# US universities 2025–2040: thinning, not shrinking — the answer is "both, asymmetrically"

**Core answer.** Some grow, more shrink, and the system overall *thins* rather than collapses. Median tuition-dependent institutions lose 20–40% of enrollment by 2040 (Unverified: projection). Total enrollment drifts down from ~19M toward ~15–17M (Unverified), but institution count falls faster than headcount: closures, mergers, and branch cuts consolidate the map while displaced demand migrates to survivors. "Grow or shrink" dissolves into "which one, on which axis."

**Direction is set by demographics, not AI.** US births fell from ~4.3M (2007) to ~3.6M (2023), a ~18% contraction in the traditional-age pipeline; high-school graduates peak ~2025 at ~3.9M (WCHS), so 2025–2026 is the high-water mark, not a plateau (Unverified on exact figures; direction confirmed by CDC/NCHS). The enrollment contraction of 2029–2037 happens regardless of what AI does. AI answers only the second-order questions: who absorbs the lost demand, which programs retain pricing power, and how much costs fall.

**What AI actually does — sorter and accelerant on three axes.**

**Program axis (sharpest).** Generalist degrees lose their screening premium because employers can observe competency directly and no longer pay for the signal; licensure-gated programs (nursing, engineering, accounting) keep pricing power because the credential remains the license to practice. This cuts *within* every institution: each campus houses both program types, so institution-level enrollment flatters the real picture. Expect quiet program-level triage — generalist majors wither or reprice to near cost, licensure programs expand — before institution-level closures become visible.

**Cost axis (slowest).** AI targets non-tenured instruction first — the roughly 60% of instructional staff without tenure protection (Unverified: share). Tenure shields incumbents, so cost reduction comes only through hiring freezes and non-replacement, phased over 10–15 years. Budgets therefore lag technology by a decade; distress from cost-side AI savings surfaces late, and the enrollment shock from