# AI splits the degree market by major and region before it shrinks it by headcount — the bachelor's survives as a license to direct AI, but only in half the curriculum

**Core claim:** Over the next 20 years, AI is a **repricing event, not a demand-collapse event**. The generic bachelor's degree loses its wage premium wherever what it certified — routine cognitive work — is now done by AI. Where the degree legally or practically gates human practice — licensure fields, trust-bearing roles, AI-direction itself — demand holds or grows. The degree survives, but as a narrower product: a credential for supervising and integrating AI-supplied skills, priced closer to a certification than to a four-year bundle.

**The split is by major first, institution second.** Generalist white-collar fields — generic business, marketing, communications, entry-level analytics — see their premium compress toward zero by the early 2030s (Unverified: timing extrapolated from current premium data, BLS/CEW wage-premium series). These programs lose pricing power and enrollment fastest, concentrated at non-selective tuition-dependent schools. Licensure-anchored fields — nursing, engineering, teaching, accounting — hold or gain, because the state requires credentialed humans regardless of AI capability. National enrollment figures stay misleadingly stable through this: headline headcount masks a collapse in roughly half the typical curriculum.

**Institutional geography follows labor-market geography, not prestige alone.** Flagships and land-grants in growth corridors — the Southeast, Texas, Florida, and the broader Sun Belt, adjacent to expanding energy, health, and defense sectors — absorb demand and diversify program mix. Small privates and regional publics in the Northeast, Rust Belt, and slow-growth rural regions shrink, hit by a compounding of weak labor-market pull and the post-2025 college-age decline (CDC/NCHS birth data; WICHE projections). The sorting axis is regional economic growth and program mix, not conference affiliation or brand per se.

**The failure mode is diagnostic, not just financial.** Institutions that read program-level enrollment decline as price sensitivity will cut aid and budgets in exactly the programs AI is killing, accelerating their own hollowing. The correct read is signal-substitution: demand isn't leaving the degree, it's leaving the parts of the degree whose signal AI made free.

**What would change this:** (1) Entry-level hiring in generalist fields stabilizes with degree premiums intact through the early 2030s — the repricing thesis fails on its stated timeline; (2) Licensure fields also lose enrollment as AI-supplied skills substitute for credentialed practice — the "licensed half holds" claim fails; (3) Regional demand converges (Sun Belt flagships shrink alongside Northeast privates) — the geography claim fails;