The university that doesn’t exist yet

Here’s a scenario that should make university administrators uncomfortable: within 15 to 20 years, a large share of American universities could be competing with institutions that do not exist yet — and losing.

The scenario rests on one assumption: AI becomes genuinely good at teaching. Not just answering questions, but guiding students through structured learning, adapting to weaknesses, giving feedback, and helping people reach mastery. If that happens, the economics of higher education could change very quickly.

What nobody has built yet is the full package: a cheap, somewhat selective, large-scale undergraduate degree that employers would respect. In practical terms, that means a degree priced much closer to WGU than to a traditional university, selective enough to carry real signaling power, flexible enough for hard-working students to finish in two or three years if they want, and delivered through a much cheaper instructional model because it relies far more on AI than on large numbers of human instructors.

The market gap is surprisingly clear. WGU showed that low-cost, large-scale, flexible undergraduate education is possible, but not selective admissions. ASU showed that online education can operate at serious scale under a mainstream university brand, but still within a more conventional cost structure. Georgia Tech showed, at the master’s level, that a highly respected institution can offer a rigorous online degree at a fraction of the usual price. The pieces exist. The full combination does not.

If AI becomes good enough at teaching, that missing combination may become feasible for the first time. A model built around AI tutoring, explanation, practice, and feedback could cut the biggest cost that keeps many universities expensive: human instruction delivered at traditional staffing ratios. Humans would still matter, but in narrower roles and at leaner ratios.

How might such an institution work? It would be fully online, with students moving at their own pace. AI would handle much of the teaching for structured material. Human mentors would still matter, but mainly for accountability, motivation, career guidance, and judgment. The point would not be to eliminate people. It would be to use them where they add the most value.

The biggest obstacle is probably not pedagogy. It is accreditation and access to federal financial aid, which still assume a model built around semesters, credit hours, faculty roles, and institutional forms that this kind of provider would not neatly fit. That is a serious barrier. But it is a regulatory barrier, not a market one. Regulatory barriers can change.

If they do, the result would probably not be one institution but a new tier of higher education. Like the current system, it would likely stratify by selectivity. The more selective versions might compete with the tier below state flagships and attract ambitious students who want to finish faster than four years. Mid-selectivity versions could compete with regional publics and commuter schools. Open-access versions might look more like assessment platforms: learn however you want, then prove what you know through trusted evaluation.

Not every student would thrive in this model. Learning through AI without a physical campus would demand a level of self-discipline that many 19-year-olds do not have. But some students clearly could, including working adults already balancing jobs, families, and education. And for many of them, the attraction would be obvious: lower price, more flexibility, and a faster path to a credible degree.

Who gets hurt depends on a variable nobody talks about enough: how much of the value proposition is place-based.

A 300th-ranked commuter university would likely be more vulnerable than a 400th-ranked residential liberal arts college. The commuter school’s main offering — accessible education at a moderate price — is exactly what this new tier could try to deliver better and cheaper. The residential college is selling something much harder to replicate: four years alongside peers, face-to-face mentorship, and the social and developmental experience of living in a physical community. That is a real moat, even if not an unlimited one.

The first institutions under pressure would be predictable. Some for-profit and low-ranked commuter-serving schools are already struggling with enrollment declines and regulatory pressure, and a credible low-cost alternative would intensify both. The more interesting question is what happens one tier up. Regional publics and mid-ranked private colleges that primarily serve commuter populations could face real existential pressure, squeezed between demographic decline from above and a cheaper alternative from below.

Residential colleges with strong campus cultures would be better protected, though even they might feel pressure at the margin if the price gap widened enough. The top 100 schools would be less exposed still. They are selling network, brand, and a developmental experience that also functions as an elite signal. Students admitted to Duke are not making a simple price comparison against an online alternative.

The institutions that survive in the middle would probably not survive unchanged. As this new tier pulls away students, traditional schools would face a brutal choice: lower tuition, cut costs, or lose enrollment. Most would do some combination of all three. That could mean more AI-delivered instruction, fewer instructors, more consolidated departments, and a thinner version of the residential experience.

The faculty implications would follow, though unevenly. Contingent faculty would likely feel the pressure first. Over time, as departments shrink and merge, tenure lines could also be affected through attrition. The professor teaching upper-division seminars at a top-100 university is probably fine. The full-time lecturer teaching general education at a regional public that just lost a fifth of its enrollment is in a much weaker position.

Still, the whole scenario hinges on whether employers change how they hire.

That is the hardest part of the argument. The model rises or falls on a social fact, not a technological one: whether employers decide that this credential means something.

The real question is whether a hiring manager in 2035, looking at two résumés — one from a mid-tier university, one from a selective AI-native academy charging $8,000 — would treat them as broadly comparable signals. If the answer is yes for even a meaningful minority of employers, then a large part of the middle of higher education could face a genuine threat. If the answer is no, then this new tier remains niche and the current system limps along, shrinking but structurally intact.

My bet: it starts slow and then accelerates. The first employers to trust these credentials would likely be firms that already use stronger skills-based hiring systems. The last would be sectors that still care deeply about pedigree, familiarity, and conventional institutional signals.

The disruption to higher education is not that AI replaces universities. It is that AI may make possible a new kind of institution — cheaper, leaner, faster, and eventually credible enough — that can outcompete a large part of the market. That would be painful for many existing colleges precisely because, for many students, it could be a better deal.

That university does not fully exist yet.

But the logic for building it already does.