U.S. Computer Science Major Applications See Largest Decline in Six Years
According to The Wall Street Journal, computer science degrees, once seen as a "golden ticket" in the U.S. job market, are losing their luster. The number of undergraduate admissions for computer science at four-year colleges in the U.S. has seen the largest single-year decline in six years, dropping by 8.1% year-on-year, while graduate admissions have decreased by 14%.
Data cited in the report shows that from 2008 to 2024, the number of computer science degrees awarded by U.S. four-year colleges grew nearly fivefold, outpacing growth in nursing and mechanical engineering. However, this expansion has clearly reversed. For example, at Princeton University, only 74 students declared a computer science major for the class of 2028, a significant drop from 117 last year and 150 two years ago. Meanwhile, admissions for mechanical engineering have increased by about 11% year-on-year, and electrical engineering by about 14%. Currently, fields such as business, healthcare, social sciences, and history have surpassed computer science in popularity, with the latter dropping to seventh place in rankings of popular majors.
The deteriorating job market for recent graduates is a direct driver of this shift—new graduates have experienced an unemployment rate that has been higher than the national average for five consecutive years. According to industry statistics, the overall unemployment rate in the tech sector reached 5.8% this year, the highest level since the dot-com bubble burst in 2001-2002. Since the beginning of the year, layoffs in the U.S. tech sector have reached 148,092, averaging a loss of 981 jobs per day, a 46% increase compared to the average of 674 layoffs per day in 2025. The annual layoff total is expected to reach around 370,000. A report from Stanford University's Human-Centered AI Institute found that employment for young software developers aged 22 to 25 has decreased by nearly 20% compared to 2024, while employment for developers over 30 in the same companies has increased by 6% to 12% during the same period.
Nobel laureate Simon Johnson commented that AI has "largely erased the value of programming as a reliable career opportunity." Investor Matt Shumer has also pointed out that "by the end of 2025, some of the world's top engineers will have delegated most programming tasks to AI." The chair of Princeton's computer science department stated that the discipline "is no longer growing at that explosive rate."
A clear divide is emerging within the job market—positions for machine learning engineers have increased by 59% compared to February 2020, while AI/machine learning-related job postings have grown by 85% and cybersecurity-related postings by 124%, showing structural strength. However, the number of traditional general software engineering positions is down 49% compared to pre-pandemic levels, with the proportion of entry-level positions in overall IT hiring dropping from 8.1% to 7.4%, while the share of senior positions has risen from 38.8% to 43.1%. Employers' attitudes are also changing: according to data from the National Association of Colleges and Employers (NACE), computer science remains the third most in-demand major among employers, with starting salary expectations at the top of all majors. However, employers are increasingly looking for "frontline deployment engineers" who can apply AI tools in specific business scenarios, rather than graduates who merely master easily replaceable programming tools. Computer science graduates with internship experience have twice the chance of being hired compared to those without.
From the perspective of talent and funding flow, this shift is essentially a direct transmission of AI capital expenditure to the labor market—companies are reallocating budgets originally intended for hiring junior programmers to deploy AI programming tools and recruit more experienced senior engineers and "AI application" hybrid talents, leading to a systematic compression of entry-level job demand. Salaries for positions requiring scarce skills like AI/machine learning and cybersecurity continue to rise (with machine learning skills commanding a 40% salary premium and TensorFlow-related skills a 38% premium). The beneficiaries are graduates and senior engineers with AI application experience, internship backgrounds, or scarce technical certifications, while those who only possess basic programming skills and lack differentiated skills and practical experience face increased difficulty in job hunting due to the contraction of entry-level positions.
ABAB AI Insight
The previous expansion path of computer science majors is a historical narrative worth revisiting. From 2008 to 2024, the number of computer science degrees awarded by U.S. four-year colleges grew nearly fivefold, driven primarily by strong recruitment demand during the golden growth period of the internet and mobile internet in the 2010s. The current largest decline in enrollment in six years is essentially a delayed correction of the collective perception that "studying CS equals stable high salaries" over the past decade. Similar historical precedents of "majors cooling off" include the brief decline in computer-related major enrollments following the burst of the internet bubble in the early 2000s, but this time the driving factor has shifted from a drop in demand to the systematic replacement of entry-level positions by AI.
The path of reallocating funds and resources is clear—companies are shifting budgets originally intended for mass hiring of junior programmers towards purchasing AI programming tool subscriptions, deploying automated development pipelines, and concentrating the saved budget on a few senior engineers with scarce skills in machine learning and cybersecurity. The recruitment of machine learning engineers has increased by 59% compared to the baseline, and cybersecurity recruitment has grown by 124%, contrasting sharply with the 49% shrinkage of general software engineering positions compared to pre-pandemic levels. Goldman Sachs estimates that AI is net reducing about 16,000 U.S. jobs per month (eliminating about 25,000 jobs while creating about 9,000 new jobs through efficiency gains), indicating that capital is replacing entry-level labor with AI tools in the most easily standardized and replicable tasks, while concentrating scarce human capital budgets on complex and senior positions that AI cannot easily replace.
The most direct historical analogy is the brief cooling of computer science majors following the burst of the internet bubble in the early 2000s, as well as the cyclical fluctuations in law school applications with the proliferation of automation tools in the legal industry. Whenever a major's previously promised "certainty of high salary paths" is proven to be no longer certain, applications tend to experience a delayed but severe self-correction. The difference this time is that the trigger for this correction is not an overall industry downturn, but the direct replacement of entry-level skills by AI, while demand for machine learning and AI-related positions continues to grow. In terms of industry positioning, the U.S. tech labor market is currently undergoing a structural re-layering phase characterized by "entry-level positions being compressed by AI, while senior and hybrid positions remain in high demand." The signaling value of a single degree is being diluted, with internship experience, project work, and verifiable AI application abilities becoming more important screening signals than degrees.
This is essentially a technological replacement—AI is redefining software engineering, previously seen as "the most certain white-collar job," into two levels: "standardized coding that can be automated" and "complex judgment and business application capabilities that AI cannot replace." The labor demand for the former is being systematically compressed, while the scarcity and salary premiums for the latter are rising simultaneously. This replacement occurs because entry-level coding tasks are highly patterned and rich in training data, making them tasks that current large language models excel at handling. In contrast, the complex abilities required to understand specific business scenarios, coordinate multiple interests, and apply AI tools are still difficult for models to fully replace. Therefore, the labor market is being redefined along the new boundary of "whether it can be standardized and replicated by AI," reshaping the scarcity and pricing of different skill combinations.
ABAB News · Cognitive Laws
- Skills that are easier to standardize are the first to be zeroed out by AI.
- Degrees were once a signal; now they are merely copies of entry tickets.
- The scarce part has never been the degree; it is the irreplaceable judgment beyond the degree.