Key Takeaways
- • Hyperscaler AI-related debt now accounts for approximately 10% of the three largest intermediate and long-term investment-grade corporate bond ETFs, up from about 7% three years ago, according to CFRA Research data cited by Pensions & Investments.
- • CFRA estimates AI-related capex by hyperscalers will reach about $1.2 trillion next year, meaning these firms will need external financing beyond internal cash flows.
- • Hyperscaler bonds yield 7.1% on average versus 6.5% for non-AI bonds, per CFRA's analysis of BlackRock's IGLB holdings as of Oct. 13; Oracle yields about 8.2%.
- • Oracle, with $125 billion in notes payable and other borrowings as of August 31, is viewed by CFRA as the name to focus on among the largest IG ETF holdings, while Amazon, Microsoft, Meta and Alphabet are seen as stronger credits.
- • DWS's George Catrambone says insurers are increasingly concerned about concentration risk, and no AI issuer has guaranteed it will keep its investment-grade ratings.
- • Defiance launched the AIBN AI infrastructure bond ETF on Oct. 20, while BondBloxx and Allspring have filed with the SEC for similar products.
Summary
AI-related debt is becoming a more visible component of U.S. investment-grade corporate bond indexes and ETFs as hyperscalers expand borrowing to fund AI infrastructure and data centers. According to data from CFRA Research cited by Pensions & Investments, the Vanguard Long-Term Corporate Bond ETF (VCLT), the iShares 10+ Year Investment Grade Corporate Bond ETF (IGLB) and the State Street SPDR Portfolio Long Term Corporate Bond ETF (SPLB) each hold approximately 10% of their portfolios in AI-related debt, up from about 7% three years ago. Because hyperscalers tend to issue long-term bonds, the exposure is concentrated in intermediate and long-term funds.
Aniket Ullal, head of ETF research and analytics at CFRA Research, said hyperscaler AI-related capital expenditure is expected to reach about $1.2 trillion next year, more than internal cash flows can fund. In VCLT, the top five issuers are AT&T, Oracle, Amazon, Meta Platforms and Verizon. CFRA views Amazon, Microsoft, Meta and Alphabet as stronger credits than Oracle, given higher ratings and cash flows. Oracle reported $125 billion in notes payable and other borrowings as of August 31 in its most recent earnings call. Ullal said Oracle is the name to focus on for active IG corporate bond investors.
CFRA found hyperscaler bonds yield 7.1% on average versus 6.5% for non-AI bonds, with Oracle at about 8.2%, SpaceX at about 7.9% and Meta at about 7.4%. CFRA recommends investors seeking higher income without high-yield exposure favor ETFs weighted toward higher-rated hyperscalers, noting that higher yields reflect greater credit risk. George Catrambone, head of fixed income Americas at DWS Group, said insurance clients are asking how much concentration risk they carry and whether to keep participating in AI-related deals. He noted that no AI issuer has guaranteed it will maintain investment-grade ratings.
JoAnne Bianco, senior investment strategist at BondBloxx, took a more constructive view, arguing the corporate bond market has absorbed AI issuance and that spreads would not be this tight if investors were seriously concerned. She said the key watch item is new AI-related issue supply, warning that forcing debt on bondholders would not work, while noting strong investment-grade credits remain. Several issuers are responding with products: Defiance launched the AI infrastructure bond ETF (AIBN) on Oct. 20, investing in hyperscalers, semiconductor and AI compute companies, and BondBloxx and Allspring have filed with the SEC for similar funds. Ullal said investors wanting to reduce AI debt exposure could consider broad aggregate funds such as the iShares Core U.S. Aggregate Bond ETF (AGG), avoiding long-dated corporate ETFs, or using short-term corporate bond ETFs and Treasuries.
Aniket Ullal, head of ETF research and analytics at CFRA Research, said hyperscaler AI-related capital expenditure is expected to reach about $1.2 trillion next year, more than internal cash flows can fund. In VCLT, the top five issuers are AT&T, Oracle, Amazon, Meta Platforms and Verizon. CFRA views Amazon, Microsoft, Meta and Alphabet as stronger credits than Oracle, given higher ratings and cash flows. Oracle reported $125 billion in notes payable and other borrowings as of August 31 in its most recent earnings call. Ullal said Oracle is the name to focus on for active IG corporate bond investors.
CFRA found hyperscaler bonds yield 7.1% on average versus 6.5% for non-AI bonds, with Oracle at about 8.2%, SpaceX at about 7.9% and Meta at about 7.4%. CFRA recommends investors seeking higher income without high-yield exposure favor ETFs weighted toward higher-rated hyperscalers, noting that higher yields reflect greater credit risk. George Catrambone, head of fixed income Americas at DWS Group, said insurance clients are asking how much concentration risk they carry and whether to keep participating in AI-related deals. He noted that no AI issuer has guaranteed it will maintain investment-grade ratings.
JoAnne Bianco, senior investment strategist at BondBloxx, took a more constructive view, arguing the corporate bond market has absorbed AI issuance and that spreads would not be this tight if investors were seriously concerned. She said the key watch item is new AI-related issue supply, warning that forcing debt on bondholders would not work, while noting strong investment-grade credits remain. Several issuers are responding with products: Defiance launched the AI infrastructure bond ETF (AIBN) on Oct. 20, investing in hyperscalers, semiconductor and AI compute companies, and BondBloxx and Allspring have filed with the SEC for similar funds. Ullal said investors wanting to reduce AI debt exposure could consider broad aggregate funds such as the iShares Core U.S. Aggregate Bond ETF (AGG), avoiding long-dated corporate ETFs, or using short-term corporate bond ETFs and Treasuries.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.