AI Compute Financings: Evaluating Re-Leasing Risk for GPUs and TPUs

The growth of artificial intelligence (AI) is creating a new category of digital infrastructure financing. These transactions are not simply data center financings or traditional equipment loans; they sit at the intersection of the two. The data center provides power, cooling, fiber connectivity, security, and the physical operating environment, while graphic processing units (GPU), tensor processing units (TPU), networking equipment, storage, and the software stack deliver the compute capacity that ultimately drives revenue.

That distinction matters because data center and compute equipment depreciate on different timelines. A well-located data center with scarce power can remain valuable for decades. A GPU or TPU cluster may physically operate for many years, but its economic relevance can decline much faster as next-generation chips deliver more computing power per watt, support newer software libraries, or become the prevailing standard for AI workloads.

For project finance lenders…