Die Chinesen tragen die Hälfte des Welt-AI-Bereichs. Mark Zuckerberg gibt Milliarden aus, um chinesische AI-Experten zu rekrutieren.
Waking up from sleep, all the top AI talents in the entire Silicon Valley circle have been poached by Zuckerberg. Altman of OpenAI and Cook of Apple, who have had their "home stolen," are on pins and needles.
According to incomplete statistics, recently Meta has poached at least 14 core researchers from OpenAI, among whom 8 are of Chinese descent, with a signing bonus of up to $100 million. Subsequently, Meta poached Ruoming Pang, a core AI executive at Apple, with a compensation package of over $200 million (equivalent to approximately 1.436 billion yuan), and these 9 Chinese talents will all be included in the "super intelligent dream team" led by Zuckerberg.
OpenAI couldn't sit still. "Someone broke into our house and stole things." Mark Chen, the chief research officer of OpenAI, described the poaching behavior in this way, but they also had to respond. So, OpenAI snatched four top talents from Meta, Tesla, and xAI.
Apple didn't follow suit because $200 million far exceeds the compensation level of Apple executives. Cook's publicly disclosed compensation in 2024 was $74.6 million. Apple's sense of crisis reached its peak. Eddy Cue, the head of Apple Services, issued an internal warning: "If Apple can't quickly adapt to the AI era, it may become the next BlackBerry or Nokia."
Behind this so - far most intense "talent scramble" war, the rise of Chinese AI talents has become a new focus.
Jensen Huang, the CEO of NVIDIA, said: "50% of the world's AI researchers are Chinese. You can't stop them from advancing AI development."
The value of this statement is being verified again and again.
Talent acquisition becomes a new "recruitment" model for big companies
In this talent war, a post - 95s is stirring up the US AI circle. Meta poached Yu Jiahui from OpenAI for $100 million. You know, Real Madrid once spent $80 million to poach Cristiano Ronaldo from Manchester United. The AI circle has surpassed the football circle in terms of sky - high "transfer fees."
However, this is no longer surprising. One "talent acquisition" drama after another has been staged in succession.
Meta spent $14.3 billion (equivalent to approximately 102.668 billion yuan) to acquire a 49% non - voting stake in Scale AI founded by Alexandr Wang. The real purpose behind this is to get Alexandr Wang to bring his core team to join Meta's super intelligent team. This is considered the most shocking "talent acquisition case."
This 28 - year - old Chinese genius has repeated the story of successful drop - out entrepreneurship. He is now appointed as Meta's chief AI officer, and Yann LeCun, a Turing Award winner and Meta's chief AI scientist, may report to him.
At the age of 19, Alexandr Wang founded Scale AI. The core business of this company is to provide data annotation solutions for large - model manufacturers such as Microsoft and OpenAI. Its revenue last year reached $870 million (equivalent to approximately 6.252 billion yuan), and it is expected that this year's revenue will soar to over $2 billion (equivalent to approximately 14.372 billion yuan). Judging from Scale AI's revenue scale, Meta's talent acquisition case is not a bad deal and will recoup the cost in just a few years.
It's not just Meta spending 102.7 billion yuan to "recruit" startup companies. US big companies like Google are also aggressively acquiring talents.
Just two days after Google acquired the core staff of AI coding startup Windsurf for $2.4 billion, on July 14 local time, another AI programming startup, Cognition, announced the acquisition of the remaining team and technical assets of Windsurf.
Google didn't acquire Windsurf in its entirety. Instead, it recruited its CEO Varun Mohan, co - founder Douglas Chen, and some key researchers, and obtained technology licenses. "This method is called'reverse acqui - hire.' This poaching model is a new product of the AI era, aiming to avoid regulatory obstacles caused by large - scale acquisitions and quickly recruit AI talents at sky - high prices."
"Big companies are just trying to poach one - in - a - thousand or one - in - ten - thousand top AI talents. They don't even want the rest of the business and the people." In the view of Yao Xin, the CEO of PPIO, talent mergers and acquisitions often cost hundreds of millions or billions of dollars, so it's not surprising to offer tens of millions of dollars in annual salaries to AI bigwigs.
Similar to the Internet era, the winner - takes - all rule still holds in the AI era. Looking around, in the search engine market, we only see Google, and in the cloud computing market, we only see a few leading manufacturers such as Amazon. Zuckerberg launched this talent war because Meta has no way back today.
After the setback of Llama 4, "in the eyes of Silicon Valley, currently, the top three in the large - model field are OpenAI, Google, and Anthropic. From Meta's heavy investment in poaching talents this time, we can see that Zuckerberg is in a difficult situation. Only by having the top - notch talent team can he turn the tables." Yao Xin believes.
Chinese talents support "half of the sky" in the global AI field
Twelve years ago, Yao Xin was also an eyewitness to a big - company acquisition. The Internet video platform PPTV he founded was controlled by Suning Commerce and Hony Capital for $420 million.
After that, he transformed from an entrepreneur into an investor. In the venture - capital circle over the years, he deeply understands the importance of talents. He even often waits at the gates of top US universities to recruit talents first.
He recalled that in 2023, when generative AI just emerged, Chinese AI talents were already in high demand, and it was getting more and more difficult to poach them. "Chinese people have a good foundation in mathematics and physics, are willing to work hard, and the results are obvious. Only Chinese people can work so hard."
When Elon Musk released the so - called "world's most powerful model" Grok 4, the public's attention was focused on a few photos. One photo exposed the behind - the - scenes team of Grok 4, and more than 70% of the team members had Asian faces. Another photo showed that before the release of Grok 4, Musk led the team to sleep in the office, and they often had meetings at 2 am. So, there were tents all over the xAI office.
"If you're good at math, you can go anywhere in the world without fear, and Chinese people are even less afraid." An 80s Chinese engineer, An Chao (pseudonym), who has worked in Silicon Valley for nearly 18 years, also believes that Chinese people can stand out in the US AI circle mainly because they have a strong foundation in mathematics and physics, and mathematics is the foundation of large AI models and other theoretical research.
According to the "Global Artificial Intelligence Talent Tracking Survey Report 2.0" released by MacroPolo, from 2019 to 2022, the proportion of top AI researchers from China increased from 29% to 47%.
"The most expensive thing in the US now is Chinese AI talents." A Chinese engineer, Yao Hua (pseudonym), who has worked at Amazon in the US for many years, said that the AI talent war initiated by Meta has pushed up the overall AI talent compensation level in Silicon Valley. The annual salary of top AI talents even exceeds the Series A financing amount of some startups.
Top AI talents are starting to differentiate
In July last year, the Chinese AI startup circle was surrounded by a group of academic elites, presenting a pattern of "the Tsinghua clique in the north and the Jiaotong clique in the south." According to incomplete statistics, there are already as many as 40 AI company founders with a Tsinghua background. If we say that "the Tsinghua clique supports half of the Chinese AI startup circle, then one year later, Chinese talents support 'half of the sky' in the global AI field."
The Chinese AI talents poached by Meta from OpenAI all graduated from top Chinese universities, such as Tsinghua University, Shanghai Jiao Tong University, Peking University, and the University of Science and Technology of China.
Moreover, many Chinese bigwigs in the US AI circle "come from the same school." After Apple's AI bigwig Ruoming Pang joined Meta's super intelligent department, he will partner with Yu Jiahui, the former head of OpenAI's perception team. Both of them are students of Wu Yonghui, who is a former Google Fellow and now the person - in - charge of ByteDance's Seed model.
Yao Xin, who returned to China from Silicon Valley in 2018, transformed from an investor into an entrepreneur, and founded "the Didi in the computing - power field" PPIO. He travels between Silicon Valley and Shanghai regularly every year, spending a lot of time contacting top AI talents and learning about the latest AI trends and technologies in the US.
Recently, he has clearly felt the changes in the atmosphere of the Chinese and US AI startup circles: Chinese companies rely on the human - wave tactic and the spirit of hard work, while US companies rely on talent quality and the spirit of in - depth study; China has strong hardware strength, and embodied intelligence has become a hot star, while the US has strong software strength, and AI Agent is the hottest trend. Many US companies have started using AI Agents to replace human employees, and the replacement rate in positions such as programming, sales, and operations is even as high as 70%. Many vertical - field AI Agent companies have started to make money, and the business models of Agent development companies in vertical industries such as programming, healthcare, and law have been verified by the market, with annual revenues exceeding $100 million.
"In the US, big companies like Apple can hardly recruit top AI talents." A Silicon Valley headhunter mentioned that for top AI talents, it's better to start their own businesses than to join a traditional big company with multiple levels, low efficiency, and difficult innovation.
Photos / Meta xAI Internet
This article is from the WeChat public account "IT Times" (ID: vittimes), author: Sun Yan, published by 36Kr with permission.