# AI bends the instructional cost curve a decade before it shrinks the campus — personnel savings arrive by attrition, so cost-per-credential falls faster than headcount

The teaching stack — instructional faculty, adjuncts, TAs, advising, tutoring — is the largest share of university operating spend, and it is the part of the university AI substitutes for most directly. A public or non-elite private that leans into AI tutors, AI-graded formative assessment, and AI-assisted advising can plausibly cut 10–20% of total operating cost over a decade (Unverified on magnitude). But the savings cannot arrive as layoffs in the near term: tenure, shared governance, faculty-senate gates, and accreditation minimums on instruction force the adjustment through hiring freezes and attrition. So the cost curve bends slowly and predictably — a glide, not a cliff — which means cost-per-credential collapses years before enrollment does.

The competitive consequence splits by market position. Institutions selling a positional good (selective flagships, elite privates) will not pass savings through — admission scarcity prices independently of instructional cost, so their revenue-per-student holds and the efficiency gain lands as margin or reinvestment. Institutions selling a generalist credential will be forced to compete on price, because their product is becoming cheap to deliver everywhere; the AI-instructional cost cut is what makes a sub-$10K-degree lane viable at regionals and defines the floor under tuition competition (Unverified on the price point). The institutions that shrink are not the ones facing weak demand but the ones facing weak demand with a rigid cost stack — tenured-heavy, tuition-dependent, unable to reprice downward or reposition upward.

For the grow-or-shrink question, cost structure is the third leg alongside demographics and credential value: falling instructional cost expands delivery capacity (cheap seats) even as raw headcount contracts, so the system can deliver more credentials for less money with fewer people — "shrinking" in dollars and personnel while not shrinking at all in delivered capacity.

What