Are enterprises struggling to recruit despite the frenzy of opening AI majors?

In the 2026 campus recruitment, Baidu, where AI positions account for 90%, and small and medium-sized enterprises with extremely low click - through rates on recruitment ads, are presenting two extremes of the same war.

According to news reports: Recently, Internet giants such as Baidu, ByteDance, Alibaba, and Tencent have all launched their 2026 campus recruitment campaigns, increasing their recruitment efforts for artificial intelligence (AI) - related positions. From the perspective of demand, leading enterprises are eager to recruit talents, and the "talent scramble" has reached a white - hot stage. However, many small and medium - sized enterprises are suffering from the inability to recruit people. From the perspective of supply, top talents hold several job offers, and an annual salary of 300,000 yuan is just the "starting price", while more job seekers are deeply trapped in the pain of mass - applying and sighing about the "involution" in job hunting.

I want to conduct a discussion and analysis of the situation behind this phenomenon, hoping to give you some inspiration.

01

First, let's take a look at the recruitment market

From Baidu, ByteDance, Alibaba, and Tencent to Huawei, Meituan, JD.com, Kuaishou, Xiaomi, Pinduoduo, and Ant Group, campus recruitment has entered an era dominated by AI. Large enterprises are paying more and more attention to campus recruitment, and the functional stratification of AI is becoming more and more obvious:

Baidu: AI positions account for over 90%. In the 2026 campus recruitment, Baidu plans to issue more than 4,000 offers, of which over 90% are concentrated in AI - related positions, focusing on areas such as large models, multi - modality, intelligent applications, and unmanned driving.

ByteDance: AI positions also account for 90%. It offers more than 5,000 positions, with a 23% year - on - year increase in R & D positions. It has also launched special programs such as the "Top Seed Elite Talent Program" and the "Jingdouyun Talent Program", focusing on cutting - edge fields such as large models, AI security, AIGC, and cross - modality.

Alibaba: AI - related positions account for as high as 60 - 80%, including fields such as algorithms, AI product managers, intelligent digital humans, medical AI, and embodied intelligence.

Other manufacturers: A similar trend is accelerating. According to Yicai Global, the proportion of AI recruitment in large enterprises such as Alibaba, Tencent, Baidu, and ByteDance is generally high. For example, among the 10,000 positions recruited by ByteDance, 2,353 are AI positions.

Partial screenshots of the AI - related recruitment for the 2026 campus recruitment of Baidu, ByteDance, Alibaba, and Tencent are as follows:

Source: Internet

02

Second, let's take a look at the domestic cultivation situation

Which schools are focusing on AI majors?

In 2018, the Ministry of Education issued the "Action Plan for AI Innovation in Higher Education Institutions", promoting the construction of a number of AI colleges/research institutes, interdisciplinary platforms, and industry - university - research cooperation mechanisms. AI has entered a stage of systematic layout. "Double First - Class" universities and universities with strong engineering disciplines have continued to increase their efforts: the Institute for AI and Industry Research (AIR) at Tsinghua University, the School of AI at Peking University, AI - related colleges/centers at Shanghai Jiao Tong University, and interdisciplinary AI research platforms at Zhejiang University have built scientific research and talent cultivation systems around "AI + industry/science".

The "Catalogue of Undergraduate Majors in General Higher Education Institutions" has listed "Artificial Intelligence", "Intelligent Science and Technology", "Data Science and Big Data Technology", "Robot Engineering" and other majors as key related majors. Each school has also set up AI directions in computer science, automation, and electronic information majors. In the past few years, the number of undergraduate programs in AI has expanded rapidly. Statistics from multiple media and university channels show that the number of universities offering undergraduate programs in "Artificial Intelligence" has reached about 500 and is still increasing!

Top universities have recently further expanded their enrollment in "strategic disciplines" (including AI, chips, computational mathematics, etc.) to meet the needs of new - quality productivity and industrial upgrading. Represented by joint laboratories/practical projects of enterprises such as Tsinghua AIR, Alibaba/Alibaba DAMO Academy, Tencent AI Lab, and Baidu, they emphasize "real data + real scenarios + engineering implementation".

03

Third, think about why the "supply" and "demand" are unbalanced

The "hot" end: Top talents are being snatched up. Leading Internet and large - model companies are placing AI positions at the forefront of campus recruitment. Taking Baidu as an example, in the 2026 campus recruitment, AI - related positions account for over 90%, and it is open to full - stack directions such as large models/multi - modality/training frameworks/cloud - native. A large number of offers and special elite programs are carried out in parallel. Industry reports generally mention that the "top campus recruitment packages" are rising, and an annual salary of 300,000 yuan is regarded as the "starting price" range, which is often reprinted and discussed in media articles and industry communities.

The "cold" end: Small and medium - sized enterprises cannot recruit, and ordinary graduates are exhausted from applying. Job - hunting platforms and research reports show that there is a recruitment boom for AI - related positions and a structural gap, but the supply side shows a "stratification". Enterprises hope that candidates have "directly applicable engineering and implementation experience" (data governance, training/evaluation and MLOps, inference and service - oriented, A/B and observation systems), which is exactly the shortcoming of most students. Macroeconomic employment data also shows a structural contradiction: the co - existence of pressure on the total volume and a structural gap, resulting in "difficulty in mass - applying and difficulty in precise matching".

What are the key factors causing this disconnect?

Mismatch in the ability spectrum: Courses focus on theory and algorithms, while enterprises value the integrated engineering ability of "computing power - framework - data - evaluation - implementation" and the understanding of interdisciplinary scenarios. There is still a gap between the "project closed - loop" in universities and the "production - level closed - loop" in enterprises.

Computing power and data barriers: Small and medium - sized enterprises lack public computing power/high - quality corpora and evaluation systems, and it is difficult for them to carry out the complete link of "training - iteration - implementation". Naturally, they tend to poach a small number of "ready - to - fight" people.

Regional and platform siphoning: First - tier/strong second - tier cities and leading platforms have better computing power, data, scenarios, and salaries, and talents are further concentrated in a few platforms.

Cognitive bias in positions: Many job seekers think that "AI = algorithm research", but the increasing demand of enterprises comes from "gray - area positions" such as AI engineering, application products, data and evaluation, and security/governance.

Uneven distribution of evaluation and internship resources: High - level scientific research/competition/internship opportunities are concentrated in a few top - tier departments and cities, widening the "resume gap".

04

Will the "Matthew effect" intensify?

First, the conclusion is yes! In the short to medium term, the talent and resources in the AI industry are indeed showing a trend of concentration where the strong become stronger, and this concentration has its logic of efficiency and economies of scale (computing power, data, ecosystem, and brand synergy). However, if there is a lack of "inclusiveness" at the institutional and infrastructure levels, the side - effects at the social level will expand:

Uneven opportunities: High - quality positions, tutors, computing power, data, and internship opportunities are further concentrated in a few platforms/few cities.

Industrial polarization: Leading companies are running on the closed - loop of "model - platform - application", while long - tail enterprises are turning to "secondary packaging/low - code use", and the spill - over effect of high - end technology dividends is insufficient.

Education involution: Universities continue to expand enrollment and open new majors, but it is difficult to quickly make up for the "engineering - industrialization" link, resulting in a structural contradiction of "hot enrollment but difficult employment".

Regional differentiation: Differences in regional computing power/data/application ecosystems amplify the imbalance in talent flow.

Is there a solution? I think whether it can be solved immediately in the short term or not, all parties must pay attention:

Small and medium - sized enterprises: Lower the threshold and focus on applications.

Use open - source and platforms to reduce the cost of trial and error: Prioritize the use of mature domestic ecosystems for PoC and make small and fast progress. These platforms provide models, toolchains, and community examples, reducing the thresholds for human resources and computing power. Reconstruct positions: Split "model training/evaluation/implementation" into recruit - able engineering particles (data/evaluation/MLOps/application front - end and back - end/security governance), and introduce T - shaped talents with both "technology and business" skills. Co - build and share: Connect with university joint laboratories and local "computing power vouchers/public computing power centers", and use the method of "promoting research through competitions and bringing in positions through projects" to lock in potential students and cultivate school - enterprise cooperation relationships of "you can use, I can use" in advance.

Universities and departments: Make up for the "last mile of engineering".

Integrate courses and engineering: On the basis of "mathematics and physics + algorithms", strengthen data engineering, evaluation and alignment, service - orientation and A/B, observation and security. Courses should be driven by real enterprise data and scenarios. Link graduation requirements with actual combat: Use "reproducible engineering projects + standardized reports + open - source repositories" as proof of graduation ability, and promote the construction of an in - school full - chain practice platform of "model - data - evaluation - deployment". Moderately combine specialization and breadth: Promote the construction of cross - disciplinary degree programs of "AI + X" (science, engineering, agriculture, medicine, and liberal arts), and encourage joint training of doctoral students and engineering master's students with industries. Make information transparent: Publish an annual white paper on the enrollment scale of majors, employment destinations, and engineering ability requirements to avoid "only expanding enrollment without cultivating talents". (Refer to the Ministry of Education's action plan and the enrollment expansion trends of universities for policy basis and trends.)

Job seekers: From the "algorithm dream" to "real engineering skills".

Supplement skills around the "full - stack closed - loop": data → training → evaluation → service - orientation → monitoring/security. Having the ability to "implement and reproduce" in one link will give you a significant advantage. Focus on portfolios: Select 2 - 3 real and runnable projects (including README, environment, test data, and indicator reproduction scripts), which is more likely to impress employers than "stacking certificates". Understand the position spectrum: Algorithm research is just one part. AI engineering, AI products, evaluation and security, data and governance, and AIGC content and toolchains are all recruiting. Adopt regional and platform strategies: Don't blindly flock to the most competitive tracks. Reasonably evaluate cities, industry scenarios, and growth curves.

When the technological dividend is monopolized by a few players, "AI inclusiveness" becomes a false proposition. Perhaps, more urgent than "recruiting talents with high salaries" is to establish a "technological democratization" ecosystem that allows students from small towns, students from ordinary universities, and small and medium - sized enterprises to participate together.

Appendix: References and data sources

① The "Action Plan for AI Innovation in Higher Education Institutions" and the catalogue of undergraduate majors (setting up majors such as Artificial Intelligence) of the Ministry of Education.

② Universities and research platforms: Tsinghua AIR, the School of AI at Peking University, AI directions and courses at Shanghai Jiao Tong University, and interdisciplinary AI platforms at Zhejiang University.

③ Campus recruitment and position trends (taking Baidu's 2026 campus recruitment as an example, the public information and campus recruitment page where AI positions account for > 90%).

④ Industry reports and discussions on salary ranges: Summaries of news from media such as the Securities Times about "the annual salary for top AI campus recruitment starting at 300,000 yuan".

⑤ Talent supply and demand and structural contradictions: Talent flow reports from Liepin and employment data from Xinhua News Agency/Zhaopin.

⑥ Top universities' enrollment expansion in strategic disciplines and AI: Comprehensive reports from Reuters.

⑦ Open - source and platform ecosystems: Official portals of PaddlePaddle, ModelScope, and MindSpore.

This article is from the WeChat public account "Niaoge Notes" (ID: niaoge8), author: June, published by 36Kr with authorization.