Bloomberg reports that US computer and information science enrollment fell by more than 12% for undergraduates and more than 10% for graduate students this fall, according to National Student Clearinghouse data, and links the exodus to AI reducing coding jobs. The report surfaced on Techmeme on October 29, 2026. We could not open the Bloomberg article itself, so this post treats the new figure as reported, not independently confirmed, and checks it against the Clearinghouse and other data that is public.
The short version: the direction is well supported, the exact fall 2026 number is Bloomberg's, and the "AI did it" explanation is an inference that enrollment counts cannot prove on their own. If you are choosing what to study, hire for, or teach, that distinction matters.
TL;DR: claimed vs verified
| Question | Answer |
|---|---|
| What is the claim? | US CS undergrad enrollment down 12%+, grad down 10%+ this fall (Bloomberg, via Techmeme) |
| Who is the data source? | National Student Clearinghouse, per Bloomberg |
| Did we read the original? | No. The article is paywalled and the Clearinghouse site blocked our fetch, so base year and sector mix are unconfirmed |
| Is a decline real? | Yes. Earlier Clearinghouse releases show falls in fall 2025 and spring 2026 |
| Is the 12% figure comparable to earlier ones? | Not necessarily. Terms, sectors and preliminary vs final data differ |
| Does it prove AI caused it? | No. The link is the reporting's interpretation; researchers call the data descriptive |
| What should a builder or student do? | Treat it as a signal about entry-level hiring, and invest in shipped projects and agent fluency |
What exactly is being claimed
The Techmeme summary of Bloomberg's piece, by Elizabeth Rembert, says computer and information science departments at US colleges "saw an exodus of students this fall," with enrollment declining more than 12% for undergraduates and more than 10% for graduate students, citing the National Student Clearinghouse and attributing the shift to AI reducing coding jobs. That is everything we can attribute to the report directly.
What we do not know from the summary: whether the 12% is year over year or term over term, whether it covers two-year colleges, four-year colleges or both, whether it is the Clearinghouse's preliminary fall snapshot, and how it treats students who moved from "computer science" into adjacent labels like data science. Each of those can move a headline number by several points.
What the earlier data confirms
Public coverage of earlier Clearinghouse releases gives the trend line the new figure would sit on:
- Fall 2025: four-year undergraduate computer science fell about 8.1%, from roughly 659,700 to 606,100 students, even as total postsecondary enrollment grew about 1.0%. The preliminary report had graduate computing programs down as much as 15%.
- Spring 2026: undergraduates majoring in computer science fell about 8.4% at four-year institutions (around 53,000 students) and 11.2% at two-year colleges, while overall undergraduate enrollment still rose roughly 1.3%.
- Computing Research Association: its Taulbee survey reported new bachelor's enrollment in computing down 12.9% in 2025 for its cohort of departments, which is the closest independent match to a "12%" headline.
- Stanford's Jacob Light: course-level data across 1,019 institutions showed computer science course enrollment down about 4.6% in 2025-26, and the analysis was presented as descriptive, not causal.
So a further fall of 10-12% is plausible against that background, but note the mismatch in definitions. An 8.1% four-year figure and a 12% all-sector figure are not the same measurement. When Bloomberg's full table is available, check the sector split first.
This is also not the first time the story has run. A Washington Post report in April about a sudden drop in CS enrollment drew a long Hacker News discussion, and TechCrunch ran "The great computer science exodus" in February. What is new this week, if Bloomberg's numbers hold, is that the decline did not flatten out.
What the data cannot tell you: why
The causal sentence in the headline, "as AI reduces coding jobs," is the part to be most careful with. Enrollment is a choice made months or years before graduation, based on perceived job prospects. Students reacting to viral posts about AI replacing programmers would produce the same chart as students reacting to a genuinely worse entry-level market, or to the end of a 2010s boom that pulled in people chasing salary.
Evidence that bears on the jobs side is mixed and mostly separate from enrollment:
- Stanford researchers including Erik Brynjolfsson have found employment for software developers aged 22 to 25 down nearly 20% since ChatGPT launched, while older developers held up. That fits an entry-level squeeze, and it is the strongest piece of evidence for the AI story.
- Anthropic has published its own view on junior hiring and agents; we covered it in Anthropic's employment and junior hiring analysis.
- Several employers say the opposite about skills: one Anthropic manager argued that hunger beats tenure and that coding interviews are being dropped, which is a hiring-process change rather than a demand collapse.
- Amazon's new engineer role guideline pushes a "renaissance developer" who leans heavily on AI tooling, which changes the job without eliminating it.
Other things are happening at once. Interest rates, big-tech layoffs, post-pandemic over-hiring and visa policy all affect new-graduate demand. Students are also moving toward data analytics, data science and AI-labelled programs, which the Clearinghouse itself has noted, so some of the "lost" CS students may simply be counted elsewhere.
What developers are saying
The April Hacker News thread on the Washington Post piece is a useful read on how practitioners interpret this kind of number. These are commenters' views, not verified facts:
- One commenter, garbawarb, argued there was a golden age when a CS bachelor's alone was an easy route to the upper middle class, and that fewer people now enter the field mainly for the money because getting a new-grad job is harder.
- cedws said Gen Z holds a doom-laden view of CS careers shaped by social media influencers calling the path a dead end because of AI, while believing studying CS can still lead to decent careers.
- rizza compared it with dentistry, law and electrical engineering, where a "golden ticket" reputation drew a crowd and a shock then reset demand, and expected a cycle of five to ten years.
- ModernMech offered a counterpoint from a high school career expo, where computer science was among the two most asked-about majors, and warned against calling a trend early.
The disagreement itself is the useful bit: even people close to the industry split between "AI scare", "market correction after a bubble" and "normal cycle".
What this means for what you build, hire or study
For a reader who builds with AI, the enrollment story is a lagging indicator of something you can act on now.
If you are a student or career changer. Treat a CS degree as necessary for some roles and not sufficient for entry-level ones. Evidence of shipped work is what separates applicants when junior roles are scarce. Build two or three projects where you used coding agents to ship something real and can explain every decision, including what the agent got wrong.
If you hire. A smaller pipeline of graduates in four years is a supply signal. Companies that stop hiring juniors now may find themselves short of mid-level engineers later; some, like the ones in our India IT entry-level hiring analysis, are already under scrutiny on exactly this.
If you teach. Curricula that treat AI assistance as cheating on one side and ignore it on the other both miss the point. The skill now is directing, reviewing and testing agent output. Our Anthropic worker retraining evidence review is a good starting point on what the research says about retraining that works.
If you are tracking the macro picture. Pair this with the Goldman Sachs estimate that AI could displace about 9% of workers, and note the same caution applies: projections and enrollment counts are not employment outcomes.
Job cards beside a bar chart, illustrating computer science enrollment decline numbers and the entry-level job market
How to check the figure yourself
- Open the National Student Clearinghouse Research Center's enrollment reports and the computer science enrollment analysis and find the fall 2026 term release, noting whether it is preliminary.
- Match the sector: two-year, four-year public, four-year private. A single national percent hides large differences.
- Compare like with like: undergraduate majors versus course enrollments versus new-student intake are three different measures (compare Clearinghouse, CRA Taulbee and Stanford's course data above).
- Look at adjacent fields such as data science, information systems and AI-labelled majors for migration.
- Ask whether any cited jobs data is for the same age band and period as the enrollment change.
What would change our read
Three things would sharpen this story. A published table from the Clearinghouse showing the exact fall 2026 figures and sectors; graduate outcome data showing whether the 2022-2024 cohorts that entered during the AI-coding boom found jobs; and employer data showing whether junior postings recover. Until then, the responsible reading is that enrollment is falling, entry-level hiring is weaker, and the causal weight of AI relative to everything else is unresolved. We will update this post when the original table is public.
Related reading
- Anthropic on junior hiring and agents
- India IT entry-level hiring and AI: claimed vs verified
- Goldman Sachs: AI could displace 9% of workers
- Amazon's renaissance developer guideline
- Anthropic manager drops coding interviews
- Anthropic worker retraining evidence review
Figures reflect public reporting as of October 29, 2026. The Bloomberg figure is as summarized by Techmeme and has not been checked against the original Clearinghouse table.