The surge in high-cost borrowing for artificial intelligence infrastructure has prompted federal officials to evaluate whether the trend could destabilize the US financial landscape. Members of both parties on Capitol Hill are monitoring AI investments for systemic risks to the wider financial system as the debt market for data centers shows emerging cracks. Investors are increasingly favoring stable, low-risk projects over speculative developments, which is raising borrowing costs and narrowing the pool of lenders available to syndicate multibillion-dollar loans.
AI borrowers issued around $55 billion in high-yield bonds this year, and debt for projects linked to Meta, Google, and Nvidia has spread into investment-grade and broad bond funds. Major lenders are reducing their exposure; in 2025, MUFG participated in 15 of the 25 largest data center deals and SMBC was in seven. These banks, along with Société Générale, provided financing for the Stargate and CoreWeave projects and a $15 billion bridge loan for SoftBank before narrowing their lending criteria.
Key sources
- SOURCE@theinformation“Financing the AI data center boom is getting more expensive as bond investors demand bigger concessions and some major banks become more selective”x.com
- SOURCE@theinformation“AI-related borrowers have sold around $55 billion of high-yield bonds this year”x.com
- SUPPORT@theinformation“Even broad bond funds now hold debt tied to projects connected to companies such as Nvidia, Google and Meta”x.com
- SUPPORT@edzitron“SMBC and MUFJ, who have been in effectively every large deal - Stargate, CoreWeave, etc - are now "stepping back"”x.com
- SUPPORT@edzitron“Out of 25 big data center deals done in 2025, MUFG/MUFJ was in 15 of them and SMBC was in 7”x.com
- SUPPORT@brendanpedersen“members of both parties say they're monitoring the potential for AI investment to pose systemic risks to the wider financial system”x.com
- SOURCEhuggingnewshuggingnews.com