Carlyle urges private credit to cap AI financing exposure

what happened

Carlyle Group is calling on private credit lenders to set clear, advance limits on artificial intelligence financing. Seeking Alpha reported that Jason Thomas, Carlyle’s head of global research and investment strategy, said a potential 40% concentration in AI data centers would create excessive risk for the asset class. He warned that lenders could repeat the overexposure mistakes seen with software-as-a-service companies earlier this year, after Anthropic's Cloud Code was released in February. Depending on the definition used, SaaS exposure had reached 20% to 40% of portfolios.

Thomas put the projected AI infrastructure buildout at $5T over the coming years, with private credit expected to provide approximately $1T. Based on current assets under management and growth expectations, that would imply roughly 40% concentration. He also estimated that the $2.5T backlog of compute remaining performance obligations represented only about $375B in net economic value after discounting and comparing it with the costs of GPU purchases, data center construction and power infrastructure. The resulting 15–17% present-value margin, he said, could narrow rapidly.

Thomas compared current lending assumptions with the 2005–2008 period, saying lenders now rely heavily on residual value guarantees from hyperscalers and semiconductor manufacturers. Deal documentation varies, while workouts remain untested. He recommended a 15% cap on private credit exposure to AI-related assets, with limits set before risk materializes rather than afterward. Thomas said having no exposure would be unreasonable, but additional risk above that threshold should carry hefty premiums.

why it matters

The warning highlights concentration and underwriting risks for private credit as AI infrastructure financing grows. Investors will be watching guarantee structures, project economics and the ability of pre-profitability AI labs to meet lease commitments.

instruments