Higher education leaders are weighing a costly computing arms race as AI workloads expand, often requiring high-performance computing far beyond what many campuses provide today. A report focused on “faster compute” describes how training large language models and running advanced analytics can more than quadruple energy needs—creating infrastructure and cost constraints for institutions. Steve Goodman, senior director of technology at Marquette University, said limited computing slows research and can degrade downstream learning quality; he cited a student-built teaching assistant chatbot that took more than two months to learn but produced inaccurate coding assessments. The example is used to illustrate the time-and-cost risk of insufficient compute for iterative, AI-enabled research and teaching. The article also links computing capacity to enrollment and curricular differentiation, noting prospective students and employers expect AI skills embedded in programs. It argues purely cloud-based models may not fit every use case due to continuous, large-scale workloads and the need for hands-on control. For campuses, the investment question is no longer limited to classroom tools. It spans campus power, HPC procurement strategies, data governance, and budgeting models that can handle sustained compute demand.