Investors struggle to offload risk while looming loan maturities threaten weaker borrowers’ refinancing prospects amid higher interest rates
Debt markets supporting AI expansion are tightening, with yields on some data center and chip-related bonds and loans reaching double digits, according to various reports.
Bankers are struggling to transfer risk to insurers and other investors, Bloomberg reported Wednesday, according to CNBC. Citadel Securities has warned that exposure extends beyond speculative “neocloud” borrowers: major technology firms could face pressure because they are backing projects with their balance sheets.
All of the more-than-30 bankers, investment managers and private equity executives Bloomberg interviewed had said data center financing has become substantially harder to secure in recent months. Investors are concerned about provisions allowing customers to cancel when construction falls behind schedule. Oracle’s force-majeure notice last month concerning a New Mexico data center had likely heightened those worries.
To date, at least 18 data center transactions across junk bonds and leveraged loans now yield double digits. JPMorgan Chase and Goldman Sachs have deepened discounts on a US$5bn Volta Infrastructure Holdings loan, lifting its yield above 11%.
Also, credit default swaps, used to protect against losses, have become more expensive for Broadcom, Nvidia, Meta and Oracle, reaching levels associated with junk-rated borrowers. Average yields in Bloomberg’s investment-grade technology index have risen to roughly 6%. Citadel analyst Tucker Roberts said the volume of forthcoming debt could make finding buyers harder than securing investment-grade backing. Still, Citadel is not forecasting an immediate credit crunch.
Circular financial arrangements detected
Other investors are retreating. Oaktree’s Armen Panossian said spreads indicate AI lending markets may be approaching saturation. Bloomberg reported that KKR is avoiding remote data centers, while Carlyle has rejected several transactions.
Separately, on 27 October 2026, Moody’s had warned that AI funding is expanding private credit beyond conventional corporate lending, Reuters reported. Loans maturing from 2027 through 2029 will test weaker borrowers’ ability to refinance at higher interest rates.
Moody’s projects that capital expenditures by the five largest US hyperscalers and CoreWeave will approach US$1tn in 2027. Around 17% of its 607 tracked borrowers use payment-in-kind interest, postponing cash interest payments while increasing outstanding debt. The agency also highlighted “circular” financing arrangements through which technology firms, chipmakers and AI developers invest in, lend to and purchase from each other.
The Herald Business has noted that these interconnected exposures could deteriorate simultaneously during a downturn.
