into revenue.