DoubleLine portfolio manager Robert Cohen is sounding the alarm on artificial intelligence debt, predicting it will almost certainly reach bubble levels in credit markets. His warning draws on a straightforward historical pattern: periods of massive capital deployment into transformative technologies tend to end with overextended balance sheets and painful corrections.

Cohen points to two well-known precedents — railroad expansion and the internet buildout — as templates for what happens when investment outruns fundamentals. Both eras generated enormous credit creation, funded ambitious infrastructure, and eventually left lenders and investors holding the bag when reality caught up to expectations. AI appears to be following the same trajectory, with debt financing accelerating as companies race to build capacity.

For credit investors, Cohen's assessment carries real weight. DoubleLine manages hundreds of billions in fixed-income assets, and its portfolio managers spend their days evaluating exactly the kind of risk that bubbles produce. If AI-related debt follows historical patterns, the credit cycle will eventually expose which borrowers overleveraged and which projects generated actual returns. Cohen's message is blunt: the bubble isn't a question of if, but when.