In 2018, Carnegie Mellon launched the first bachelor’s degree in artificial intelligence in the United States.
At the time, this was a mildly controversial move. The prevailing view in academia was that AI didn’t deserve its own degree. It was a specialization inside computer science, the way databases or graphics were. Giving it a standalone major felt like giving “spreadsheets” a standalone major.
Eight years later, there are 193 AI bachelor’s programs in the US. And 310 master’s programs.
I work on a degree directory, so part of my role is to count these things for a living. And I’ve never seen a credential category grow this fast. AI master’s programs went from 116 to 310 between 2022 and 2026. Bachelor’s programs more than doubled in just two years, from 90 in 2024 to 193 today.
For context: it took cybersecurity degrees roughly 15 years to go through the same adoption curve. AI did it in four.
Students figured this out before universities did
The demand side of this story is even more lopsided than the supply side.
SUNY Buffalo’s AI master’s program had 5 students in 2020. By 2024 it had 103. That’s a 20x increase at a school nobody would call an AI powerhouse.
Meanwhile, the Computing Research Association reported that 62% of traditional computer science programs saw undergraduate enrollment decline this past fall. Students aren’t abandoning tech. They’re re-labeling themselves, and they’re doing it faster than the institutions are.
The economic signal driving this is not subtle. Lightcast analyzed over a billion job postings and found that roles mentioning AI skills advertise salaries roughly 28% higher, about $18,000 more per year. List two or more AI skills and the premium jumps to 43%.
An 18-year-old choosing a major can read that data as well as any provost.
Here’s my problem with a lot of these programs
I’ve spent 15+ years watching companies chase job/field demand, and I recognize the pattern when I see it in higher ed.
“MS in Computer Science, Machine Learning Concentration” doesn’t stand out that much in 2026. “MS in Artificial Intelligence” does. So a meaningful chunk of these 500+ programs are the same courses, same faculty, same syllabi, wearing a new name. The university didn’t build an AI program. It did keyword optimization on an existing one.
That’s not automatically bad. A rebranded CS degree with strong ML coursework is still a good degree. But students are paying dedicated-AI-program prices for it, and they deserve to know which one they’re buying.
A quick metric I’d give any prospective student: pull up the course catalog and check when the core AI courses were created. If the “deep learning” course predates 2020 and hasn’t been revised, you’re looking at a rebrand. Also check whether the faculty teaching those courses actually publish in the field, and whether the curriculum touches modern infrastructure (training pipelines, evaluation, deployment) or stops at a survey of techniques from 2015.
Carnegie Mellon and Penn (which launched the first Ivy League undergrad AI degree in fall 2024) built their programs from the ground up. Plenty of schools did the honest work. Plenty didn’t.
The obvious objection, and why I only half agree with it
The standard critique of AI degrees goes like this: the field reinvents itself every 18 months, so a four-year curriculum is teaching 2026’s frontier to the graduating class of 2030. By the time you finish, half of what you learned is obsolete.
This is true, and it’s also kind of a bad argument against the degree.
The stuff that goes stale is the tooling layer. Frameworks, model architectures, whatever technique is currently winning benchmarks. No degree should be built on that layer anyway.
The stuff that doesn’t go stale is linear algebra, probability, optimization, and systems thinking. That math hasn’t changed in decades and it’s the actual substrate of everything happening in AI right now.
Which means the programs that will age worst aren’t the rigorous ones. They’re the trend-chasing ones, the programs padding credit hours with “prompt engineering” coursework. Prompt engineering is a skill you pick up in a weekend, and universities charging tuition for it are selling a depreciating asset at a premium.
If a program is 70% math and fundamentals and 30% current techniques, it’ll hold value. If those ratios are flipped, it won’t.
The gap universities have to keep in mind
Here’s the number that actually stopped me when I first saw it.
Roughly 57 million Americans say they’re interested in learning AI skills. About 8.7 million are actively learning them in any format. And as of last year, only around 7,000 were doing it through a credit-bearing program at a higher education institution.
Seven thousand. Out of 57 million.
Higher ed is capturing something like 0.01% of the demand for AI education. Everything else flows to YouTube, bootcamps, vendor certificates, and increasingly to the AI tools themselves, which are genuinely decent at teaching you how to use them.
Universities move at committee speed. Approving a new degree takes 18 to 36 months at most institutions. The market they’re trying to serve moves in quarters. Even with 500+ programs live, supply isn’t remotely close to demand, which tells you the real growth in AI education is happening outside the degree system entirely.
The e-commerce degree question
There’s a historical precedent worth remembering. Around 2000, universities launched “e-commerce degrees.” They were everywhere for about five years, then vanished, because e-commerce stopped being a specialty and became just commerce.
Some people think AI degrees follow the same arc. I don’t, but it’s still an interesting little tidbit.
Lightcast found that 51% of job postings requiring AI skills are now outside IT and computer science entirely. That half of the market (the marketer, the analyst, the HR lead who needs AI fluency) will never enroll in an AI degree. For them, AI dissolves into every existing program the way Excel did. No one gets a spreadsheet degree.
But the builder path is very different. The people designing, training, and deploying these systems need depth that a bolted-on elective can’t provide. For that track, I’d bet the dedicated AI degree doesn’t just survive. Within a decade, it quietly becomes what computer science was for the last thirty years: the default technical degree, with CS as the legacy label.
The 62% of CS programs watching their enrollment shrink are already living in that future. They just haven’t renamed themselves yet.