Entry-level software engineering postings in the United States dropped 67% between 2023 and 2024. That number comes from Stanford's Digital Economy Lab, working with ADP payroll data covering millions of workers. It is not a survey. It is not a projection. It is a body count.
The share of tech jobs requiring three years of experience or less fell from 43% in 2018 to 28% in 2024, according to IEEE Spectrum. Google and Meta are hiring roughly 50% fewer new graduates compared to 2021. Salesforce halted junior hiring entirely for 2025. In the UK, entry-level technology roles dropped 46% last year, with projections hitting 53% by the end of 2026.
Computer science graduates now face a 6.1% unemployment rate. Computer engineers are at 7.5% — one of the highest rates across all majors. The pipeline that turns students into engineers is collapsing in real time.
What the Stanford Data Actually Shows
Erik Brynjolfsson's team at the Stanford Digital Economy Lab published the clearest evidence yet in August 2025. Using ADP payroll records — not LinkedIn postings, not surveys, but actual employer data — they tracked what happened to workers in AI-exposed occupations from 2021 through mid-2025.
For software developers aged 22 to 25, employment declined nearly 20% from its peak in late 2022. That is a one-in-five reduction in the youngest cohort of working programmers.
Workers aged 30 and over in the same high-exposure fields saw employment grow 6% to 12% over the same period. AI is not eliminating software engineering. It is eliminating the entry point into software engineering.
The mechanism is straightforward. When companies adopt generative AI tools, junior employment drops 9% to 10% within six quarters. Senior employment barely moves. AI handles the tasks that used to justify hiring someone who needed training: boilerplate code, documentation, routine bug fixes, test scaffolding. The grunt work was the classroom. The classroom is gone.
The Seed Corn Problem
Heather Doshay, head of people at SignalFire, told the New York Times: "Nobody has patience or time for hand-holding in this new environment, where a lot of the work can be done by A.I. autonomously."
She is describing what agronomists call eating your seed corn. If you consume the grain you need to plant next season, you get one good meal and a permanent famine.
Senior engineers exist because they were once junior engineers who shipped bad code, broke production databases, wrote functions that didn't scale, and learned. That process takes five to ten years. There is no shortcut. AI can write syntactically correct code, but it cannot teach someone to reason about distributed systems under load, to recognize when an architecture won't survive its second year, or to debug a failure that crosses six services and three time zones.
If companies stop hiring juniors now, the senior pool starts shrinking in 2030. By 2033, every company that cut its junior pipeline in 2024 will be competing for the same diminishing set of experienced engineers.
The Inversion Nobody Expected
Meanwhile, the salaries tell a story the layoff headlines don't. Netflix, Anthropic, and OpenAI are paying up to $775,000 for non-coding roles — technical writers, prompt engineers, AI safety researchers, people who can think clearly about systems they don't personally build. The premium isn't on writing code anymore. It is on understanding what code should do and whether it did it correctly.
Morgan Stanley's research group argues AI will ultimately create more software engineering positions than it destroys. That may be true in aggregate. But the new positions require skills that take years to develop — the same years that companies are no longer willing to fund.
The industry is simultaneously saying AI will create a golden age of software and refusing to train anyone to participate in it.
What Actually Happens Next
The realistic near-term looks like this. Companies offshore junior work to cheaper markets. Senior salaries in the US and Europe climb because the supply is fixed and demand is growing. Mid-career engineers become bottlenecks, burning out under workloads that used to be distributed across larger teams. Startups that need to move fast hit a wall because they cannot hire seniors and refuse to train juniors.
Forrester predicts that half of all AI-motivated layoffs will be quietly reversed — rehired at lower pay, offshore, or repackaged under new titles. The work still needs doing. The humans who do it just get paid less and receive no training budget.
The 1.6% increase in Class of 2026 hiring that employers project is noise. It does not offset a 67% collapse. It does not rebuild a pipeline. It does not solve the problem of a generation of CS graduates who are credentialed but unemployable because nobody will give them their first production commit.
The Real Question
Five years from now, AI will still need humans who understand what it cannot: why a system was built a certain way, what trade-offs were made, where the edge cases live that no training data covers. Those humans come from somewhere. Right now, that somewhere is a shrinking doorway that companies keep making narrower while wondering why nobody senior is walking through it.
The industry is building its future on a workforce it is actively refusing to create.
Sources: Stanford Digital Economy Lab/ADP payroll data (Brynjolfsson, Chandara, Chen 2025), IEEE Spectrum, TIME, Fortune, CNBC, CIO, Stack Overflow, Rest of World, Renewed Right
Top comments (0)