Marquette University’s technology leadership is putting a spotlight on a new constraint for higher ed: high-performance computing demand is rising faster than many campus IT infrastructures can support. In a report on whether “faster compute” is becoming an arms race, administrators warn that training large language models and running advanced analytics require dramatically more power and cooling than standard institutional clusters. Steve Goodman, Marquette’s senior director of technology, describes how limited computing capacity can directly impair academic work—citing a student-built teaching assistant chatbot that spent more than two months training before researchers found inaccurate assessments. He frames the risk as more than technical: slower iteration can delay research outcomes, including in medicine and engineering. The article also links compute constraints to recruitment and curriculum changes. Goodman argues that AI fluency and access to AI-enabled labs are quickly becoming “differentiators” for prospective students, and that cloud-only approaches may not fit institutions that need continuous, hands-on workloads. For university CIOs and research leaders, the operational issue is investment sequencing—how to fund upgrades now while scaling energy, hardware, and staffing capacity to keep research and instruction competitive.