Can China compete in the AI talent war?
Half of DeepSeek’s team never left China for education or work, and those who did ultimately return to pursue AI development, say Stanford researchers
THE latest eye-watering artificial intelligence (AI) outlays are not going towards high-end chips or data centre build-outs, but individuals.
The competition for AI talent prompted Meta to reportedly offer sign-on bonuses of US$100 million to lure senior staff from rivals. It feels “as if someone has broken into our home and stolen something”, OpenAI’s chief research officer said of the aggressive poaching in a memo to staff obtained by Wired.
The latest victim: Apple, which just lost top executive Pang Ruoming to Meta.
It is telling that so many of the superstar players US tech titans are boasting about adding to their rosters are of Chinese origin. Including Pang, eight out of the 12 new recruits to the Meta Superintelligence Labs team graduated from universities in the mainland before pursuing careers abroad.
It means that a key driver of the global AI race is an intense scramble for the people building it: Chinese talent.
The outsize role they play in developing AI systems for its geopolitical rival likely is not lost on Beijing. In other tech fields where workers hold a knowledge advantage, the government has not been afraid of asking them to return home.
Authorities have already reportedly restricted travel for some of DeepSeek’s employees. Instead of cracking down on immigration, US policymakers must now do more to entice the best and brightest from China and beyond, and create an environment where they are likely to stay.
But American business leaders should not assume that the big pay cheques alone will win an international talent contest.
Researchers at Harvard University last month said the number of high-impact scientific publications showed that China dominates in “raw human capital for AI”. This helps drive indigenous research, despite US advantages in computing power and investment. Top workers may still be keen on making money overseas, but that does not mean a lot of them will not stay at home.
Separate researchers at Stanford University in May analysed data on the more than 200 authors listed on DeepSeek’s technical papers. The firm’s success story is “fundamentally, one of home-grown talent”, they found. Half of DeepSeek’s team never left China for education or work, and those who did ultimately returned to pursue AI development. This has policy implications for the US.
China looks at international experience less as a brain drain and more as a way for researchers to acquire knowledge before returning home, the Stanford paper said. The US “may be mistakenly assuming it has a permanent talent lead”.
It aligns with other data that suggests America has been losing its allure as a destination for top-tier AI researchers. Only 42 per cent of these individuals worked in the US in 2022, compared to 59 per cent in 2019. During that same period, China was closing the gap fast, rising to 28 per cent from 11 per cent.
The Chinese government, meanwhile, has been funding AI labs and research at universities as part of industrial policy. It is not clear how well this investment has paid off, but it has helped incubate talent who went on to support breakthroughs at private companies.
One of DeepSeek’s keystone papers, for example, was co-authored by scholars at Tsinghua University, Peking University and Nanjing University. In this way, China has been building up an ecosystem of innovation that does not centre around poaching individual star players.
Domestic firms are less able to spend so lavishly to attract top talent. US private investment in AI was nearly 12 times the amount in China, according to one analysis.
Earlier this year, state-backed news outlet the Global Times reported on the “high-paying job offers” from DeepSeek, which could amount to annual income of some 1.5 million yuan (S$267,200) per year. It is a significant sum in urban China, but hardly the instant millionaire-minting figures being tossed around in Silicon Valley.
DeepSeek is nonetheless in the middle of a recruiting blitz – one that is trying to attract overseas Chinese AI researchers to come back home.
It has posted a spate of roles on LinkedIn, a platform that is not used domestically. As my colleague Dave Lee has written, this is about more than just money, but instead convincing workers that their contribution “will matter most in the history books”. DeepSeek may be hoping that this pitch will work on homesick Chinese talent.
Ultimately, just under half of the world’s top-tier AI researchers come from China, compared to 18 per cent from the US. Many may be seeking opportunities abroad, but Beijing is pulling all its levers to convince at least some to stay, at a time when America is not signalling a warm welcome.
Mind-boggling sign-on bonuses from Silicon Valley may be enough to win a cross-border battle for talent, but time will tell if it is enough to win the war. BLOOMBERG