By Chen Jiahui and Zhao Lei
China’s race to dominate artificial intelligence is driving researcher pay packages into territory once reserved for entertainment celebrities and top athletes, as companies battle for a tiny pool of elite engineers capable of building world-leading large language models.
Just three years ago, leading AI researchers in China typically earned annual compensation packages worth about 1 million yuan ($147,000). By 2024, a small number were commanding tens of millions of yuan. Now, according to industry insiders, a handful are approaching — or even surpassing — the 100-million-yuan threshold.
The escalation reflects the extraordinary value placed on the few hundred researchers globally who are considered capable of materially advancing frontier AI models. Chinese technology groups including Tencent, ByteDance, Alibaba andDeepSeek are competing not only with one another, but also with Silicon Valley groups such as Meta and OpenAI for the same small talent pool.
The phenomenon has created a new class of highly mobile “AI stars”, whose movements between companies are closely watched by the industry. Recruiters, investors, and executives describe a market in which senior researchers are courted directly by company founders, offered unprecedented autonomy and compensation, and sometimes subjected to strict non-compete arrangements.
One widely discussed case involved researcher Yao Shunyu, who joined Tencent from OpenAI at the age of 27 to oversee the company’s LLM efforts, reportedly reporting directly to president Martin Lau. Tencent denied rumours that his package exceeded 100 million yuan annually, but the speculation itself highlighted how dramatically expectations around AI pay have shifted.
From academic career path to bidding war
Industry executives say the turning point came after the release of ChatGPT in late 2022, which triggered a surge of investment into generative AI in China. Hundreds of start-ups were launched, while established technology companies scrambled to build in-house model teams.
Initially, compensation remained broadly aligned with conventional big-tech salary structures. But that began to change in late 2023, when emerging players such as DeepSeek started aggressively targeting top researchers with higher cash compensation and greater research freedom.
ByteDance soon intensified the competition. The owner of TikTok dramatically expanded hiring for its Seed AI model division in 2024, recruiting heavily from rivals including Alibaba, Baidu and start-ups such as Moonshot AI and Zhipu AI. Recruiters said ByteDance routinely offered salary increases of 100% or more to attract experienced researchers.
One high-profile move came when Zhou Chang left Alibaba for ByteDance in the second half of 2024 on what industry insiders described as an eight-figure compensation package. Several members of his team reportedly followed him.
By 2025, the success of DeepSeek’s open-source models further intensified the competition. Companies increasingly concluded that a small number of exceptional researchers could generate outsized breakthroughs, leading to another wave of aggressive hiring.
Recruiters said compensation expectations were rising so rapidly that offers made in March 2025 were often no longer competitive by April. ByteDance’s TopSeed recruitment programme for elite graduates reportedly raised annual packages for new hires to between 3 million yuan and 5 million yuan, up from roughly 1.5 million yuan a year earlier. Some top students in 2026 recruitment rounds were reportedly offered as much as 6 million yuan.
Silicon Valley pushes salaries even higher
The global nature of the race has amplified the inflationary pressure.
Meta, led by Mark Zuckerberg, has emerged as one of the most aggressive recruiters worldwide. According to people familiar with the offers, the U.S. company paid huge packages to lure senior researchers from rivals including OpenAI and Apple. Reports of compensation worth hundreds of millions of dollars over multiple years have reverberated across the global AI sector.
Chinese groups have increasingly sought to compete directly for overseas Chinese talent. Tencent, in particular, has spent heavily poaching researchers from ByteDance and other rivals. Industry headhunters said the company targeted experienced staff whose compensation had stagnated relative to newly hired recruits, creating dissatisfaction inside some AI teams.
Several prominent researchers from ByteDance’s Seed division later appeared within Tencent’s Hunyuan AIorganisation, including specialists in reinforcement learning, infrastructure and visual AI.
Another landmark case was the recruitment of Guo Daya by ByteDance in 2026. Unlike Yao, whose appeal stemmed partly from his elite academic pedigree and OpenAI experience, Guo built his reputation through contributions to DeepSeek’s R1, Coder and Math models. His hiring reflected a shift in the market: practical achievements in training successful models are now valued as highly as experience at U.S. AI laboratories.
People familiar with the negotiations said Alibaba, Tencent and ByteDance all attempted to recruit Guo, with senior executives personally involved in discussions. ByteDance ultimately prevailed, partly because of its focus on AI coding agents, which reportedly aligned more closely with Guo’s research interests.
A handful of people, billions at stake
The enormous salaries reflect the economics of frontier AI development.
Training advanced models already costs billions of dollars in computing resources, while operating them requires vast spending on inference. Against that backdrop, companies increasingly believe that a single outstanding researcher can improve computing efficiency dramatically or accelerate model performance by months.
Industry executives estimate that only a few hundred people worldwide are genuinely capable of driving major advances in large language models. Many share similar backgrounds: mathematics or informatics competition winners who later attended elite universities such as Tsinghua University or Peking University before moving into top research laboratories.
The result is a talent market unlike anything China’s technology sector has previously experienced. For leading AI companies, the fear of falling behind has made time more valuable than money — and securing the right researchers worth almost any price
Source:
LatePost
