Artificial Intelligence Investments and Bank Loan Contracting

Abstract

Measuring firm-level artificial intelligence (AI) investments based on employee resumes, we find robust evidence that firms’ investments in AI are associated with significantly higher interest rates of bank loans. To mitigate omitted variable concerns, we use the initiation of the ImageNet contest as a quasi-natural experiment. The impact of AI investments is stronger when banks lack expertise in lending to AI firms and for firms with more information uncertainty, suggesting that information frictions contribute to the higher cost of debt. Firms’ investments in AI are also accompanied by a more prevalent use of contingencies in debt contracts (i.e., more covenants and performance pricing provisions). Lastly, AI investments are associated with a contraction in debt issuing activities. Overall, our study highlights the credit market frictions in promoting transformative technologies.   

Keywords: Artificial intelligence, Bank lending, Cost of debt, Bank expertise, Debt covenants