The numbers behind the slide
Connecticut has about 3,000 fewer students taking computer science courses than it did in the 2022-23 school year, the third straight year of decline in a state that recently passed a law mandating computer science in its public school curriculum. John Tusch, president of the Connecticut chapter of the Computer Science Teachers Association, put it bluntly: 'I think the opportunities are drying up,' adding, 'I think we're fumbling the ball.' He also noted the shift in how students now talk about AI tools: 'People began to realize that it is a tool.'
The state-level data tracks a national pattern. Spring 2026 figures showed a decrease of 53,000 computer science students at four-year colleges compared with spring 2025, and Stanford's 2026 AI Index reports undergraduate CS enrollment at U.S. four-year universities fell 11 percent from 2024 to 2025 even as AI-related graduate enrollment kept climbing. The decline is not evenly distributed: courses closest to a traditional computer science curriculum, like coding and robotics, saw the steepest drops, while interdisciplinary computer science courses posted a modest increase.
Funding doesn't protect you from this one
One of the more counterintuitive findings in the Connecticut data undercuts a comfortable assumption. Among 20 districts studied, higher-funded districts fell 2.3 percentage points (about 550 students) from 2023-24 to 2025-26, while lower-funded districts showed almost no change. Broken out by free and reduced-price lunch status, higher-income districts fell 3.5 percentage points and lower-income districts fell 2.2 percentage points. Tusch pointed to other contributing factors too, including competing graduation requirements, budget constraints, and a persistent shortage of computer science teachers.
That matters for anyone running a technical hiring pipeline. The conventional worry about AI and education has been unequal access, with wealthier districts assumed to adapt faster and poorer districts left behind. This data suggests the opposite dynamic in at least one state: students with more resources and more exposure to AI tools are the ones opting out of the traditional CS track fastest, possibly because they have the most convincing evidence that a chatbot can do the entry-level coding work they once learned by hand.
Campuses are rebuilding around the shift, not fighting it
Universities are not waiting for computer science departments to solve this alone. Emory launched an AI minor combining psychology, economics, and English. Cornell is developing an AI and society minor aimed at both public and private sector careers. The University of Florida now offers more than 200 AI courses across all 16 of its colleges, teaching roughly 14,000 students a year, while the University of South Florida is standing up an entire Bellini College of Artificial Intelligence, Cybersecurity, and Computing on a $40 million gift. Purdue added an AI graduation requirement, Ohio State built an AI fluency requirement with hands-on workshops, and Northwestern is adding a full AI major on top of its existing minor.
The scale mismatch is the part CTOs and CIOs should sit with. The 2025 AI Degree Report counts 304 U.S. colleges now offering AI or AI-related degrees, with AI master's programs rising from 116 in 2022 to 310 in 2025 and bachelor's programs rising from 90 in 2024 to 193 in 2025. Yet a Validated Insights study found nearly 57 million Americans want to learn AI skills, and only about 7,000 of them, roughly 0.2 percent, are enrolled in a credit-earning AI program. LinkedIn's Work Change report found the share of job postings requiring AI literacy rose more than sixfold in the past year, and Microsoft's Work Trend Index found 71 percent of business leaders would rather hire someone with AI skills than traditional experience.
What this means for your hiring pipeline
If you run engineering at a PE-backed SaaS company, the entry-level funnel you have relied on is narrowing at the source. Fewer students are taking foundational coding and robotics courses, and the ones who do are increasingly folding AI tools into the work from day one rather than mastering the fundamentals first. Northwestern's computer science staff doubled to meet demand even as incoming CS majors started to shrink, which tells you departments are investing in depth for a smaller, more self-selected population rather than broad throughput.
The practical takeaway is to stop assuming a steady supply of junior engineers who learned to code the way you did. University of Washington doctoral candidate Murtaza Ali's warning is worth taking seriously: students may end up lacking understanding of basic concepts in their own field if AI does the work for them. Build your junior hiring and onboarding programs assuming new grads arrive AI-fluent but with shakier fundamentals, and budget more time for the kind of fundamentals training your own senior engineers used to take for granted.
The interdisciplinary bet is the one to watch
The one enrollment category that grew, interdisciplinary computer science, is a signal about where the next generation of technical hires will actually come from. Harvard now teaches how large language models work inside freshman writing classes, covering copyright and disinformation alongside the technical mechanics. Northwestern's Bienen School of Music offers a music and AI certificate. Eastern Mennonite University is co-teaching a class that mixes art, music, and technology majors with STEM majors. None of this looks like a traditional computer science pipeline, and that is the point.
University of Texas at Austin computer science chair Peter Stone built an introductory AI course specifically for non-computer science majors, on the premise that 'everybody needs a degree of AI literacy.' Harvard visiting scholar Paul LeBlanc's observation captures the operational challenge for every university registrar and every company building a technical bench: 'It's impossible to keep up with the technology.' Virginia Commonwealth University's Andrew Arroyo said course approval that used to take a year and a half now takes months or sometimes weeks. That pace of curricular change is itself a signal CIOs should read: the skills your next hires carry will look different every single admissions cycle.
