Where Data Center Risk Can Hide in the Portfolio

A data center may arrive as one giant project, but the credit effects can spread well beyond the site itself. 

The boom can change local labor costs, land values, infrastructure needs, housing demand, municipal decisions, and borrower behavior. That means exposure may show up even when the bank is not financing the data center itself. 

That was the core credit question in a ProSight Risk Readiness webinar on the data center boom: Which borrowers are changing because of the boom, and would their finances still hold up if today’s conditions do not last? 

The panelists offered several places for lenders to start: 

Follow the constraint. Construction and commercial real estate are obvious areas to watch when a data center project changes demand for labor, land, or infrastructure. But financial services executive Subramanian Narayanaswamy said lenders should look at borrower characteristics, too. Thin margins, fixed-price contracts, limited pricing power, lease renewals, and competition for scarce labor or land may matter as much as industry classification. His shorthand: “follow the constraint rather than just the industry.” 

Look beyond the headline contract. Brian Strawberry, chief economist with FMI Corporation, said data center work can create concentration risk for construction firms. For example, a contractor may pull crews from long-term customers or turn down recurring smaller jobs to chase the large project. Those customer relationships may not appear on the balance sheet, but they have economic value. Narayanaswamy said lenders should look past the attractiveness of the new contract and ask what the borrower’s business will look like after it ends. 

Stress test for timing, duration, and magnitude. Narayanaswamy suggested asking what happens if the expected economic lift from a data center project arrives 12 or 18 months late, lasts two years instead of five, or is only half as large as anticipated. Those scenarios should be translated into familiar credit variables such as revenue, margins, rents, collateral values, leverage, and liquidity. “The data center doesn’t have to fail for the assumptions surrounding it to fail,” he said. 

Do not underwrite to the peak. Strawberry said the construction phase may create a short-term labor swell. Ermengarde Jabir of Moody’s added an important caution: Once built, an operational data center “creates very few jobs.” That distinction matters because contractors and other borrowers serving the boom may make longer-term commitments—hiring permanent employees at higher wages, taking on multi-year equipment leases, or adding debt. 

Map indirect exposure. Narayanaswamy offered three questions: Who is connected to the development? How is it changing the economics of the business? Which businesses are changing financial behavior by hiring, buying equipment, expanding capacity, or taking on more debt? Together, those questions help lenders see when exposure to the boom has become a dependency—and when a borrower’s risk profile may have changed. 

The takeaway: Data center exposure reaches well beyond direct lending into those projects. Lenders shoudl ask whether the boom is changing borrower behavior—through hiring, equipment purchases, added debt, or dependence on a large new customer. Borrowers that look materially better than they did two years ago deserve a closer look before stronger numbers hide a new dependency.