AI Borrowers Take 44% of Long-Dated US Investment-Grade Bonds

Artificial-intelligence-related borrowers sold about 44% of this year's long-dated US investment-grade bond supply outside the financial sector, according to PIMCO, a jump from roughly 35% in 2025 and just 6% in 2024. The figure covers US dollar investment-grade issuance by non-financial companies with maturities beyond 15 years, based on Bloomberg data through Oct. 30 and published in PIMCO's Nov. 9 note, The Credit Market Lens.

The concentration problem is the point for bond investors. Anyone who holds a long-duration corporate bond index or fund now owns a growing share of AI capital spending, whether or not they chose to. "The AI investment boom is increasingly becoming a bond-market story," Kevin Thozet, a member of Carmignac's investment committee, told CNBC.

How big is the AI borrowing wave?

Hyperscaler bond issuance will total about $250 billion in dollar-denominated debt this year, according to a Goldman Sachs forecast cited by CNBC. Estimates for the wider AI ecosystem — chipmakers, utilities and data center developers — range from $300 billion to $570 billion.

The largest deals show the scale:

• Microsoft sold $25 billion of bonds on Oct. 29.
• Apple priced a $7.5 billion seven-part deal on Nov. 5.
• Alphabet plans to sell its first Hong Kong and Singapore dollar bonds.

Lower down the credit curve, the same boom looks expensive. CoreWeave's 2032 bonds yielded more than 16% last week, and MLQ.ai reported on Nov. 9 that lenders are pulling back from AI debt.

What happens after the midterms?

PIMCO's multi-asset strategist Lotfi Karoui expects the AI share of supply to stay high even if politics intervene. US midterm results could bring restrictions on data center development and slow near-term issuance, he wrote, but stretching spending over more years does not shrink the total funding need of the buildout. Such limits are "more likely to change the journey than the destination," Karoui wrote, and "the concentration challenge is unlikely to dissipate."

Do the usual hedges still work?

Partly. From Sept. 30 to Nov. 6, the Nasdaq 100 lost about 11% while 10-year Treasuries returned about 4%, by PIMCO's measure — although the hedge worked less well in early November. Corporate sectors far from AI capital spending, such as investment-grade financials and high-yield energy, also diversify, but PIMCO finds their spreads tight relative to their risk-adjusted potential. "Both, however, are expensive, in our view," Karoui told CNBC.

What does PIMCO recommend instead?

The firm names three diversifiers, disclosed with the caveat that it manages funds holding these assets:

• Agency mortgage-backed securities. Bonds guaranteed by Fannie Mae and Freddie Mac carry no corporate credit risk. The current-coupon agency MBS basis offered about 33 basis points more than US investment-grade option-adjusted spreads as of Nov. 5. That pickup was negative in 2021 and 2022 and has been mostly positive since 2023.
• Senior AAA non-QM mortgage bonds. Spreads on these tranches over agency MBS narrowed to near zero in 2024 and 2025 and have turned positive again this year, which PIMCO attributes mostly to extension risk rather than credit risk. New deals are structured to pay down senior holders faster. "We think the repricing in AAA non-QM creates an intriguing opportunity," Karoui said.
• Emerging-market debt. In a hypothetical portfolio exercise, adding EM external and local-currency debt to a mix of US aggregate, investment-grade and high-yield bonds improved risk-adjusted returns by up to 0.13 in Sharpe ratio terms. PIMCO singles out semi-liquid EM investment-grade deals, often used to fund infrastructure, which paid a median pickup of about 40 basis points over comparable bonds this year. "It is not that EM is a lower-risk asset class," Karoui told CNBC; the case is that EM debt offers a "relatively differentiated source of yield, duration, and country-specific risk."

PIMCO notes that its publication is the firm's own market view rather than independent research. With AI borrowers locked into multi-year buildout funding, their share of the long-dated corporate bond market appears set to stay near record levels through the coming issuance cycle.