US computer-science enrolment falls for first time in 20 years

US computer-science enrolment has fallen for the first time in roughly 20 years. During the 2025–26 academic year, the average college saw participation in computer-science classes decline 4.6% year on year.

The finding comes from Jacob Light, an economist at Stanford’s Hoover Institution, whose research was reported by Business Insider. His dataset covers more than 50 million course sections since 1996 across 1,019 US colleges.

National data points in the same direction. The National Student Clearinghouse reported that undergraduate computer-science enrolment at four-year colleges fell 8.1% in autumn 2025, from about 659,700 students to 606,100.

From degree boom to first decline

The drop follows a long expansion. National Science Foundation figures show that the number of computer-science bachelor’s degrees more than doubled in a decade, rising from about 56,000 in 2014 to 122,000 in 2024.

Recommended reading

AI learns ship navigation from experienced captains

The timing coincides with two major shifts: AI assistants have become mainstream tools for writing, debugging, and completing coding assignments, while the entry-level software job market has cooled.

Why AI is not the only explanation

Light does not claim that AI caused the decline. He treats the results as descriptive rather than causal and identifies five possible contributors: a weaker junior-developer job market, doubts about the value of a computer-science degree, weaker mathematics preparation, immigration-related demographic shifts, and the rise of data science.

The data cannot distinguish among those explanations. Computer science has also experienced boom-and-bust cycles before, and enrolment remains above its 2022 level.

Still, students appear to be responding to signals from the labour market. Companies such as Verizon are cutting staff, while JPMorgan’s boss has said AI has already eliminated a third of jobs in some units. Some demand is shifting toward other fields, including skilled trades.

That creates a potential contradiction: the discipline responsible for building AI could be losing some of the students who would develop its next generation. Google DeepMind’s chief argues that a STEM degree still makes people far more effective with AI, and Light agrees that many graduates will find substantial value in one.

Whether this is a temporary dip or the start of a broader rethink will become clearer as students compare a degree with the appeal of “vibe coding.”

Published July 22, 2026, at 11:31 a.m. UTC.