Setting limits on AI exposure: Carlyle's Jason Thomas on the risks for private credit

The private credit industry faces significant risk from over-concentration in the $5 trillion AI infrastructure build-out, potentially mirroring the 2005–2008 financial crisis. Jason Thomas, Carlyle’s head of global research and investment strategy, warns that private credit may provide $1 trillion of this funding, leading to a dangerous 40% concentration level. While current compute backlogs reach $2.5 trillion, the net economic value sits at only $375 billion due to high GPU and data center costs. This margin remains vulnerable to power shortages, political opposition, and the potential loss of capital market access for pre-profitability AI labs. Despite reliance on residual value guarantees from semiconductor manufacturers and hyperscalers, these credit supports remain untested in a workout scenario. Investors should set explicit concentration limits at approximately 15% and demand hefty premiums for incremental exposure to avoid the pitfalls recently experienced in the SaaS sector.