AI Industry Chain Recruitment Bids Farewell to "30 Million Yuan Talent Wars" Craze, Small and Medium Enterprises Still Struggle to Find Talent

Tuchong Creative/Provided Image Chart data source: Maimai, Core International

Reporter Zhang Shuxian from Securities Times

On the first day of the Year of the Horse, a well-known recruitment agency’s headhunter, Li Yun (pseudonym), received a hiring request from a small to medium AI industry company, requiring a core talent in large models, specifically asking for someone from a particular team at a major tech company.

When seeing AI industry showcases like humanoid robots on the Spring Festival Gala, Li Yun sensed a new wave of talent recruitment in AI was coming. “Demand for positions in AI algorithms, foundational infrastructure for large models, and applications of large models is surging, and the requirements are becoming more stringent,” Li Yun told Securities Times.

While recruitment demand remains high, companies are becoming more rational in their hiring standards.

Li Yun has been working as an AI talent headhunter for over two years. She told reporters that two years ago, the market was in a “frenzy,” with many companies not even clear about what kind of algorithm talent they needed or what specific problems their positions aimed to solve. As long as candidates had relevant backgrounds in algorithms or from prestigious universities, companies were willing to pay high salaries to attract them, even if job responsibilities were unclear. “Now, companies focus more on whether candidates can adapt to their specific business scenarios. They want to know what projects candidates have participated in, whether they have real project delivery capabilities, and even how much computing power was used in those projects, which are important evaluation criteria.”

Demand is soaring, but hiring standards are tightening

Since the beginning of this year, Li Yun has received several recruitment requests from clients, one of whom was particularly urgent, needing the candidate to start in April. Li Yun began working overtime to find suitable candidates, “I was communicating with candidates during the Spring Festival.”

Ma Yue, partner and chief talent officer of Shanghai Leap Star Intelligent Technology Co., Ltd. (“Leap Star”), told reporters that this year, Leap Star still has over a hundred hiring needs, including personnel for organizational iteration and upgrades, new positions arising from transitioning from research to product and commercialization, and talent reserves for breakthroughs in learning paradigms.

“According to real-time monitoring data from Core International’s data platform, the current recruitment demand in the AI industry chain remains strong,” said Zeng Cheng, CEO of Core International. As of now, over 35,000 companies nationwide have posted AI-related positions, with a total of more than 145,000 jobs.

A report from Maimai, a professional community platform, also shows that AI has become the most competitive talent track in this spring’s recruitment season. In the first two months of this year, the number of AI positions increased about 12 times year-on-year, and the share of AI jobs among all new economy positions jumped from 2.29% in the same period in 2025 to 26.23%.

Amid the surge in AI talent demand, there were reports of the “30 million hiring a CFO” in the embodied intelligence industry. Interviewing sources revealed that such reports are not typical industry practice. Even when they occur, they are limited to a few star companies and are often temporary.

“However, to some extent, this reflects the hotness of the AI industry and the acceleration of financing and listing processes. From the positions we handle, starting in the second half of 2025, high-end financial talents with deep understanding of the AI industry, successful IPO experience, and top-tier capital operation experience are indeed in high demand and will command very competitive salaries,” Zeng Cheng said.

Although the demand for talent in the AI industry chain remains vigorous, the frenzy of “just grabbing anyone related” from previous years has subsided, and the industry has entered a rational development stage from “imagination-driven” to “ability-to-implement.”

This shift began in the second half of 2025. Since then, the market has favored candidates with real project experience and tangible results. Zeng Cheng explained that, based on the current core development path of “model capability—engineering deployment—scene implementation—business transformation,” the standards for AI talent are becoming more stringent, with deeper technical skills, more refined job divisions, and closer alignment with practical applications.

In terms of compensation, Maimai’s report shows that the average monthly salary for AI positions is 60,738 RMB. The highest-paying roles are AI scientists/heads, with an average of 137,153 RMB per month. Positions like algorithm researchers and large model algorithm specialists also generally earn around 70,000 RMB per month.

This may only be the cash component. Li Yun told reporters that recruitment in the AI industry often adopts a diversified incentive model of “cash + options.” She has handled positions with a maximum cash salary of over 5 million RMB, mainly for senior experts in large model algorithms and embodied intelligence core R&D roles. “Overall, core positions like large models and algorithms still command high salaries, with annual cash compensation generally exceeding one million RMB.”

Tighter hiring standards have increased recruitment difficulty. Li Yun said that two years ago, many positions could be advanced through phone interviews, but now, more face-to-face communication is needed to understand both the company’s and candidate’s needs thoroughly for better matching. Recently, she plans to quickly communicate with the leaders of small and medium enterprises who specifically require personnel from certain teams at major tech firms. “Salaries for specific team members at big companies are above 3 million RMB, and the conditions offered by this company are hard to match.”

Sustained or structural demand

The shift from “hype” to “ability” in recruitment is seen as a correction for the entire industry. Zeng Cheng believes that the current AI industry chain recruitment ecology is moving from an early stage of high enthusiasm and strong emotion to a more rational and structured phase.

He predicts that this year, AI industry chain recruitment will continue its structural growth, remaining high but with a steady slowdown in growth rate, entering a “rational prosperity” stage. “Previously, demand was mainly concentrated in highly digitalized fields like internet and finance, but now it is rapidly penetrating manufacturing, energy, agriculture, healthcare, and other traditional industries. Digital and intelligent upgrades in each of these sectors will generate sustained and stable talent needs.” However, he added that in the future, not all AI positions will be hot; instead, the focus will be on those that are truly needed, with others cooling down naturally.

Ma Yue also noted that AI talent recruitment shows a clear “polarization”: on one hand, there is a scarcity of key technical talents like models and algorithms—“the smartest people and the most capital are gathering in this field, and top talent is fiercely contested”; on the other hand, demand for operational roles is similar to non-AI industries.

She observed that talents in large model breakthroughs, scene implementation, product definition, and commercialization are very scarce, with a gradual shift from technical to product and business-oriented roles, and eventually to engineering, product, and commercialization expansion.

In fact, salary increases are not uniform across the AI industry chain. The most significant increases are concentrated in scarce tracks and core positions. “High-quality talent switching jobs typically see salary increases of 20%–30%, with key technical and leadership roles showing even greater flexibility,” Zeng Cheng analyzed. Premium salaries mainly cluster in three categories: first, multi-modal and embodied intelligence roles, especially those with combined algorithm, system, and control skills—such as senior large model algorithm experts earning between 1 million and 2 million RMB, and senior multi-modal algorithm engineers earning 500,000 to 900,000 RMB; second, model engineering and large-scale deployment roles—those capable of translating laboratory models into real-world, stable operations—these are in high demand with significant salary growth; third, “technical + industry + product” hybrid roles like AI product managers and solution architects, who need to understand technology, industry, and business needs, with salaries rising steadily—for example, senior AI product managers earning 800,000 to 1 million RMB annually.

“Previously, talent recruitment relied on a uniform salary system, but in the AI era, compensation is more personalized, emphasizing ‘one person, one policy,’ especially for key positions where individual super talents play a crucial role,” Ma Yue said.

SMEs face “difficult to find a key person”

While rationality is increasing, there are still many urgent issues in the AI recruitment ecosystem that hinder healthy and orderly industry development. One major concern is the over-concentration of high-end talent and the difficulty for SMEs to find key personnel.

“The company urgently needs embodied intelligence model engineers and algorithm engineers, but these talents, if influential in the industry, command annual salaries over one million RMB, with C-levels even reaching tens of millions,” said Wang Lei, chairman of Shanghai Qingbao Engine Robotics. “Most high salaries are offered by large firms; startups can’t afford that, with the highest salaries only around 700,000–800,000 RMB.”

Therefore, Wang Lei has to personally recruit talent, closely scouting at Tsinghua University. “Besides cash salaries, we also offer equity incentives for core R&D talents, ranging from tens of thousands to several million shares. We have already signed some equity subscription agreements.”

A second concern is the preference of companies for “plug-and-play” mature talents, which severely limits the growth space for junior talents. Zeng Cheng noted that many companies prefer to hire senior talents with over 8 years of experience and invest less in 1–3-year junior talents. Some companies lack systematic talent development systems; after hiring, they cannot provide suitable growth platforms, leading to high turnover. Without long-term, structured training mechanisms, future talent gaps may emerge.

“High salaries for talent poaching do exist, like Meta recruiting people at high prices. Domestically, it’s not as extreme yet, but similar trends are emerging,” Ma Yue revealed. Some well-known companies are offering annual salaries over ten million RMB to attract talent.

Unregulated poaching not only raises operational costs but also hampers national AI strategies and damages industry long-term health. Recently, relevant departments convened HR leaders from AI-related companies to explicitly stop unregulated talent poaching.

A third concern is the rising short-term profit mentality, which risks resource misallocation. Zeng Cheng said some companies and individuals focus excessively on short-term salary increases, neglecting long-term capability building and business value. When market conditions change, this can lead to high costs and low output.

Wang Lei disclosed that last year, a colleague left the company but quickly resigned again at a new company, indicating that matching talent with the right company may be more important than just high salary.

Additionally, the disconnect between talent cultivation and market demand is a current challenge. Wang Lei has long been aware of this issue; as early as 2022, he suggested universities train humanoid robot talents. “Currently, university training for AI talents is still lagging behind, unable to meet the rapid development needs of enterprises,” Wang Lei said. Companies need to cultivate talents internally, tailoring training to their business directions.

Zeng Cheng recommended that, at the policy level, efforts should be made to strengthen industry-university-research collaboration, support universities, enterprises, and training institutions in building an AI talent ecosystem, and shorten training cycles. The industry should establish more open talent mobility mechanisms, encouraging talent flow from large firms to SMEs and traditional industries through talent sharing and consulting models, to broadly empower the real economy with AI capabilities. Companies should shift from “poaching” to “training + utilizing,” by flexibly engaging independent consultants for scarce skills and increasing internal training investments to develop “AI + business” hybrid talents. They should also define roles based on real business problems, maintain rational hiring, and improve talent retention and development systems.