# AI Pushes the Research Mission Toward ~50–100 Physical-Infrastructure Hubs; Everyone Else Becomes Teaching-First (2025–2045)

Paragraphs:

The mechanism. AI collapses the cost of desk science — literature synthesis, simulation, data analysis become commodity inputs available to any faculty member at any institution. What stays scarce is what AI cannot substitute: wet labs, patient cohorts, beamlines, national-lab adjacency, protected compute at scale. Federal and philanthropic research dollars therefore concentrate further on institutions holding these physical complements, and the marginal PhD-granting program without them loses its grant pipeline. NSF HERD data already shows federal academic R&D concentrated in a few dozen institutions; AI widens that concentration because it removes the labor-cost advantage mid-tier programs once had.

The two-tier map and regional expression. Expect ~50–100 infrastructure hubs — academic medical centers, big-engineering flagships, land-grants with DoE/DoD lab adjacency — that keep or grow their research mission and function partly as compute-and-cohort utilities (Unverified on the count). Growth-state systems (Texas, Florida, SEC-country land-grants, Sun Belt systems buying GPUs) convert state money into hub status fast, extending the filed pattern that geography rotates south. The long tail of mid-tier PhD programs converts into