12.7 Million Graduates, One Vanishing Entry Level: Inside China's AI Jobs Squeeze

A record 12.7 million young people are pouring out of Chinese universities this year into a job market that is being restructured underneath them in real time. As The New York Times reported on September 6, China’s newest graduating class — the largest in the country’s history, roughly 4% bigger than the 12.22 million who graduated in 2025 — is confronting a double bind: a labor market already strained by years of weak growth and a housing downturn, plus an artificial-intelligence adoption wave that is systematically absorbing the entry-level white-collar work that used to be the first rung of a career.

The headline number that frames the story is youth unemployment. The jobless rate for urban residents aged 16 to 24, excluding students, stood at 17.9% in July — roughly triple the overall urban unemployment rate of around 5%. That is not merely cyclical. Reporters tracing this year’s graduates find that the skills universities are producing are increasingly misaligned with what employers now say they need, because the starter tasks those employers used to pay juniors to learn on — drafting, translating, basic coding, summarizing, first-pass analysis — are increasingly done by software.

The first rung is being sawed off

The structural worry is easy to state and hard to fix. Careers in knowledge work have traditionally been pyramids: graduates enter at the bottom doing routine work, and the routine work is where they absorb judgment, context, and domain knowledge. When AI does the routine work, two things happen at once. Companies save money on junior headcount, and the pipeline that produces senior people narrows.

China is feeling this earlier and harder than most economies for a specific reason: adoption is state-accelerated. Under the “AI Plus” initiative and the five-year plan running through 2030, Beijing has been pushing AI diffusion across every industry as a matter of national technology strategy. The share of Chinese industrial enterprises reporting use of AI models and agents jumped to 47.5% last year, up from 9.6% in 2024, according to market-intelligence firm IDC. Zilan Qian of the Oxford China Policy Lab puts it plainly: what is specific to China is that the government is actively diffusing AI across the economy, so AI enters different domains faster than in other countries.

That velocity shows up in the job listings. Recruitment portal Maimai found that AI skills are now required for four out of ten graduate job openings in China, and in Beijing the ratio is five out of ten. The jobs that remain for fresh graduates are not the jobs their degrees were designed for.

The human texture behind the statistics

An Associated Press report published via PBS NewsHour in late August captured what this looks like at desk level. Fei Zhaojun, a 40-year-old programmer in Beijing, was asked by his boss whether AI could soon replace humans in coding jobs. Two weeks later he was laid off, along with about 160 colleagues. His verdict on his own profession is bleak: “Mid-level coders’ jobs are essentially replaceable in most of the cases.” He is now making vlog-style short videos during a career break.

Du Qinchun, a part-time translator in Chengdu, has been hired to help train the very AI models that compete with him — more work, at least temporarily, but “the pay in the industry has been cut by more than half compared to what it was years ago.” Foreign-language programs, once among the most popular college majors in China, have fallen out of favor as machine translation becomes ubiquitous. Wang Zhicheng, a 32-year-old scriptwriter for children’s educational content, watched his company lay off roughly half of its 13 scriptwriters after AI took over much of the drafting. He resigned and now runs his own studio making picture books. “You can treat AI as a tool just like Word,” he says. “Humans are still the decision makers on which one to pick or pursue among all that AI generates.”

The short-drama industry gives the displacement a number: live-action short and vertical video series produced for mobile phones fell about 75% year over year in the first quarter, according to Chinese media reports, as generative AI took over creation, production, and distribution. The International Labour Organization adds a distributional warning — women face higher displacement risk than men because they are concentrated in automation-suitable roles like electronics assembly and remain underrepresented in science and technology fields.

Beijing’s answer: jobs-first AI, whether or not that coheres

The policy response is already visible, and it is distinctly Chinese in flavor. Beijing’s 2026 employment strategy leans on “AI-adjacent” new occupations — steering graduates toward roles that wrap around the machines rather than compete with them: data annotation, model training, AI operations, prompt engineering, robot maintenance. Local governments are holding targeted job fairs for graduates and pressing state-aligned employers to expand hiring quotas.

Courts, meanwhile, have been quietly building a doctrine. A series of precedent-setting rulings this year awarded compensation to white-collar workers laid off and replaced by AI, signaling that Chinese judges expect companies to bear some of the transition cost. Reuters Breakingviews framed the tension precisely in June: with youth unemployment at 16%, Beijing wants companies to protect workers while adopting new technologies — a contradictory policy that risks blunting AI’s productivity gains even as it props up employment. It is the classic corporatist bargain, renegotiated at machine speed.

The counterargument worth taking seriously

Not every analyst reads this as a death spiral, and the strongest counterpoint is demographic. China’s population of 1.4 billion is rapidly aging and shrinking. Xuenan Cao, a professor at San Francisco Bay University who studies technology and society, notes that by 2050 China is projected to have fewer than two working-age adults supporting each retiree, compared with more than 2.5 in the United States. On that view, automation could partially offset a shrinking workforce rather than being purely a threat to it. Cornell economist Eswar Prasad pushes back on the comfort: AI may lift productivity across the board, but it “could have a severe disruptive effect on employment, worsening the employment growth problem and resulting in a detrimental effect on social stability.”

Both can be true at once, and that is the uncomfortable synthesis. In the long run, China may need every robot and model it can deploy just to keep the care economy and industrial base staffed. In the short run — this graduating class’s five-year horizon — the pyramid’s bottom rung is being sawed off faster than new rungs are being built, and 12.7 million people are standing at the base of the ladder.

Why this story travels beyond China

Two features make China’s graduate squeeze a leading indicator rather than a local curiosity. First, the adoption curve: China’s combination of open-source model ecosystems (which drive down the cost of industrial deployment), state mandates, and hyper-competitive enterprise software has produced the fastest workplace AI diffusion of any major economy. What happens to Chinese junior coders and translators this year is a preview of the question every labor market will face — who gets the first job when AI can do it?

Second, the political economy: China’s government is uniquely positioned to force a jobs-first compromise — court rulings, hiring quotas, industrial policy — in a way that Western governments mostly are not. If Beijing succeeds in decelerating displacement without killing the productivity gains, it will have built something like a template. If it fails, and youth unemployment keeps climbing while the economy’s AI dividend accrues only to capital, the social-stability warning stops being a footnote and becomes the story.

For the graduates themselves, the advice embedded in every account is remarkably consistent, and remarkably unsentimental: use the tools, don’t compete with them. The translator trains his replacement because the training pays. The coder who doubted AI “just uses it, as everyone else is using it.” The scriptwriter becomes a decision-maker over machine output rather than a producer of first drafts. It is adaptation advice, not career advice — and the fact that a record 12.7 million people are receiving it simultaneously is itself the news.