NextFin News - The most lucrative technological boom in a generation is unfolding at the same time as the weakest job market for college graduates since the pandemic. In the second quarter of 2026, the unemployment rate for recent college graduates held at about 5.6 percent while the share working in jobs that do not require a degree edged up to 42 percent, according to the Federal Reserve Bank of New York's college labor-market tracker. The paradox is the story: artificial intelligence is demanding new skills, reshaping white-collar work, and lifting corporate productivity — yet for the class of 2026 and the cohorts just behind it, the bottom rung of the career ladder has never looked thinner.
For more than three decades, new graduates have typically fared better than the broader workforce. That historical rule has broken. The overall unemployment rate stood at 4.3 percent in August 2026, but workers ages 22 to 27 with only a recent bachelor's degree are jobless at a rate roughly 30 percent higher than the national average — and well above the 3.6 percent recorded in March 2019, before the pandemic rewired the labor market. More than four in ten employed recent graduates are in roles that do not call for their degrees, the highest level since 2020.
This is not a simple tale of machines taking graduate jobs. The evidence points to something more subtle and, for young workers, more damaging: the entry-level job itself is being hollowed out. Employers are hiring fewer trainees, automating the routine tasks that used to teach juniors their craft, and leaning on remote work to recruit experienced staff from cheaper labor markets instead of training local recruits. The result is an experience gap that no amount of prompting skill can close.
The Numbers: A Generation Entering a Cooling Market
The New York Fed's data, which stretches back to 1990, shows how far conditions have deteriorated. Recent-graduate unemployment peaked above 7 percent in 2011 in the aftermath of the Great Recession, then fell steadily through the late 2010s to 3.9 percent in December 2019. Underemployment — the share of graduates in jobs that typically do not require a degree — climbed to nearly 50 percent after the financial crisis, a level not seen since the early 1990s, before easing to 41 percent at the end of 2019. The 2026 readings of 5.6 percent unemployment and 42 percent underemployment show a labor market that has not simply failed to recover; it has regressed toward its post-crisis weak point even as the headline economy appears stable.
The contrast with older graduates is stark. In August 2026, the unemployment rate for workers 25 and older with a bachelor's degree or higher was 2.7 percent, according to the Bureau of Labor Statistics — barely above its pre-pandemic lows. For workers the same age with only a high school diploma, the rate was 4.4 percent. Recent graduates sit awkwardly between the two: more expensive than experienced degree holders, less proven than mid-career workers, and competing for a shrinking pool of trainee positions. Youth unemployment tells the same story from a wider angle: the jobless rate for workers ages 20 to 24 reached 7.1 percent in August, compared with 4.1 percent for the overall labor force.
The wage data reveal a market that has repriced faster than expectations. Undergraduates expect to earn $80,004 on average one year after graduation; actual starting salaries average $56,153, a gap of nearly 30 percent between the classroom's pricing and the market's. There is one countervailing statistic, and it matters. More than three in four recent graduates — 77.2 percent — landed a role within three months of graduating, up from 63.3 percent a year earlier, according to ZipRecruiter's Annual Grad Report. Employers still want graduates. What they want less of, the data suggest, is the kind of job that launches a career.
Why the Bottom Rung Is Missing
The first-order explanation most observers reach for is generative AI. Since the release of ChatGPT, the information sector — the industry most closely watched for signs of AI-induced displacement — has shed 8.5 percent of its employment since 2023, according to the Economic Policy Institute. The share of full-time job postings mentioning AI has nearly doubled in a year to 4.2 percent, and 35 percent of entry-level jobs now require AI skills, according to Handshake and the National Association of Colleges and Employers. Most pointedly, employment of workers ages 22 to 25 in AI-exposed occupations now sits 19 percent below where it would be had it tracked their less-exposed peers, up from a 15 percent shortfall a year earlier, according to Stanford University's Digital Economy Lab, which analyzed administrative payroll records covering more than 26 million U.S. workers.
But correlation is not causation, and here the mechanism deserves scrutiny. The Stanford authors are explicit that their finding is descriptive, not causal. AI does not primarily eliminate graduate jobs through layoffs of existing workers. It eliminates the tasks that used to constitute the apprenticeship: drafting first-pass code, summarizing documents, building baseline financial models, answering routine customer queries. When those tasks become cheap, the business case for hiring a junior employee to do them weakens — even as demand for the senior worker who supervises the AI rises. The displacement is not of people; it is of the training ground.
The entry-level job is disappearing in its own right, independent of AI. Postings that welcome candidates with no prior experience have fallen roughly 73 percent over four years and now account for about one in 50 salaried openings, down from about one in 15 in 2022, according to data from the hiring platform Cadient. This trend spans healthcare, government, utilities, and transit — industries where generative AI is not the obvious driver. Employers are increasingly filtering for demonstrated experience before a candidate ever reaches a human recruiter, so the first job that used to require little more than a degree and a willingness to learn now requires proof of work that only a previous job can provide.
This is why the AI story and the weak-pay story coincide without AI being the whole cause. A graduate today faces a double compression. On the demand side, fewer entry-level tasks justify fewer entry-level hires. On the supply side, the skills that command a premium have shifted faster than curricula can adapt, so graduates arrive priced for a job market that no longer exists. The $80,004 expectation is not greed; it is the price signal from a 2021-2022 hiring frenzy that has already reversed.
Remote work is a second, independent structural pressure — and on the New York Fed's own evidence, a larger one than AI. New York Fed researchers estimate that remote work can explain 64 percent of the increase in unemployment among young college graduates, and that the uptick in youth unemployment rates predates the rapid diffusion of AI. When a firm in San Francisco or New York can hire an experienced analyst in a lower-cost city, the incentive to recruit and train a local recent graduate falls. Remote work expands the experienced-worker labor pool while shrinking the local trainee pipeline. The economists put the concern plainly:
The high unemployment rates of young college graduates are particularly concerning because early-career experiences can have lasting consequences.
Cyclical Downturn or Structural Shift?
Getting this classification right determines the conclusion. Is the class of 2026 unlucky — caught in a cyclical cooling that will revert when hiring picks up — or is it the first cohort of a new regime in which the entry-level rung never returns?
The cyclical case is strong, and it is backed by the institution that tracks these graduates most closely. New York Fed researchers, examining real-time job-postings data, concluded that it is "difficult to attribute the recent slowdown in entry-level hiring to AI alone," noting that labor demand for junior and senior roles within highly AI-exposed occupations is moving "broadly in parallel." The Economic Policy Institute finds that the industries where young college graduates actually work are not experiencing more weakness than other industries. Job gains are as strong, if not stronger, and hiring has not fallen as far elsewhere. On this reading, the dominant cause is a depressed hires rate — a broad, cyclical slowing in labor market churn — not a profound structural change from AI or anything else. When the hires rate recovers, the argument goes, recent graduates will recover with it, as they did after 2011.
History offers support for mean reversion. New York Fed research following the Great Recession concluded that underemployment is a temporary phase for many recent graduates as they transition to better jobs after spending time in the labor market. The underemployment rate for recent graduates has fluctuated between 40 and 50 percent for more than three decades — a floor that predates the smartphone, let alone generative AI. On that evidence, today's 42 percent is not an outlier; it is the normal state of a new graduate's first years, and the current pain is a cyclical overlay.
Timing, however, complicates the pure-AI narrative. Harvard economist David Deming has pointed out that the decline in junior hiring began roughly six months before ChatGPT was released — a sequence that does not fit a story in which generative AI is the originating cause. The 73 percent collapse in no-experience postings and the 19 percent shortfall in young employment in AI-exposed roles are real, but they sit on top of a softening that started earlier and runs broader than the technology.
The cleanest judgment separates the forces rather than blending them. The level of underemployment is largely structural and cyclical-resistant — it has been the baseline for thirty years and will likely remain near 40 percent for the class of 2026. The deterioration from the 2019 trough is a mix of three distinct channels: a cyclical depressed hires rate; a structural remote-work channel that accounts for the majority of the unemployment rise and predates AI; and a smaller but career-shaping AI channel that concentrates in the routine cognitive tasks that used to train juniors. The structural piece is smaller in aggregate employment terms but larger in career-formation terms, because it attacks the specific jobs that convert a graduate into an experienced worker.
The Second-Order Consequence: An Experience Gap
The first-order story — fewer entry-level jobs, weaker pay — is already widely understood. The second-order consequence is not. If fewer graduates get trained, the supply of experienced workers tightens three to five years out, even in fields where AI raises productivity. Firms that cut trainee hiring today are borrowing against their own future talent pipeline. The displacement is visible in the cohorts: early-career professionals ages 22 to 25 in AI-exposed roles — particularly software development and customer service — have seen employment fall by roughly a fifth from recent peaks, while older, more experienced workers in the same fields have remained stable or grown.
The market has not priced this lag. Equity valuations reward the near-term margin expansion from AI-driven headcount discipline, particularly in information and professional services. What they do not price is the cohort scar: a generation that enters the workforce underemployed is less likely to catch up later, and the wage penalty of a bad first job can persist for a decade. The New York Fed's warning about "lasting consequences" is not rhetoric; it is the empirical finding of labor economics on cohort effects.
There is also an expectation gap embedded in the education market itself. Forty percent of students say they have considered changing their field of study because of AI, and 10 percent of those have already switched, according to a quarterly survey of workers and employers. That is a rational response to a shifting price signal — but it is a response to yesterday's signal. By the time a cohort pivots en masse toward AI-safe fields, those fields have often become the next crowded trade. The students who expected $80,004 are not wrong about the value of skills; they are wrong about the timing of the market's reward.
Employers themselves are divided on what comes next. While some senior talent leaders are nearly three times as likely to expect AI to increase rather than decrease entry-level hiring, according to the Strada Institute, other surveys show entry-level openings falling 35 percent in a single year. That divergence is itself the signal: nobody knows yet whether AI is a net creator or destroyer of first jobs, but the transition cost is being paid entirely by the youngest workers.
The Strongest Counter-Thesis
The most serious challenge to the structural-ladder thesis comes from the Federal Reserve Bank of New York itself. Its researchers found it "difficult to attribute the recent slowdown in entry-level hiring to AI alone" and observed that demand for junior and senior roles in highly AI-exposed occupations is moving "broadly in parallel" — exactly the pattern you would expect if the weakness is cyclical rather than technology-driven. A separate New York Fed study attributes 64 percent of the rise in young college graduate unemployment to remote work, noting that the uptick predates the rapid diffusion of AI. Harvard's David Deming adds the timing problem: junior hiring began declining about six months before ChatGPT's release. The Economic Policy Institute completes the case: the industries where young graduates work are not uniquely weak, and there is no profound structural change in the industry composition of employment that easily explains the softening.
Taken together, this evidence correctly defeats the crude claim that AI has already depopulated graduate occupations. Aggregate employment in graduate-heavy industries does not yet show an AI-shaped hole. But the counter-thesis underweights two mechanisms. First, task-level displacement precedes employment-level displacement; a firm can automate 30 percent of a junior role's tasks long before it eliminates the position, and during that window the employment data look benign while the training content evaporates. Second, remote work is itself a structural change, and it operates through a different channel than industry composition — it changes who competes for each job, not which industries are growing. A cyclical verdict on AI does not erase a structural verdict on remote work.
The falsifying signal is observable. If AI-exposed occupations — information services, software development, customer support, paralegal work, junior analysis — show entry-level employment growth re-accelerating relative to non-exposed occupations while the overall hires rate recovers, then the "AI removed the bottom rung" thesis is wrong and the cyclical view wins. Conversely, if entry-level counts keep falling even as mid-career roles in the same occupations expand, the structural-ladder interpretation holds. Watch the BLS JOLTS hires rate for the 22-to-27 cohort and the information-sector payrolls report over the next two quarters.
What Comes Next
In the short term, the class of 2026 and the class of 2027 will continue to face a market that pays less and trains less than the one their older siblings entered. The National Association of Colleges and Employers projects employer hiring for the class of 2026 to be up 5.6 percent from the prior year, with the information sector, engineering services, wholesale trade, construction, and professional services leading the increases. That is a cyclical tailwind, and it should modestly improve placement rates. It will not, by itself, restore the apprenticeship content of those roles.
In the medium term, the key variable is the hires rate, not the unemployment rate. A recovery in labor market churn would absorb underemployed graduates into better-matched roles, as it did after 2014. The New York Fed's quarterly update due in November 2026 will show whether the 5.6 percent unemployment and 42 percent underemployment readings are a plateau or a trough. The Stanford Digital Economy Lab's next data vintage will show whether the 19 percent shortfall for young workers in AI-exposed occupations is widening or stabilizing.
In the long term, the structural question resolves around one variable: whether the routine cognitive tasks that used to train juniors stay automated. If they do, the entry-level job will permanently reconfigure around supervision, judgment, and client-facing work — and the graduates who thrive will be those who arrive with those skills already demonstrated, not those who expected to learn them on the job. The underemployment floor of 40 to 50 percent may prove to be not a post-recession anomaly but the permanent cost of a degree in an economy that no longer needs beginners to do its beginner work.
The AI era is not taking the college degree away. It is taking away the first job that used to prove the degree's value — and that is a harder problem to hire your way out of.
Explore more exclusive insights at nextfin.ai.
