What If the AI Industry Is Just the Next Higher Education?
From the AI is the New Higher Education collection
Most attempts to understand the AI industry reach for industrial metaphors. AI is the new electricity. AI is the new oil. These framings describe what AI powers — appliances, workflows, entire sectors — but they say almost nothing about how AI produces its core product. Electricity does not have a campus. Oil does not have a curriculum. Neither has an admissions office, a faculty, a graduation ceremony, or a credential.
The university does. And the AI lab does, too.
A frontier AI lab is an institution whose core output is trained intelligence. It recruits researchers the way a university recruits faculty. It selects and curates data the way a university selects students. It designs a training recipe the way a university designs a curriculum. It evaluates its models the way a university administers exams. It releases a model into the labor market the way a university sends a graduate into a first job. The model's knowledge cutoff is its graduation year, its model card is its transcript, its benchmark score is its exam grade, and its API price is its wage.
The structural parallel runs deeper than analogy. It is the same industry — the production, certification, pricing, and deployment of trained intelligence — operating at a new level. Higher education trains minds. The AI industry manufactures them. The name for this level is highest education, and the eight-century record of higher education is the closest thing available to a map of where it is headed.
The map is only useful if the reader knows where on it the AI industry currently stands. The method that makes the comparison precise is phase-matching: aligning the AI industry with the developmental phase of higher education that shares its structural conditions, rather than with higher education as it exists today.
Higher education's history follows a recognizable sequence. A long era of small, elite, teaching-focused colleges gives way to a research revolution in which the fusion of teaching and research creates a new kind of institution. A founding wave follows, as the new model is copied, adapted, and spread across countries within a few decades. Then come standardization, massification, the hardening of tiers and rankings, and finally commoditization and crisis.
The AI industry sits in the founding wave — the phase that ran from roughly the 1870s to the 1920s in higher education. Frontier labs are multiplying, converging on one production paradigm, and competing for the talent and capital that will decide which institutions survive. Comparing today's AI labs with today's universities — mature, bureaucratic, embedded in a mass system — produces false conclusions. The later phases of higher education's history are not descriptions of where AI is; they are forecasts of where it is going.
Within the founding wave, three structural mechanisms borrowed from higher education explain much of the AI industry's current behavior.
The first is the prestige flywheel. Education is a market in which buyers cannot easily inspect the product before purchase. A prospective student cannot test the quality of a degree before enrolling; an enterprise buyer cannot fully evaluate a model's capability before committing. In both industries, reputation fills the gap. Reputation attracts the strongest researchers, who produce the work that builds reputation, which attracts more talent and more users, which funds the next round of investment. The loop is self-reinforcing, and it is the mechanism that sorts institutions into a hierarchy.
In higher education, the flywheel explains why the same handful of universities have sat at the top of the prestige hierarchy for a century or more. In AI, the flywheel is already visible: the labs that attract the strongest researchers produce the most capable models, which attract the largest user bases, which generate the revenue that funds the next training run. The question the history poses is whether AI's hierarchy will freeze the way higher education's did — and if so, what accumulating assets will lock it in place.
The second mechanism is Bowen's law and the arms race it produces. The economist Howard Bowen observed that universities raise all the money they can and spend all they raise, not because they are wasteful but because prestige is a positional good — what matters is rank relative to competitors, and there is always another investment that could improve position. Frontier AI labs follow the same rule. Every funding round is spent, not saved. Compute budgets rise with each model generation. The spending is not irrational; it is the rational behavior of institutions competing for position in a market where position is the product.
The arms race extends beyond spending. Universities compete through research facilities, star faculty hires, and campus investments that no individual student demands. Labs compete through ever-larger training runs, infrastructure buildouts, and researcher compensation packages that no individual customer demands. In both cases, the race is for position in the hierarchy, and position is what customers are actually buying when they choose one institution over another.
The third mechanism is the deepest structural break in the entire parallel: who captures the graduate's wage. A human graduate keeps the salary premium that an education creates. The university trains the student, certifies the student, and sends the student into the labor market — but the graduate's paycheck belongs to the graduate. In highest education, the model graduate's wage flows back to the institution that trained it. The lab sets the API price, collects the revenue, and keeps the margin. The AI lab is a university fused with a staffing agency, an institutional form that has no precedent in the history of education.
This single difference reshapes every other parallel. It explains why AI labs can fund training from future revenue rather than from tuition. It explains why they can scale without limit — a model graduate can be hired a million times at once, while a human graduate works one job at a time. And it explains why elite AI labs cannot rely on scarcity the way elite universities do. A university that admits two thousand students a year sells exclusivity as part of its product. A lab whose model serves millions of concurrent users must sell capability and trust instead, because scarcity is precisely what it does not have.
The founding wave is also the phase in which an industry's permanent units are fixed. In a few decades, the research university settled the structures that have lasted a century: the academic major, the credit hour, the PhD, the department, the semester. These units were not inevitable. They were choices made under competitive pressure during a narrow window, and they hardened into standards because the institutions that adopted them survived and the ones that did not were absorbed or abandoned.
The AI industry is inside that window now. The units being settled today — model generations, token-based pricing, context windows, benchmark suites, release cadences — are the forms most likely to become permanent. The history of higher education suggests that whatever structures emerge from the founding wave will persist far longer than the people designing them expect.
The same history offers a warning. Every founding wave in higher education produced more institutions than survived. The survivors were not always the first movers or the biggest spenders. They were the institutions that built durable research cultures, maintained leadership continuity, secured funding that outlasted any single patron, and earned public trust. The AI industry's founding wave will end the same way — not with a ranking of who spent the most, but with a sorting of who built the institutions that lasted.
The value of reading the AI industry through higher education's record is not that the parallel is perfect. It is not. The breaks — copyability, wage capture, speed, the absence of graduate agency — are as informative as the matches. The value is that higher education provides a developmental map drawn over eight centuries, covering phases the AI industry has not yet entered: massification, designed tier systems, credential inflation, cost crises, the long negotiation between institutional autonomy and public accountability.
These phases are coming. The founding wave always ends. What follows — who builds the tier system, how credentials harden, where the cost crisis lands, whether the hierarchy freezes or stays fluid — is already written in the record of the industry that did all of this first. The AI industry is the newest education industry on earth, and the oldest one has already mapped its road.