CS Enrollment Dropped 11%. The Industry That Needs Developers Told Students Not to Bother.

You know the scene in The Karate Kid where Daniel-san spends weeks waxing cars, painting fences, and sanding floors before Mr. Miyagi lets him throw a single punch? He hates every second. It feels pointless. He almost quits twice. But when the fight comes, his body already knows what to do — the fundamentals are in his muscles, not his head.

Now imagine a version where Daniel just downloads a karate app. He skips the wax-on-wax-off. He generates kicks with a prompt. He looks great in practice. And then Cobra Kai shows up, and he doesn’t even know how to fall without breaking something.

That’s computer science education in 2026.

The Numbers Don’t Lie (But They Do Sting)

According to the National Student Clearinghouse, computer science enrollment dropped 8.1% in the 2025-2026 school year. That’s the steepest decline of any field of study. Not English. Not philosophy. Not even art history. Computer science.

Look at the CS-specific breakdown and it’s worse: an 11.2% plunge. The major fell from the 4th to the 6th most popular degree in the United States. Over 62% of universities surveyed by the Computing Research Association reported declining enrollment. Only 13% saw growth. The rest? Flatlined.

The University of California system — the crown jewel of public tech education — saw a 6% drop. Graduate CS enrollment? Down 15%.

These aren’t rounding errors. CS degrees roughly quintupled between 2008 and 2024, from about 51,000 to 112,000 graduates per year. For the first time in a generation, that number is going the wrong way.

Why Students Are Bailing

Ask any freshman why they switched from CS to electrical engineering (up 13.8%) or a shiny new AI-specific program, and you’ll hear the same answer.

64% of pessimistic CS majors cite generative AI as the reason.

They watched Anthropic’s CEO say AI could eliminate 50% of entry-level jobs. They read the headlines about 80,000 developers getting laid off for AI. They saw Salesforce announce it would halt junior hiring entirely. They noticed that bootcamp graduate conversion rates dropped below 2% application-to-interview. And they did the math.

Why spend four years learning algorithms, data structures, and operating systems when Claude Code can generate a working app from a paragraph of English? Why take on $100K in student debt for a 6.1% unemployment rate — nearly double most other majors — when the CEO of the company you’re applying to told CNBC that one senior engineer with AI tools now does the work of a senior plus a junior?

A junior developer at a San Francisco tech company told the SF Standard: “I’m basically a proxy to Claude Code. My manager tells me what to do, and I tell Claude to do it.” He’s not complaining about his education. He’s questioning whether he needed one.

The Irony Is Thick Enough to Deploy to Production

Here’s what makes this story feel like a bad comedy sketch.

The same industry that told students “AI will do your job” is simultaneously posting about a developer talent shortage. The same companies that cut junior positions by 73% in a single year are now complaining they can’t find mid-level engineers. The same executives who bragged about replacing entire teams with AI are quietly rehiring the people they fired — sometimes at higher salaries.

It’s like a restaurant that fires all its line cooks, replaces them with a fancy food processor, discovers the food processor can’t plate or season or adjust on the fly, and then puts up a sign saying “EXPERIENCED COOKS WANTED — 5 YEARS MINIMUM.”

Where exactly are those experienced cooks supposed to come from if you never let anyone into the kitchen?

The Pipeline Problem Nobody Wants to Solve

This is the part that should worry anyone building software — including iOS developers.

If students don’t enroll in CS, they don’t become junior developers. If they don’t become junior developers, they never become mid-level. If they don’t become mid-level, the pipeline to senior engineer, tech lead, and architect just… stops.

We already wrote about this using the youth academy analogy: if you shut down the farm team, you don’t have any stars in five years. The enrollment data now shows it’s worse than we thought. The students aren’t just failing to get hired after graduation — they’re not showing up to class in the first place.

The Bureau of Labor Statistics still projects software developer employment to grow 17% through 2034. The demand isn’t disappearing. But the supply is choosing to go somewhere else.

Some students are pivoting to electrical engineering. Some are going into dedicated AI programs. UC-San Diego reports that roughly 20% of CS applications now target its AI-specific track. The University of South Florida’s new AI college already has nearly 3,000 students. Northwestern, Columbia, and USC are all launching AI programs this fall.

The message is clear: students aren’t leaving tech. They’re leaving traditional CS. They’re betting that the future belongs to the person who builds the AI, not the person who writes the code the AI generates.

They might be right. They might also be making the same mistake as someone who studies automotive design without ever learning to change a tire.

The Wax-On-Wax-Off Problem

Here’s what gets lost in the enrollment data: the fundamentals still matter more than ever.

AI made developers 19% slower in a peer-reviewed study — while those same developers swore it made them 20% faster. The 39-point perception gap wasn’t because AI is bad. It’s because developers who understand what the AI is doing can catch its mistakes. Developers who don’t… can’t.

When Claude Code hallucinates a function that doesn’t exist, you need to know enough to recognize it. When it generates a race condition wrapped in perfectly formatted Swift, you need to understand concurrency to spot it. When it architects a solution that technically works but scales like a shopping cart with a flat tire, you need the systems thinking that comes from — yes — four years of sanding floors.

The developer who never learned to debug without AI assistance is the developer who can’t debug when AI assistance fails. And AI assistance fails a lot. CodeRabbit’s 470-PR analysis found AI-generated code has 1.7x more issues than human-written code, with 8x more performance problems.

The wax-on-wax-off isn’t pointless. It’s the whole point.

What’s Actually Growing (and What It Means for iOS)

For anyone in the Apple ecosystem, the enrollment shift has a specific flavor.

Apple’s Foundation Models framework requires developers who understand both AI capabilities and native platform fundamentals. SwiftUI, concurrency, App Intents, CloudKit — these aren’t things an AI-only education covers. You need someone who understands @Observable, who can debug a ModelContainer migration, who knows why nonisolated(unsafe) exists and when it’ll bite you.

The students flooding into AI programs will learn transformer architectures and loss functions. The students staying in CS will learn algorithms and systems. But the students who learn both — native platform development plus AI integration — those are the ones building the next generation of apps.

If you’re a student reading this and wondering whether to stick with CS: yes. But don’t just study it. Build things. Ship an app. Learn SwiftUI not because a degree requires it, but because understanding how a real platform works is the difference between prompting an AI and actually knowing if the AI’s output is any good. Our SwiftUI courses exist precisely for this — building real apps, from module architecture to CloudKit to the App Store.

This Happened Before

If you’re old enough to remember 2001, you remember the first time CS enrollment cratered. The dot-com bubble burst, and overnight, freshmen decided computers were over. Enrollment didn’t recover until 2007.

Then it happened again after the 2008 financial crisis, though milder.

Every time the pattern is the same: a shock hits the job market, students flee, the industry spends five years complaining about talent shortages, and then enrollment slowly recovers — usually right around the time the next shock arrives.

The 2026 version has a new ingredient: AI. But the cycle is familiar. The question is whether this time the recovery looks different — whether “CS degree” evolves into something broader, or whether it slowly fades into the same niche as a library science degree.

My bet? CS adapts. It always has. The degree in 2030 will look nothing like the degree in 2020, the same way a 2010 CS program looked nothing like a 1995 one. But the core — algorithms, data structures, systems thinking, the ability to reason about correctness — that’s permanent. That’s the wax-on-wax-off that survives every hype cycle.

The Bottom Line

CS enrollment is dropping because the industry sent a confusing message: “We don’t need you, but also, where are you?”

Students heard the first half. The second half is a quiet panic happening in HR departments across Silicon Valley.

Here’s the thing nobody in a keynote will say: AI tools are only as good as the person auditing them. And auditing requires the exact skills a CS degree teaches — logic, systems thinking, debugging, understanding trade-offs. The industry doesn’t need fewer CS graduates. It needs CS graduates who also understand AI. It needs people who can wax and punch.

If you’re a student: don’t drop out. Specialize. Learn to build, not just to prompt. The developers who thrive in 2030 won’t be the ones who skipped fundamentals — they’ll be the ones who used AI to go faster while knowing exactly where they were going.

If you’re a hiring manager: stop telling students AI will take their jobs and then complaining when they believe you. Reopen the youth academy. Hire juniors. Train them. It’s cheaper than the alternative, and the alternative is already showing up in your empty interview pipeline.

Daniel-san didn’t beat Cobra Kai because he was stronger. He beat them because he’d done the boring work. In 2026, the boring work is still the shortcut.

Cover photo by Fabio Sasso on Unsplash.