**Research hubs vs teaching institutions diverge on different timelines (2025–2045).**

Research universities feel the funding squeeze first — falling federal overhead rates and compute replacing lab equipment as the binding capital constraint — but retain a moat: frontier AI work still needs physical labs, clinical cohorts, graduate pipelines, and trust relationships industry cannot cheaply replicate. Expect the top 50–80 research universities to consolidate grants, PhD students, and industry contracts and to grow in output even as aggregate federal dollars flatten.

Teaching-only institutions have no comparable moat. AI tutors plus employer-issued micro-credentials unbundle the lecture and the credential; a tuition-dependent model with no research revenue breaks within 10–15 years.

Regional pattern: flagship and land-grant systems in population-growing states (SEC footprint, Texas, Florida, North Carolina, Arizona) absorb both roles — applied research plus workforce pipelines — and expand. Small private teaching colleges in the Northeast and Midwest decline fastest, an accelerated continuation of the 2015–2025 closure wave, roughly 3x its prior rate.