# The research function bifurcates: compute hubs vs teaching institutions

**Claim:** Over the next 20 years, the American university's research role contracts to a small set of well-capitalized hubs while most institutions revert to teaching-focused operation — and the split is regional.

**Mechanism:**
1. **Compute is the new gate.** AI-era research demands GPU infrastructure, data, and industry partnerships only elite-endowed institutions can afford. Federal research dollars were already concentrated in a few dozen institutions; compute economics concentrate further.
2. **Corporate siphon.** Frontier AI research is migrating out of universities into corporate labs (OpenAI, DeepMind, Microsoft, Meta). Universities keep the research that requires physical presence and regulatory access — wet labs, clinical trials, field science — where their comparative advantage holds.
3. **Mid-tier squeeze.** R2 institutions without industry ties get priced out of competitive AI research and revert to teaching colleges. The *effective number* of research universities shrinks even as hub budgets grow.

**Projected shape by ~2045:** ~30–50 research hubs with heavy industry co-funding; the long tail becomes AI-augmented teaching institutions — cheaper to run, smaller faculty, credential-focused.

**Regional pattern:** Hubs cluster where tech industry, state investment, and population growth align (Texas, Georgia, Florida, North Carolina, coastal California). Demographic-decline regions — small-town Northeast, rural Midwest — lose their research role first, hollowing out anchor employers in the process.