Landlords, tenant quality, municipal pushback and electricity costs are increasingly factoring into underwriting strategies for data center hardware as equipment lenders look to mitigate risk.
Amid the data center boom, many lenders have financed GPUs and CPUs with little hesitation over the past year to capitalize on surging CapEx spending.
But now context matters more than ever, Riley Thompson, vice president and head of direct sales at Mitsubishi HC Capital America, told Equipment Finance News, a sister publication of FinAi News.
“These days, we want to know where you’re putting [the equipment] and what data center you’re putting them in,” he said. “How well is it capitalized? What are the other tenants that are in there? What is your kilowatt cost per hour?
“Those are all things that now go into the lending decision that didn’t a year ago.”
For example, Mitsubishi HC Capital America struck a $100 million deal with a data center tenant occupying roughly 15% of the facility and the landlord required the lender to pay it “way more money than we’re even making on the deal” in the event of a default and repossession, Thompson said.
Thompson said the landlord told the lender, “’What you guys need to understand is that because I have the water, the power, the land and municipal support, I have gold, so I can use whatever terms and conditions I want.’”
Many lenders have financed data center equipment and later realize that the power purchase agreement (PPA) wasn’t baked into the deal and they’re “left holding the bag” when a borrower defaults, Thompson said.
“You financed them, they have the GPUs, they’ve made a couple payments to you, but they literally don’t have a data center that they can put the equipment in,” he said
PPAs are long-term contracts in which an energy company agrees to supply a power source and the data center owner or developer agrees to buy the electricity at a set price.
It’s crucial to factor PPAs into data center hardware financing deals because electricity must be treated as a core part of the collateral, Thompson said.
Kilowatt costs increased roughly 23% over the past five years, according to the U.S. Energy Information Administration.
Municipal pushback
Pushback from municipalities also is increasingly weighing on data center financing, Thompson said.
There are 210 active data center moratoriums across 43 states, with most imposed by counties, cities and townships, according to technology investment firm Interconnected Capital.
And now state officials are putting up roadblocks. This week, Texas Gov. Greg Abbott ordered a pause on all new data center approvals and Pennsylvania Gov. Josh Shapiro imposed restrictions on permit approvals for data center projects.
Lenders are not only concerned about developments in progress but also data centers that are fully operational, Thompson said.
“Some of those are in danger of getting shut down by the municipal pushback that’s going on,” he said.
Reducing risk
While many financial institutions are understandably targeting the largest data center players with top-tier credit, lenders also should consider smaller players to diversify their portfolios and help maintain balanced market dynamics, Thompson said.
“We do not want the power of AI consolidated into three or four companies in China and three or four companies in the U.S.,” he said.
“I understand why the lending community is changing their underwriting criteria, but we really need to support the mid-sized, small-sized companies to make sure that we don’t have a much, much bigger problem on our hands.”
If this consolidation occurs, “we basically report to tech companies,” he added.
Ultimately, lenders must stay disciplined and understand that the largest deals aren’t always the best deals when financing data center equipment, Wintrust Specialty Finance President and Chief Executive David Normandin told EFN.
“It’s a very shiny object,” he said. “You can do a lot of volume very quickly, but it’s important to make sure that you don’t lose sight of what your mission is,” Normandin said.
“You can put your entire portfolio at risk if you don’t use discipline around the size of and scale the opportunities that really fit your business.”
Check out our exclusive industry data here.
Facts Only
* Equipment lenders are incorporating landlord quality, tenant quality, municipal pushback, and electricity costs into data center hardware underwriting.
* Mitsubishi HC Capital America and Wintrust Specialty Finance are active lenders in this sector.
* Mitsubishi HC Capital America executed a $100 million deal with a tenant occupying 15% of a facility.
* U.S. Energy Information Administration data shows kilowatt costs increased approximately 23% over the last five years.
* Interconnected Capital reports 210 active data center moratoriums across 43 U.S. states.
* Texas Governor Greg Abbott paused all new data center approvals.
* Pennsylvania Governor Josh Shapiro imposed restrictions on permit approvals for data center projects.
* Power Purchase Agreements (PPAs) are long-term contracts for electricity supply at set prices.
* Lenders are observing cases where borrowers possess GPUs but lack accessible data center space.
Executive Summary
Equipment lenders are shifting from aggressive, volume-driven financing of GPUs and CPUs toward more disciplined underwriting as the data center boom matures. The primary driver of this shift is a heightened focus on collateral risk; specifically, the realization that hardware value is dependent on the availability of power, land, and municipal support. Lenders are now scrutinizing Power Purchase Agreements (PPAs) and the financial stability of landlords to avoid "holding the bag" during borrower defaults.
The landscape is further complicated by increasing regulatory friction. With over 200 active moratoriums and recent executive pauses on approvals in Texas and Pennsylvania, the viability of both new and operational facilities is under pressure. While there is a natural tendency to gravitate toward top-tier credit players, some industry leaders argue for supporting mid-sized and small firms to prevent the consolidation of AI infrastructure power into a handful of global corporations. However, the overarching trend is a move toward rigorous discipline to prevent portfolio-wide risk.
Full Take
The strongest version of this narrative is a cautionary tale of "infrastructure lag." It posits that while the demand for compute (GPUs) is astronomical, the physical prerequisites—electricity, zoning, and municipal goodwill—are finite and fragile. The core argument is that hardware is useless without a "plug," making the utility contract the actual collateral, not the silicon.
This situation follows a classic speculative cycle: an initial gold rush characterized by low-friction lending, followed by a "correction" where the physical constraints of reality (power grids and local laws) impose a ceiling on growth. The root cause is a mismatch between the speed of digital scaling and the inertia of physical infrastructure.
There is a subtle tension regarding the democratization of AI. The push to fund smaller players is framed as a necessity to avoid a corporate oligopoly, yet this contradicts the primary theme of risk mitigation. Funding smaller players typically increases risk, suggesting that the "systemic risk" of consolidation may be viewed by some as more dangerous than the "financial risk" of a loan default.
The second-order consequence is a power shift toward the "landlords of the grid." As seen in the $100 million deal example, those who control the water and power hold absolute leverage over both the tenant and the lender, effectively extracting rents from the financing process itself.
Patterns detected: none
Bridge Questions:
1. If municipal pushback continues to mount, will data center financing migrate toward decentralized "edge" computing or shift geographically to regions with fewer regulatory hurdles?
2. Is the drive to support mid-sized AI firms a genuine strategy for competition, or a rhetorical shield to justify higher-risk lending?
Counterstrike Scan: A coordinated campaign to manipulate this narrative would use "fear of consolidation" to lobby for government subsidies for small-scale data centers. The current content does not match this pattern; it is a straightforward discussion of risk management among lenders.