What Banks Should Examine in a Data Center Deal

A data center leased to a major technology company can look like a reassuring lending opportunity. Then come the questions about power, construction, local opposition, and how much computing capacity that company will need years from now. 

Today’s massive, single-tenant AI facilities are a different proposition from older data centers, many of which served multiple tenants and could be converted to other uses. The new facilities can require their own substations, substantial power and water infrastructure, and specialized construction expertise. 

In a ProSight Banking Strategies Podcast conversation, Edwin Anderson, a partner in Oliver Wyman’s banking and financial services practice, compared the challenge to early toll-road financing. Some lenders focused on expected cash flow without fully understanding the technical assumptions behind it. He sees similar concerns in data center lending, warning that “people that are just following the pack may inadvertently be getting the worst of the loans.” 

Anderson’s discussion highlights several areas worth examining: 

Study the infrastructure requirements. Reliable power is essential, and securing it can be complicated. Banks need to understand grid capacity, legal constraints on power access, and the availability of specialized equipment and labor. Shortages of substation components, electrical equipment, and skilled workers can disrupt construction. 

Read the lease carefully. A familiar technology company may be associated with the project, but the actual tenant could be a subsidiary. Anderson said parent-company guarantees may be limited, expire, or require renewal. Lease termination provisions also deserve scrutiny. Those details can significantly affect the protection a lender receives. 

Look beyond government approval. A project may have the necessary permits and still face substantial community opposition. Anderson cited concerns about water use, noise, pollution, and limited permanent employment. Lawsuits and moratoriums can derail projects, particularly when opposition emerges after substantial construction spending. 

Consider how AI demand might change. Today’s data center expansion reflects a particular approach to developing AI, one that requires enormous computing resources. Anderson noted that alternative models, changes in hardware, or customer preferences for less expensive AI products could alter demand. He cautioned that “there’s a huge correlated risk here on the kind of AI bet being made.” 

Know where the exposure ends up. Some lenders originate construction loans and sell them, while others retain longer-term exposure. “The risk sits with the person who ends up holding it,” Anderson said. He also urged institutions to consider how much of their lending depends on the same assumptions about AI’s future. 

The takeaway: Data center lending offers substantial opportunities, but these projects demand specialized underwriting expertise. Careful scrutiny of infrastructure, contracts, construction, and long-term demand can help banks understand the commitments they are making—and the risks they may be keeping.