This paper argues that universities must rethink how they provide AI computing resources as demand for GPU access continues to outpace supply. Researchers increasingly expect rapid access to AI infrastructure, yet traditional procurement and deployment processes often take months, while shared national resources remain heavily oversubscribed. At the same time, existing GPU resources are frequently underutilized due to fragmented, manually managed environments created for individual groups. The paper proposes operating campus AI infrastructure as a standardized, governed “factory” that transforms hardware into self-service AI environments accessible to researchers, students, and departments. It examines the funding, policy, and operational challenges driving this shift, outlines key factors

Join for free to read