Employment of developers aged 22 to 25 has declined by approximately 20% since 2024, according to the Stanford AI Index 2026. Not because of a recession. Not because of a trade war. Because of a deliberate recruitment choice that dozens of tech companies have made quietly, often without measuring the consequences over a ten-year horizon.

The dynamic is simple to understand, difficult to correct. AI agents can now absorb a significant share of the tasks that made up the first years of a junior developer: writing basic code, documenting, testing, integrating. Companies saw this and drew immediate conclusions: fewer juniors hired, more agents supervised by seniors. Short-term balance sheets are flattering. Payroll decreases for entry-level profiles. Team productivity improves. And seniors — those who know how to pilot these agents, correct their errors, arbitrate their limitations — see their salaries rise because their relative scarcity increases.

What these balance sheets don’t show is what happens in ten years when these seniors retire and no mid-level cohort has been trained to replace them.

The Essentials

  • Approximately 128,270 total tech layoffs were recorded in the first five months of 2026, according to HeroHunt.ai — only a portion of which were directly attributed to AI
  • Employment of developers aged 22 to 25 has declined by approximately 20% since 2024 (Stanford AI Index 2026)
  • Salaries for senior developers and AI architects are rising, driven by a scarcity of available experienced profiles
  • The missing cohort today will produce a mid-level deficit in 2033-2035, with a scissors effect on salaries and potentially on innovation capacity
  • Several companies are beginning to create hybrid training pathways involving AI agent supervision to anticipate this shortage, but at still marginal scale

128,270 Layoffs in Five Months

The raw figure is considerable. In the first five months of 2026, HeroHunt.ai records approximately 128,270 job cuts in the tech sector, across all causes — not only those explicitly linked to AI. The same source specifies that only about 9,200 of these layoffs were directly attributed to AI. Nevertheless, these are not merely cyclical cutbacks — the sector remains profitable, valuations hold. They are largely strategic reorganizations, choices about human allocation around new work architectures.

What the aggregate figure masks is its distribution by profile. Cuts don’t affect teams uniformly. They fall selectively on entry-level positions and the most codifiable functions: QA, level-1 technical support, standardized frontend development, documentation management. In other words: functions historically occupied by junior profiles early in their careers.

Data from the Stanford AI Index 2026 clarifies this distribution. Employment of 22-25 year-olds in the tech sector has declined by approximately 20% in two years. This is a sharp drop over a very short horizon, without equivalent in previous automation cycles. For comparison, the decline in clerical employment linked to office automation in the 1990s took a decade to produce structural effects of this magnitude. The current adjustment is two to three times faster.

Seniors Benefit from a Relative Shortage They Didn’t Create

While junior job openings contract, salaries for senior profiles and AI architects accelerate. The mechanism is classical: when demand remains strong and supply becomes scarce, price rises. Except the scarcity isn’t natural — it’s the direct product of hiring decisions made since 2023.

Companies that reduced their junior cohorts inadvertently reduced the pipeline for their own future seniors. A mid-level developer in 2030 is a junior from 2025 who will have had four or five years to learn on the job, accumulate real errors, understand complex architectures through practice. If this person wasn’t hired in 2025 because an AI agent did the same thing at marginal cost, they simply won’t exist in the available talent pool in 2030.

Senior salaries rise therefore for two simultaneous reasons. First, their operational value is real: piloting AI agents, detecting their hallucinations, arbitrating decisions beyond their scope, supervising complex architectures requires expertise that years form, not prompts. Second, their relative supply falls because the junior cohorts meant to replace them have been sacrificed to short-term optimization.

This isn’t a collapse of the tech labor market, as some feared during early layoff waves. It’s a bifurcation. Work is polarizing between highly qualified, rare, and well-paid profiles, and AI agents that absorb standardized tasks, leaving a void where learning once happened.

A Generational Gap Opens in Real Time

The 20% decline figure for 22-25 year-olds must be read with its temporal dynamic. If this pace holds for two more years — a conservative hypothesis given hiring intentions stated in sector surveys — the entry cohort will have fallen 35 to 40% compared to 2022 levels. The effect in ten years calculates quite simply.

A 35-year-old senior developer in 2035 must have started working around 2022-2025, during the period of maximum compression in junior hiring. The size of this talent pool will be structurally reduced. If today’s seniors, those aged 40-50, begin retiring in 2032-2038, the market will find itself in scissors: strong demand, reduced pool, salaries erupting. This is the scenario several HR prospecting consultancies already anticipate under the name “skills cliff” — a cliff rather than a slope.

This mechanism has precedents in other sectors. France’s civilian nuclear industry experienced something structurally comparable in the 1990s-2010s: a generation of engineers recruited en masse during the reactor construction program, followed by a long hiring pause, then acute shortage when staff had to be trained for maintenance and the large-scale refurbishment. EDF had to recreate emergency training pipelines at high cost. The parallel isn’t perfect — the tech sector renews faster — but the pipeline logic is identical.

The question facing the United States therefore is this: who pays the cost of training that’s no longer being done by the market? Historically, it was the company that trained its juniors by paying them less than their immediate productivity, recovering the difference during their mid-level years. This implicit contract is breaking.

What Pioneer Companies Are Doing to Correct Course

The market isn’t blind. Some companies sensed the problem before aggregate data quantified it clearly, and began corrections.

The model emerging in a growing number of large tech companies is what’s called “AI-augmented apprenticeship”: a hybrid entry pathway where the junior isn’t recruited to write basic code, but to supervise AI agents on tasks of increasing complexity, progressively developing capacity to identify their limits and intervene. The training logic is inverted: you learn to correct before learning to build, which raises serious pedagogical questions.

Other companies, notably in scale-ups of fewer than 500 people where HR budgets are tighter, are experimenting with partnerships with bootcamps and universities to co-finance cohorts of “AI oversight engineers” — a still-vague title that actually covers the junior developer profession reimagined for the age of agents. These partnerships remain small-scale: several thousand positions total, vastly insufficient to compensate for the decline of more than 128,000 tech jobs recorded in five months.

Perspective must be maintained. These initiatives are real, documented, and interesting as signals of market self-regulation. They don’t compensate, in the short term, for the scope of compression. But they prove something important: the future shortage is already readable by those looking ten years out, and several private actors are beginning to invest to avoid it. That’s not nothing.

Why Work Reorganization, More than AI Itself, Produces This Result

It’s tempting to read this situation as an inevitable consequence of technology. That would be analytical error. AI agents don’t eliminate juniors because they can. They eliminate them because companies chose to deploy them that way, in a context of cost pressure and valuation of immediate productivity gains.

The most convincing demonstration of this point comes from a sectoral comparison. In fintech, where regulation imposes minimum human-supervisor ratios and auditors require traceable decision trails, junior hiring has declined much less markedly — significantly but well below what’s observed in unregulated segments. Not because AI agents are less capable there, but because the institutional framework imposes a different work organization.

The same logic applies to technology consulting firms working for clients in regulated sectors: defense, health, critical infrastructure. These firms have maintained junior hiring at levels close to 2022, precisely because their clients impose human supervision constraints that agents alone cannot satisfy.

What these cases show is that the generational gap isn’t a technological destiny. It’s the result of an organizational choice in an unregulated market, where short-term optimization doesn’t internalize long-term costs. When external constraint forces internalization of these costs — regulatory, contractual, or cultural — hiring behavior changes.

What the Market Can Do Alone, and What It Won’t

It’s reasonable to anticipate that the market will partially self-correct. Senior salaries will continue rising, making investment in junior pathways increasingly profitable for companies that can read ten years ahead. Hybrid apprenticeship initiatives will multiply as shortage becomes more visible. A handful of large companies — those with balance sheets and time horizons to absorb training costs — have interest in recreating these pipelines to secure their own future needs.

But spontaneous market correction will be neither quick nor universal. It will arrive too late for the 2025-2027 cohort, already absent from the pipeline. It will favor large companies able to invest in long pathways, at the expense of small structures that will find themselves buying seniors trained elsewhere at premium prices. It will probably reproduce access inequalities already existing in the sector: partner bootcamps will recruit from the same pools as prestigious engineering schools, profiles from less-valued educational backgrounds will remain outside new hybrid pathways.

The political question this movement poses, and which almost no one yet poses publicly, is that of bearing the cost of training a generation. If companies externalize this cost — by no longer recruiting the juniors learning within them — someone else will have to bear it. Universities? States via apprenticeship programs? Individuals themselves, through expensive training whose return on investment is increasingly uncertain for less well-positioned profiles?

The market has begun posing the question. It remains to see who grasps the answer.

Sources

  1. HeroHunt.ai — Tech layoffs and AI: the 2026 reality check : https://www.herohunt.ai/blog/tech-layoffs-and-ai-the-2026-reality-check/
  2. Stanford HAI — Artificial Intelligence Index Report 2026 : https://aiindex.stanford.edu
  3. Stanford HAI — 2026 AI Index Report (official Economy page) : https://hai.stanford.edu/ai-index/2026-ai-index-report/economy
  4. Stanford HAI — 12 Takeaways from 2026 AI Index : https://hai.stanford.edu/news/inside-the-ai-index-12-takeaways-from-the-2026-report
  5. Burning Glass Technologies — Labor market data, tech hiring trends 2024-2026 (report without stable public URL)
  6. Burning Glass Institute — No Country for Young Grads : https://www.burningglassinstitute.org/research/no-country-for-young-grads
  7. Microsoft — AI skills and workforce transformation, public statements 2025-2026
  8. Salesforce — Trailhead AI apprenticeship program, public announcements 2025
  9. Salesforce — AI Builder Emerging Talent Program : https://careers.salesforce.com/en/jobs/jr341276/ai-builder-emerging-talent/
  10. SignalFire — State of Talent Report 2026 : https://www.signalfire.com/blog/signalfire-state-of-talent-report-2026
  11. TechCrunch — Running list major tech layoffs 2026 citing AI : https://techcrunch.com/2026/07/06/the-running-list-major-tech-layoffs-in-2026-where-employers-cited-ai/