Fierce Midfield Battle for AI Talent: Domestic Shortage Reaches 5 Million While Overseas Giants Offer Multimillion-Dollar Salaries

"Currently, internship salaries for core AI technical roles at major tech firms have reached 5,000 yuan per day, which translates to more than 100,000 yuan per month," revealed an industry recruiter.

As the AI wave continues to surge, the competition for talent has only intensified. "Last year, the talent supply-demand ratio in AI was one-to-one; in the first half of this year, four candidates are competing for three positions," said Lin Fan, founder and CEO of Maimai. Lin noted that the number of applications for AI roles has increased by 30% year-on-year, with a talent gap of approximately 5 million nationwide, particularly in technical roles.

Following the emergence of DeepSeek, which disrupted the AI landscape, leading startups such as Zhipu, MiniMax (Xiyu Technology), Moonshot AI, and StepStone are now facing even fiercer competition, where talent acquisition is the key battleground.

The AI talent war has expanded beyond domestic borders into Silicon Valley, where Chinese talent is a prime focus for major tech firms. NVIDIA CEO Jensen Huang, who recently visited China, remarked, "Fifty percent of global AI researchers are Chinese, and there's no stopping them from advancing AI."

The high-paying allure of AI roles continues to attract attention. Salaries within the AI industry are stratified, with technical roles commanding top-tier compensation. Lin explained that, based on campus recruitment data, 32.45% of AI industry roles offer monthly salaries below 20,000 yuan, while only 18.17% exceed 50,000 yuan. Over 55% of salaries fall below 30,000 yuan.

According to Liepin, over 30% of AI technical positions offer annual salaries above 500,000 yuan, compared to less than 10% across all roles. "All in AI" has become the central theme for tech companies in the first half of the year. Unlike previous trends where talent shifted from tech giants to startups, this year large internet firms like ByteDance and Tencent have maintained a net inflow of AI talent.

Maimai data shows that 40% of employees at leading AI startups are exploring new opportunities, significantly higher than the 15% average in the broader internet industry. Since last year, at least 22 senior executives have departed from these startups, including 12 in the first half of this year alone. Some have returned to major tech firms, while others have launched their own ventures. Lin attributes this to the maturing of foundational model competition, leading some AI professionals to leave foundational model companies.

ByteDance stands out as the most aggressive recruiter among domestic firms, with a first-half talent inflow-to-outflow ratio of 1.71. There are reports that ByteDance offered an eight-figure annual salary to recruit a senior technical leader from a major tech company.

According to sources, although no longer CEO, ByteDance founder Zhang Yiming is deeply invested in the AI business and has shifted his focus to AI strategy and research. He frequently travels between Beijing and Singapore and has repeatedly emphasized the pursuit of Artificial General Intelligence (AGI) within the company.

"Large firms offer stable resources, clear business scenarios, and mature commercialization pathways, especially appealing to core AI talent," an industry insider explained. For top-tier professionals, major tech companies spare no expense. At ByteDance, annual salaries ranging from 30 million to 40 million yuan are not uncommon.

While tech giants provide ceiling-level salaries, startups tend to offer incentives through equity and profit-sharing. "Some startups may offer 1% equity, potentially worth tens of millions of dollars," said Lin. Many AI startup executives and HR professionals are actively recruiting on Maimai, with HR from Zhipu and MiniMax reportedly engaging in real-time.

Moreover, multiple industries are rapidly absorbing AI talent, including new energy vehicles, autonomous driving, robotics, e-commerce, internet, gaming, new lifestyle services, and companies like BYD, Li Auto, XPeng Motors, Horizon Robotics, QCraft, Taobao, and Pinduoduo. These sectors have seen consistent net inflows of AI talent and have identified concrete application scenarios.
For those working on foundational models, DeepSeek is seen as the disruptive force that reshaped the industry earlier this year.

"Last year, the foundational model space experienced an intense cash-burning and promotional race, but DeepSeek's breakout brought a more grounded perspective," said a source from a Shanghai-based foundational model company. "DeepSeek helped the relatively niche foundational model field break into the mainstream. Many previously unfamiliar with AI became more aware of the market through DeepSeek."

From a focus on advertising spend to technical prowess, AI companies are entering a maturity phase post-shakeup. For example, Shanghai-based AI unicorn MiniMax (Xiyu Technology) secured $600 million in Series A funding in March 2024, with a post-investment valuation of $2.5 billion. Alibaba led the round, with previous investors including Tencent, Future Capital, and miHoYo. Reports suggest MiniMax closed a $3 billion valuation funding round at the end of last year. In June, insiders confirmed MiniMax is preparing for a Hong Kong IPO.

Previously, MiniMax launched a "Tech Release Week," unveiling its first open-source large-scale hybrid architecture inference model MiniMax-M1, seen as a direct competitor to DeepSeek R1. Data indicates that during 80,000-token deep inference, M1 uses only 30% of DeepSeek R1's computational power; for 100,000-token generation, M1 needs just 25% of DeepSeek R1's resources.

Another startup, Zhipu, has also initiated its IPO process. On April 15, the website of the China Securities Regulatory Commission revealed that Beijing Zhipu Huazhang Technology Co., Ltd. (Zhipu) filed for IPO counseling registration with the Beijing Securities Regulatory Bureau, becoming the first foundational model startup to do so. On July 2, Zhipu announced a 1 billion yuan strategic investment from Shanghai state-owned capital, including Pudong Venture Capital Group and Zhangjiang Group, and recently completed the first installment. As a global tech hub, Silicon Valley is now a hotspot for AI talent competition, with Chinese professionals at the center of attention.

"China is not just another market—it is unique in its vitality, innovation, growth momentum, and industrial development speed," said NVIDIA CEO Jensen Huang in a July 21 interview with CCTV News.

Recently, AI scientists Banghua Zhu and Jiantao Jiao, both Tsinghua University alumni, announced they had joined NVIDIA. Zhu completed his undergraduate studies at Tsinghua University and obtained a Ph.D. in Electrical Engineering and Computer Science from the University of California, Berkeley. Jiao also graduated from Tsinghua, won the university’s top scholarship in 2011, and earned a Ph.D. in Electrical Engineering from Stanford University, specializing in information theory.

Meanwhile, Meta has recruited more than 14 core researchers from OpenAI, including eight Chinese scientists, offering a total bonus of $100 million. It was later revealed that Meta also brought on board former Apple AI executive Ruoming Pang with a similar compensation package. All nine Chinese scientists will join Meta's "superintelligence dream team" led by Mark Zuckerberg.
On June 25, Tesla CEO Elon Musk announced the official launch of Robotaxi on X, describing the event as "the pinnacle of a decade of hard work." In the team photo Musk shared, the prominent figure in the center is Duan Pengfei, a young Chinese engineer and key technical lead for Tesla Autopilot. Duan earned his undergraduate degree in Optoelectronic Information Science and Engineering from Wuhan University of Technology and his Ph.D. in Electrical Engineering from the University of Ohio.

Duan currently serves as Chief Software Engineer on Tesla's AI team, leading the Fleet Learning group. This team plays a vital role in Tesla's autonomous driving system by enabling data and experience sharing among vehicles, optimizing models through collaborative learning.

"China has the world’s largest AI engineering workforce and produces the highest volume of AI research papers," said Fang Han, CEO of Kunlun Tech, in an interview with The Paper. Fang added that China and the United States significantly outpace other nations in AI capabilities. With 50% of top AI talent being Chinese, the country's innovation strength and vitality are undeniable.
"As industry consolidation intensifies, China's foundational model sector is transitioning from a phase of frenzied spending to one of value-driven rationality," Fang said. "The interactions among technological innovation, commercial implementation, and capital markets will deepen. Every entrepreneurial boom is a game of musical chairs—once the number of chairs shrinks, the AI revolution will enter a new phase."