Global AI-related debt issuance fell sharply in September as investors grew increasingly cautious over rising leverage, infrastructure costs and uncertainty surrounding returns on investment, according to a report by the Financial Times.
Data compiled by Morgan Stanley shows that AI-linked debt issuance fell to $23bn in September, less than half the amount raised in August and well below the record $113bn issued in June. The figures cover public bond markets and private placements.
The slowdown follows a surge in borrowing to finance AI infrastructure, including data centres and advanced semiconductors. Companies have raised an estimated $466bn in AI-related debt so far in 2026, compared with $101bn over the same period last year.
In the US investment-grade market, AI-related bond issuance stalled in September after major technology companies raised approximately $306bn between January and August.
Morgan Stanley attributed the decline primarily to the volume of financing completed earlier in the year, although concerns over investment returns and growing opposition to data centre developments have also weighed on sentiment.
The increase in debt supply has given credit investors greater negotiating power over yields, raising borrowing costs for technology companies and infrastructure developers.
Speaking at the FT and Latham & Watkins Private Capital Summit in London, John Aylward, founder of Sona Asset Management, said financing conditions were becoming more challenging as lenders approached their exposure limits.
The change in sentiment could affect several large financing transactions currently under development. Wall Street banks are arranging a $60bn financing package for Broadcom and Anthropic, while SpaceX is reportedly seeking $40bn to purchase Nvidia chips. Both transactions are expected to be syndicated to investors in the coming months.
Concerns have also emerged around the infrastructure underpinning AI expansion, particularly access to electricity and local opposition to data centre construction.
Oracle’s Project Jupiter data centre development in New Mexico has faced difficulties securing power, placing approximately $18bn of associated loans under pressure. The loans were recently quoted privately at around 85 cents on the dollar, according to the FT.
Jordan Rieger, managing principal and head of European credit at Monarch Alternative Capital, said the slowdown could provide an opportunity for lenders to reassess the structure and risks of AI-related financing facilities following months of heavy issuance.
Rob Dafforn, chief investment officer at distressed debt specialist Polus Capital, also questioned whether investors were adequately pricing the risks associated with highly leveraged AI transactions, particularly given uncertainty over development timelines and eventual returns.
Polus has so far avoided making investments in the sector, although it has begun assessing existing AI-related financing transactions for potential opportunities.
The slowdown highlights growing scrutiny of the financing structures supporting AI infrastructure expansion, with credit investors increasingly focused on leverage, execution risk and the ability of projects to generate sufficient returns to service their debt.
