I am going to stand by the title. The American university is built like a Ponzi scheme in the narrow, technical sense, which is that the promises owed to earlier cohorts can only be paid if each new cohort is larger, or pays more, or borrows more than the one before, forever. That dependency was invisible for as long as it kept working. It is visible now because it has stopped.
Walk through what a university actually owes. Tenure promised to faculty hired forty years ago. Pension obligations that run past 2060. Maintenance and replacement on a physical plant sized for peak enrollment. Bond payments left over from the construction binges of the growth years. Salaries for an administrative headcount that has roughly doubled in real terms since 1990. None of these shrink when enrollment shrinks, and all of them rise with inflation. The money to meet them comes from tuition, from federal loan guarantees, from state appropriations in some places, and from endowment draws at the small handful of rich schools. Tuition is the elastic part, and tuition is the part that has done all the stretching. Between the mid-1980s and 2020, sticker tuition at American four-year schools rose at something like five times the pace of median wages, and tuition net of aid at around three times. That gap between the posted price and the real price is what you always see in a system where the sticker exists for signalling and the actual number gets negotiated in private. The new money went to cover the old promises. New students paid for old commitments.
The federal student loan program is the outside capital that kept the whole arrangement upright far longer than it would have stood on its own. Once a student could borrow whatever the school chose to charge, the school could charge more or less anything, because the price had been cut loose from the student's ability to pay. The debt then sat on the graduate's balance sheet rather than the school's, could not be discharged in bankruptcy, and stayed collectible for life. That is the piece that looks most like a fraud mechanic when you describe it coldly, even though it was assembled out of good intentions one statute at a time.
The part nobody prices in until late is what happens when the incoming cohort finally shrinks. The American high-school-age population peaked around 2025 and is now declining, and projections put undergraduate enrollment somewhere between four and fifteen percent lower in 2040 than it was in 2020. Over the same stretch the earnings premium for a bachelor's degree has flattened or slipped once you control for who was going to do well anyway. And the credential is, for the first time in seventy years, facing real competition, from apprenticeships, from portfolio-based hiring in tech, from AI-assisted skills assessments that cost a few hundred dollars instead of a few hundred thousand. So the new cohort is smaller, poorer on the expected lifetime payoff, and holding alternatives it never had before. That particular combination is exactly the one under which this kind of structure tips over.
The top ten or twenty schools are fine, because they long ago stopped selling education and started openly selling membership in a class, and class membership does not face any of these pressures. The bottom two-thirds of the sector is already tipping. Small private colleges have been closing at roughly a dozen a year for most of this decade. Regional state schools are shutting humanities departments and calling it restructuring, which is what you call it when you cannot meet the obligations and cannot say so out loud. Community colleges are mostly holding on, because their costs actually track their revenue. The damage is landing hardest on the middle, which is also where most of the actually enrolled students are.
The campus panic about AI cheating is a symptom of the same disease. If the real product were learning, AI would be treated as the pedagogical earthquake it is and the teaching would be rebuilt around it inside two years. A few individual professors are doing exactly that. The institution as a whole is not, because its real product is the credential, and the credential is legitimized by assessment, and assessment depends on a polite fiction that students are not using the tools that can simply do the assessment for them. Letting AI honestly into the room would mean rebuilding the whole assessment regime, which would mean rebuilding the teaching, which takes an institutional energy that does not exist in a body whose attention is already consumed by servicing its legacy debts.
The staffing tells the same story. Tenure-track faculty are a shrinking slice of who actually teaches. Most of the teaching is done by adjuncts paid close to minimum wage once you divide it out by the hours. Administrative headcount keeps climbing, because administration is the machinery through which the legacy obligations get administered, and a typical research university now carries more middle management than a tech company of the same size, most of it employed to comply with rules that other middle management wrote.
What the collapse looks like in practice is slow, because large institutions rarely collapse, they grind. Small colleges close one at a time, announced in a regional paper and absorbed into the noise. Mid-tier state schools shed programs on a timetable that tracks their bond covenants. Expensive buildings get mothballed or sold off, often to a hospital system. The endowed schools at the very top stay insulated more or less indefinitely, because their money does not come from tuition at the margin. The middle hollows out. This is already underway and has another fifteen or twenty years left to run. Nobody had to do anything wrong for any of it to happen. The structure was a reasonable response to postwar demand and to a loan policy built on the assumption that demand would keep climbing, and most of the people running it genuinely believed they were stewards of a teaching institution, and many still do. The dynamics only ever needed a set of promises and a stream of revenue that held together as long as the revenue grew faster than the promises. When the revenue breaks, the promises do not care why.
If you are a high-school student weighing this right now, the thing to understand is that the system pricing your decision is feeding you bad information on purpose. The sticker is a signal, the real price is a negotiation, the value of the credential is slipping at the margin, and the default path is the one optimized for the institution's survival rather than yours. It is worth it when it is free or close to it. It is worth it when you need the specific credential for a licensed profession. Outside those cases the maths is much weaker than the brochure suggests, the books in the library are still there, and most of them are on the internet for less. The thing that is actually in trouble is the expensive institutional envelope wrapped around them.