Higher Education Is About to Find Out What It Actually Sells

American universities, at their current price points, are selling four different products bundled together. The first is knowledge: access to faculty, libraries, courses, and the intellectual infrastructure for learning. The second is credential: the degree that signals to employers that you completed a recognized program of study. The third is network: proximity to peers who will become professionals, connections to alumni, access to recruiters who show up specifically because of where you study. The fourth is status: the social fact of having attended a particular institution, which confers prestige independently of what you learned or who you met.

These four products have very different cost structures, very different substitutes, and very different competitive positions relative to AI. The crisis facing higher education is that AI threatens one of them severely, competes partially with two others, and leaves the fourth untouched, but the institutions have never been transparent about which part of their product people are actually paying for.

Knowledge access is the most thoroughly disrupted. This was already happening before AI. Wikipedia, YouTube lectures, online textbooks, and JSTOR reduced the marginal cost of accessing information to near zero for anyone with internet access. AI completes the disruption. An undergraduate today with access to GPT-4 or Claude can get explanations of complex concepts more patient and tailored than most office hours, research assistance more thorough than most campus libraries, feedback on writing more immediate than most professors provide. The knowledge access value proposition of a $75,000-per-year university education is increasingly difficult to defend on its merits.

Credential is more complicated. A degree from a recognized institution still functions as a labor market signal, but the signal is eroding in several directions simultaneously. Employers who have learned to distrust grades, because AI-assisted coursework makes grades an unreliable signal of genuine competence, are substituting their own assessments for degree-based screening. Google, Apple, and IBM formally removed four-year degree requirements from most positions in the early 2020s, citing the availability of alternative competency signals. This is still the exception rather than the rule in most sectors, but the direction is clear. If the credential’s value as a competence signal degrades, the price universities can charge for it degrades with it.

The network product is more defensible. The people you meet at a good university, particularly the people you work with in competitive environments, who reveal their actual capabilities under pressure, are not something AI replicates. The friendships formed in shared adversity, the professional relationships built before anyone had status to protect, the exposure to people from different backgrounds who challenge your assumptions. These are real, valuable, and not available elsewhere. LinkedIn is not a substitute. An AI cannot introduce you to the person who will be your co-founder, your most valuable professional reference, or your most important intellectual sparring partner.

The status product is also defended, for reasons that are almost purely sociological. An MIT degree does not just certify competence. It signals that you survived a selection process that many people attempted and few passed. This signal is valuable independent of what it says about any specific skill set. Employers use it as a quality filter not because they believe MIT graduates can code better than graduates of every other institution, but because MIT’s selective admissions process does much of the filtering work for them. AI does not disrupt this mechanism. Harvard’s admissions selectivity will continue to confer status even if Harvard’s knowledge transmission is fully substituted by AI tools.

The crisis is that the institutions bundling these four products together are charging prices anchored on the knowledge access component, or at least on the cost of maintaining the infrastructure that delivers it, while more and more of their actual perceived value comes from the credential and status components. This mismatch creates a vulnerability. If the knowledge access component can be provided for near zero marginal cost, and if the credential and status components are the ones people are actually paying for, then the university is overcharging for knowledge access while the real product, credential and status, has a different competitive structure.

The universities that are genuinely at risk are the ones in the middle of the status hierarchy. Top-20 research universities sell enough status and network value that demand is essentially inelastic. Community colleges and regional public universities serve students for whom credential access and price are the primary considerations. The institutions at risk are the ones in between — moderately selective private universities that are expensive enough to price out cost-sensitive students but not prestigious enough to capture the status premium that makes elite institutions price-inelastic.

These institutions are already struggling. Between 2020 and 2026, more than 200 small private colleges either closed or merged. AI accelerates the pressure, particularly on their knowledge access argument. If their primary selling point has always been “you can get a quality education here at lower cost than the elite institutions,” and if AI makes quality education increasingly available outside any institution, the argument becomes harder to make.

The universities that will navigate this well have already started articulating what they actually sell. MIT’s OpenCourseWare project, which put all MIT course materials online for free in 2001, was an early and explicit statement: “we are not primarily selling knowledge access, we are selling something you can only get by being here.” MIT built the argument that presence mattered. Their subsequent investments, in maker culture, in research access, in the network of alumni and industry partners who recruit specifically at MIT, are all consistent with that argument.

The institutions that are still primarily arguing “come here because we have better faculty and libraries and instruction” are in the most precarious position. Those arguments were weakening before AI and are much weaker now. It’s not that good faculty and libraries and instruction don’t matter — they do. It’s that the marginal value of a moderately good professor over a good AI assistant is not $50,000 per year, and students, or their parents, are starting to do that math.

There is a version of higher education reform that responds to AI productively: a deliberate shift toward making the network and experience dimensions of university life more central, with AI handling more of the knowledge transmission work. This would look like smaller seminars, more project-based learning, more emphasis on collaborative work and peer engagement, more investment in the parts of the campus experience that require physical presence and human relationship. Less lecture, more dialogue. Less homework, more apprenticeship.

This costs about the same as the current model and is educationally better, which is why almost no institution has moved decisively in this direction. The barriers are structural: faculty who are rewarded for research, not teaching; class sizes driven by cost pressure rather than pedagogical design; accreditation systems that count credit hours and course completions rather than experiential outcomes.

The students who are making decisions about higher education in 2026 are the first generation for whom the AI alternative to traditional coursework is genuinely capable. The enrollment data is beginning to reflect this: applications to moderately selective institutions were down 8% nationally between 2024 and 2026, while applications to highly selective institutions were up 3%. The gap is not about demographics. It’s about the value proposition. When AI reduces the knowledge access premium of attending a university, the residual premium goes to the institutions whose credential and network are most valuable. The institutions in the middle are discovering, in real time, that their value proposition rested more on knowledge access than they thought.

The international dimension is worth noting specifically. American universities built a $47 billion annual international student tuition market over the past 30 years, largely on the logic that an American degree gave foreign students access to American academic culture, American professional networks, and a credential recognizable by American employers. AI is eroding the first of these, you can access American academic content without flying to America, leaving the second, professional networks require physical presence and remain valuable, and the third, the credential’s labor market value depends heavily on the US labor market context.

The Chinese middle class, which produced an enormous fraction of international student tuition revenue for American universities through the 2010s, is now asking different questions about the value of an American degree. The combination of US-China geopolitical tension, the rising quality of Chinese domestic universities, and AI-enabled access to high-quality educational content regardless of geography is reducing the calculus that used to make spending $200,000 on an American degree a straightforward family investment decision. Several American universities that built financial models around international tuition, New York University’s global campuses were a notable example, are quietly renegotiating those models.

There’s a scenario where American universities emerge from the AI disruption in a stronger position than they entered it: if the network and status products hold their value, and if the institutions that invested in genuine intellectual culture, the ones that produce graduates who can think clearly about complex problems, find that this distinguishes their credential from those of institutions that sold primarily content access. This is not impossible. It requires the institutions to be honest about what they actually provide, invest in the things that make presence valuable, and stop defending assessment structures that AI has already broken.

The institutions that get this right will be the ones that stopped trying to be all things to all students and became very good at something specific: developing the kind of judgment that comes from real intellectual challenge in a community of peers. That’s a smaller market than “education for everyone.” It’s a more honest product than what most universities are currently selling. Whether the scale of the contraction required to get there is survivable for the middle of the institutional distribution is genuinely unclear.

The unbundling process has started and will not reverse. Students who are primarily seeking knowledge access have alternatives that didn’t exist five years ago and are much better than they were two years ago. Students who are primarily seeking credential signals have a growing range of alternatives, professional certifications, portfolio-based hiring, employer-designed assessment, competing with the traditional degree. The universities that survive, and the ones that thrive, will be the ones that can clearly articulate what their bundle provides that the unbundled alternatives don’t. Presence, relationship, intellectual culture, network. These are real. They’re worth paying for. But they need to be worth the price being charged, and right now, for too many institutions in the middle of the status hierarchy, the honest answer is that they aren’t.