AI Lending Surge Boosts Bank Activity Amid Caution
U.S. banks are experiencing a notable increase in lending to AI-related sectors, with commercial and industrial loan commitments rising from approximately $150 billion in early 2022 to about $191 billion by late 2025, according to the Chicago Federal Reserve. This surge is driven by growing investments in AI infrastructure, including data centers and semiconductors.
Moody's reports that the five largest U.S. hyperscalers and CoreWeave are expected to spend nearly $1 trillion on AI infrastructure in 2027, financed through various channels, including debt. However, this expansion is accompanied by concerns over refinancing risks, as many loans are set to mature between 2027 and 2029.
The San Francisco Federal Reserve highlights that banks with higher AI adoption have shown a 0.38 percentage point increase in average return on assets but also face slightly higher problem-loan shares. These banks tend to have lower shares of SME loans, holding 12% compared to 21% in banks with lower AI adoption.
Regulatory bodies, including the Office of the Comptroller of the Currency and the Federal Reserve, are increasing their scrutiny of AI applications in high-risk areas such as lending and sanctions screening. This regulatory attention comes as banks continue to navigate the complexities of AI integration into their operations.
In a significant financing move, a consortium of banks, including Bank of America, Citi, Morgan Stanley, and Blackstone, announced a $63 billion transaction to lease AI compute capacity to Anthropic, leveraging Broadcom technology, with deliveries expected in 2027.