Source: InvestDaily
Author: Feng Yuchen
Original Title丨Recent Graduates Offered Annual Salaries of Up to $8 Million
“I’ve been busy recruiting PhD graduates from the Class of 2027 at Tsinghua and Peking University,” said a headhunter friend.
Ph.D. graduates from top universities, specializing in AI, with published papers and competition achievements—this is likely the most valuable cohort of new graduates on the market.
How valuable is it? Recently, a post circulating on social media showed that NVIDIA offered a recent PhD graduate from Peking University a first-year pre-tax package of approximately 2.9 million yuan.
Indeed, the circle limits imagination.
The AI world is still too surreal. A headhunter actively recruiting talent revealed that major companies have already offered recent graduates packages of up to 8 million yuan, with one specific offer from JD.com standing out as astonishing. This scene alone reveals the intense demand from top companies for the brightest minds of this era.
Most expensive new graduate: Big tech companies compete for talent with offers of 8 million
An offer for an NVIDIA AI computing powerchip development position has been leaked:
Ph.D. in Integrated Circuits, Peking University, Class of 2026; first-year pre-tax compensation package of approximately 2.9 million RMB, including signing bonus and performance bonus; eligible for overseas assignment to the Silicon Valley headquarters after two years of employment.
As soon as this post was published, there was endless debate online about whether the offer was genuine or not. Looking back five years, the total compensation range for PhD campus recruits in NVIDIA China was generally between RMB 450,000 and 600,000; at that time, a Tsinghua University computer science PhD who joined a core technology role at NVIDIA China receiving an annual package of RMB 550,000 (excluding stock) was still considered normal, albeit on the higher end.
Today, the AI boom has pushed talent salaries to record highs.
Among China's major tech companies, the demand for recent PhD graduates from top universities is even greater. "Recent PhD graduates who interned at ByteDance's Seed program can expect annual packages of around 2 million RMB," multiple recruiters told us. It is no longer uncommon for recent PhDs from Tsinghua and Peking University to secure salaries in the millions.
For example, recent PhD graduates from the Interdisciplinary Information Sciences Institute at Tsinghua University specializing in large models can start with salaries of four to five million. It is well known that Tsinghua’s Interdisciplinary Information Sciences Institute, led by Turing Award winner Andrew Yao, is a research institution focused on computer science, AI, and quantum information. It includes the Tsinghua Honor Computer Science Program (commonly known as the "Yao Class"), and has produced leading experts such as Jiajun Wu in deep learning, Yao Shunyu, Chief AI Scientist at Tencent, Lou Tiancheng, co-founder of Pony.ai, Yin Qi, Chairman of StepFun, and Hu Yuanming, founder of Meshy AI.
Headhunter Amy from a major company shared the current salary range trends with us: for recent PhD graduates from top universities, the typical annual compensation package ranges from 2 to 3 million RMB; for preferred candidates, it starts at 5 to 6 million RMB and above; and for the very best, it reaches 8 to 9 million RMB and above.
It is reported that recent graduates specializing in large models from top 985 universities have received offers of 6 million yuan annually from Xiaohongshu, and some are master’s degree holders. Even more striking, JD.com has offered recent graduates salaries of up to 8 million yuan annually. “However, landing a salary at this level isn’t just about attending a prestigious university—it also requires publications in top conferences or journals, competition achievements, and strong research expertise,” explains headhunter Amy.
In embodied intelligence, talent is also very expensive.
Embodied industry headhunter John said: "Compensation ranges vary by direction; for example, embodied algorithm cerebellum roles offer annual packages around 1 million RMB, while those in world model directions typically exceed 1.5 million RMB." Earlier this year, Luo Tianqi, partner and CFO of StarMap, mentioned in a media group interview that StarMap has extended offers with total packages as high as 5 million RMB to several recent PhD graduates, directly competing with top AI giants for elite research talent.
The end of competition is talent
Crazy, the outside world can't help but marvel.
Is it worth offering salaries in the millions to recent graduates? First, consider this: building an artificial intelligence model can easily cost hundreds of millions of dollars, and training the next-generation model like GPT-5 could exceed $500 million. Meanwhile, a top-tier talent optimizing a single training strategy could save tens or even hundreds of millions in computing costs.
Previously, DeepSeek-V3’s technical report attributed training efficiency to the coordinated optimization of algorithms, frameworks, and hardware; the V4 technical report also revealed specific engineering optimizations that accelerated performance relative to the baseline—these examples illustrate how technical and engineering decisions can significantly impact the efficiency of compute investment.
Therefore, what the big companies are paying high salaries for is scarce technical judgment and the potential to stay competitive in the next round. Top-tier university PhDs are the most visible vehicles in this competition.
It is clear that this year, major companies have loudly announced their calls for AI talent.
ByteDance's 2026 "Top Seed" campus recruitment program for top talent in large models aims to recruit approximately 30 outstanding PhD students focused on cutting-edge artificial intelligence areas, including large language models, machine learning algorithms, and systems. ByteDance hopes this program will identify the top 5% of talent in the large model field to accomplish what 95% of people cannot.
Tencent has launched the "Qingyun Program" for top global tech students, offering customized training programs and access to core business roles, emphasizing exposure to forward-looking technical challenges alongside highly competitive compensation. Additionally, Alibaba, Kuaishou, and others have respectively rolled out the Alibaba Star Top Talent Program and the Kuaishou Star Top Technical Talent Program.
Among these major companies' talent programs, there are many third-year undergraduates who have not yet graduated, some even receiving salaries in the millions. In the Qingyun Program, the average annual salary for newly trained PhDs is 3 million.
The competition for AI talent is heating up, backed by significant investments in AI from each company.
Rumors in the market suggest that ByteDance’s 2026 AI investment budget is close to RMB 160 billion; Alibaba has announced it will invest RMB 380 billion in AI and cloud computing over three years, with additional capital expenditures.
On the Tencent side, the company has also invested over RMB 100 billion in AI infrastructure in a single quarter. On October 8, it was reported that Tencent Holdings is considering issuing offshore bonds of up to $5 billion, with the issuance potentially launching as early as this month. The final use of proceeds has not been determined, but the market generally believes the funds will primarily be used to increase investment in AI.
Hop on the big boat
At the height of its popularity, an untimely question was still raised: How long can high salaries last in this most capital-intensive sector?
Consider the supply side of talent. Previously, central authorities explicitly required that 80% of newly established doctoral programs be concentrated in scarce fields such as science, engineering, medicine, and agriculture, with artificial intelligence being one of the most critical areas. In 2026, universities across the country are increasing doctoral slots in artificial intelligence, with Tsinghua University, Chongqing University, Southeast University, Shandong University, and others actively expanding interdisciplinary AI programs.
This also means that the supply and demand curve for talent in the AI industry will certainly not remain stable over the coming years. There is a precedent for talent inflation in the internet era: as the mobile internet economy reached its peak, rapid business expansion and intensified homogenized competition led to a tsunami-like overexpansion of workforce resources.
On the other hand, when people ask whether high salaries in a certain industry will last, they are also concerned about whether the industry’s growth will continue to thrive.
Over the past two years of the booming large models, AI talent salaries have also seen their steepest increases. Historically, industry booms often begin with widespread optimism but ultimately end in differentiation—not just among companies, but also among the very people who built the industry.
Of course, for the top recent PhD graduates currently in the midst of this situation, since it's impossible to stop the tide, boarding the big ship and going with the flow may be the most practical choice.
As the saying goes on the internet: “In a person’s lifetime, there are only three to five opportunities to change one’s fate.” When an opportunity is right in front of you, why miss it?
Twitter: https://twitter.com/BitpushNewsCN
BitPush Telegram community: https://t.me/BitPushCommunity
BitPush TG subscription: https://t.me/bitpush