Carlyle Warns AI Infrastructure Spending Risks Lender Limits

Jason Thomas of Carlyle warned that the massive capital spending required for AI infrastructure could create significant financing risks by 2027. Projections suggest a need for $4 trillion in external finance by 2030, much of which will rely on debt through special purpose vehicles and off-balance-sheet borrowing.

Thomas cautioned that this volume of spending may breach the prudential risk concentration limits of lenders, who are being asked to allocate more of their portfolios to AI than is typically comfortable. He noted that the economic value of these projects remains highly sensitive to political opposition to data center construction, turbine procurement for natural gas solutions, and delays in grid interconnection.

These factors could narrow margins and complicate the ability of hyperscalers to meet compute rental agreements. Additionally, Thomas highlighted the Federal Reserve System's desire to tighten financial conditions to guide inflation down toward its target, adding further pressure to the financing landscape.