Computer science enrollment falls for the first time in 20 years as AI reshapes who gets hired

Undergraduate computer science enrollment fell for the first time in roughly two decades, with the University of California system reporting a 6% decline from 2024 and a 9% drop over two years, while national data from the National Student Clearinghouse shows an 8.1% decrease in computer and information science programs in the 2025-2026 academic year. Stanford economist Erik Brynjolfsson found that employment among software developers aged 22 to 25 has declined nearly 20% since ChatGPT launched, as entry-level job postings dropped 60% between 2022 and 2024 and 54% of companies have stopped hiring junior developers due to AI. The shift reflects students and employers moving away from traditional CS paths toward AI-specific roles, with UC San Diego's dedicated AI major seeing enrollment rise roughly 20%.

ChatGPT didn't just change how developers work , it's now convincing students not to become developers at all, and Stanford has the data to prove the job market is already responding. The boom was supposed to be permanent. After ChatGPT launched in late 2022, computer science felt like the safest bet in higher education: the one major that could pay off your student loans before you'd unpacked your dorm room. Enrollment in CS programs had climbed almost without interruption since the early 2000s. Then 2025 happened. For the first time in roughly two decades, undergraduate computer science enrollment fell across the University of California system, down 6% from 2024 and 9% over two years, according to data reported by TechSpot. The UC system isn't an outlier. Nationally, the National Student Clearinghouse found enrollment in computer and information science programs dropped 8.1% in the 2025-2026 academic year, the steepest decline of any field of study tracked. Enrollment in CS and programming majors each fell more than 10% that year. Across 62% of computing programs surveyed by the Computing Research Association, units reported falling enrollment. Only 13% saw growth. Students aren't walking away from technology. They're walking away from the version of it that used to guarantee a job. The students making this call aren't being irrational. Stanford economist Erik Brynjolfsson and his co-authors studied payroll data from ADP through mid-2025 and found something stark: employment among software developers aged 22 to 25 has declined nearly 20% since ChatGPT launched, as Fortune reported in June. Older workers in the same field are doing fine , employment for developers over 30 in high AI-exposure roles actually grew between 6% and 12% over the same period. It's specifically the entry tier that's collapsing. Brynjolfsson's explanation is direct. Large language models learn from written, codified knowledge , the textbooks, documentation, and tutorials that form the curriculum of a CS undergraduate. Senior engineers carry tacit knowledge: the institutional memory, the judgment calls, the pattern recognition that doesn't live in any repo or Stack Overflow thread. Junior developers, by definition, haven't built that yet. Their knowledge overlaps most precisely with what AI can replicate. They were, as Brynjolfsson's paper calls them, canaries in the coal mine. The hiring data makes the same point from the employer side. Entry-level developer job postings dropped 60% between 2022 and 2024, according to analysis covered by Built In. A survey found 54% of companies have stopped hiring junior developers because of AI, and junior roles now represent just 7% of new IT hires - down from 15% before the tools took hold. Big Tech cut new-grad hiring by 25% in 2024 compared to 2023; startups cut it by 11%. The economic logic for a seed-stage company is blunt. One senior engineer with Cursor or GitHub Copilot can produce output that would previously have required two or three juniors. Why carry the mentorship overhead? UC San Diego is one of the few bright spots in the enrollment data , and the reason is telling. It's the only UC campus that launched a dedicated AI major, and the only campus where enrollment went up, by roughly 20%. Students aren't fleeing technology. They're re-routing toward the parts of it that don't look replaceable yet. What this means for the next five years of startup hiring Here's the problem that's invisible in the current moment: the mid-level and senior developers that startups rely on today were once the junior developers that companies hired five and ten years ago. If that intake valve closes, the talent pool doesn't just get more expensive , it gets smaller. The pipeline breaks. Founders building teams right now should probably think about this in two phases. Phase one, the next two or three years, genuinely favors a leaner model: fewer engineers doing more with AI tooling, higher average seniority, and compensation concentrated at the top of the stack rather than spread across a large team. That's real. The math works. But phase two, the 2028-to-2032 window, is when the enrollment drop becomes a supply shock. Fewer students entering CS programs today means fewer experienced engineers available to hire in five years, precisely when AI-native applications will be scaling hardest and competition for senior talent will be fiercest. The irony is sharp. The technology that made coding look automated has convinced a generation of students not to learn it deeply , and that may produce exactly the developer shortage that makes the next generation of engineers worth more than ever. Don't be surprised if the CS enrollment rebound, when it comes, is as sudden as the drop. Also read: CuspAI raises $450 million to let AI design the next generation of chip materials https://startupfortune.com/cuspai-raises-450-million-to-let-ai-design-the-next-generation-of-chip-materials/ • Synthesia turns its AI avatars into performance coaches with Roleplay Sessions https://startupfortune.com/synthesia-turns-its-ai-avatars-into-performance-coaches-with-roleplay-sessions/ • Omen AI raises $31M to detect the bacterial threat quietly killing GPU clusters https://startupfortune.com/omen-ai-raises-31m-to-detect-the-bacterial-threat-quietly-killing-gpu-clusters/