The intensifying competition in artificial intelligence (AI) is often framed around advanced chips, computing infrastructure, and advanced models. However, a new analysis suggests that talent—and the ability to attract and retain it—continues to be the foundation of AI leadership.
According to the latest Global AI Talent tracker by MacroPolo, China is the largest source of elite AI researchers globally, overtaking the United States in terms of where top talent is educated and trained. Despite producing more AI researchers, however, China continues to see a major share of its best talent move abroad, especially to the US.
The findings suggest that 38 per cent of the world’s elite AI researchers were educated in China, making it the largest origin point of top-tier AI talent. However, the tracker also claims that 72 per cent of the China-educated researchers are currently serving in the US.
The trend shows a paradox at the centre of the AI rivalry between the two nations. It also highlights that regardless of the source, the US continues to be the most preferred destination for many AI scholars. “It has always been about talent—the chip race and the model race are downstream of that,” said Kelly Forbes, global AI policy advisor, commenting on the report’s findings.
“What this data makes very clear is that the US lead in AI is not a story of American innovation in isolation. It’s a story of American institutions being extraordinarily good at attracting the world’s best minds,” Forbes told indianexpress.com.
Not just a technology advantage
According to Forbes, this dynamic has accorded the US a significant competitive advantage, but one that may be more fragile than many policymakers assume. “Thirty-eight per cent of top AI researchers were educated in China, and 72 per cent of those are now working in the United States. That’s not just a technology advantage. And that’s a much more fragile thing to build a national strategy around,” she said.
The findings surface at a time when both the US and China are investing hundreds of billions in AI infrastructure, talent development, and enhancing their respective semiconductor capabilities. Even though the debates around AI competition mostly focus on export controls and restrictions on advanced chips, Forbes argued that immigration and talent mobility may be just as important.
Asked about the impact of tighter immigration rules on the US AI ecosystem, Forbes said, “extremely vulnerable, and I doubt that’s well understood in the policy conversation.”
“When you look at the numbers—72 per cent of China-educated elite AI researchers are working in US institutions—you realise that any significant tightening of visa policy or student exchange programmes doesn’t just affect individuals, it degrades US AI capacity directly,” she added.
The data shows the crucial role international students and researchers play in sustaining America’s leadership in AI. For several decades, leading US universities and research institutions have attracted top talent from around the world, particularly from China and India. This has helped in creating a concentration of expertise that has fuelled many of the breakthroughs behind today’s AI systems.
Although China produces a larger share of elite AI researchers than any other country, according to the analysis, retaining them continues to be a challenge. MacroPolo’s findings show that only 11 per cent of the country’s top AI researchers currently work in China, down from 16 per cent in 2019.
According to Forbes, the numbers suggest that Beijing has yet to fully solve the retention problem despite its significant investments in AI research and innovation. “The retention number, which is 11 per cent, down from 16 per cent in 2019, tells you that despite enormous domestic investment, China hasn’t cracked this yet,” she said.
However, she cautioned against assuming the trend will continue indefinitely. “I wouldn’t be complacent. The early warning signs to watch are compensation parity at top Chinese labs closing the gap with US counterparts, whether researchers who did PhDs in the US start returning to China in larger numbers, and whether Chinese institutions start appearing more prominently at the top tier of conference publications from China-based affiliations rather than US ones.”
According to Forbes, current figures offer only a snapshot, while long-term trends may prove more important. “We are not there yet, but the trajectory matters as much as the current snapshot,” she said.
The tracker also takes into account the overall effectiveness of strategies that mainly focus on limiting access to advanced semiconductors. The US has imposed a series of export controls aimed at restricting China’s access to cutting-edge AI chips and semiconductor manufacturing technologies. However, the MacroPolo findings suggest that talent concentration may be a more enduring source of competitive advantage. “Both matter, but I think talent is the more durable advantage and the harder one to replicate quickly,” Forbes said.
“Chips you can redesign around, stockpile, or eventually produce domestically, and China is investing heavily in that. But the concentration of elite AI researchers, the research culture, and the ecosystem of labs and universities…that takes decades to build,” she added.
Forbes’s comments reflect a growing view among policymakers and researchers that AI leadership is determined not only by hardware and funding but also by the ability to attract, develop and retain highly skilled researchers.
Where does India stand?
The report also offers important lessons for India, which has long been recognised as a major source of global technology talent. According to MacroPolo, 10 per cent of elite AI researchers were educated in India. However, only 2 per cent currently work in the country.
Forbes described the findings as a warning sign for India. “India’s numbers in this data should be a wake-up call. Ten per cent of elite AI researchers were educated in India, but only 2 per cent are working there, a 20 per cent retention rate,” she said. “India is producing serious talent and then effectively gifting it to the US and others.”
According to the report, 462 top AI researchers were educated in India, of which 80 per cent moved to the US, 60 per cent to the UK, and 5 per cent to Canada. In India, top institutions producing these talents are IIT Bombay, IIT Delhi, and IISc Bangalore.
The report claims that after China, India is the second-largest producer of AI talent. However, both countries share similar patterns of US-bound migration.
As governments worldwide race to build domestic AI ecosystems, Forbes believes India has a unique opportunity to position itself as a major AI hub instead of simply being a supplier of talent to foreign companies and universities. “The opportunity is real, but it requires deliberate policy choices,” she said.
“India needs to move beyond celebrating the success of its diaspora in Silicon Valley and start asking harder questions about what it would take to keep or attract back that talent,” she added.
According to the policy expert, this will require investments in research infrastructure, competitive compensation and a stronger domestic AI industry capable of offering world-class opportunities. “The window is open, but it won’t stay open indefinitely. As China improves retention and the US potentially tightens access, the countries that have invested seriously in domestic AI ecosystems will be best positioned,” she said.
The same report claimed that China produced 1,756 researchers in 2024-25, the US 1,108, and Europe 416. China’s retention rate stands at 11 per cent, compared with 80 per cent in the US, 65 per cent in the UK, and 55 per cent in Germany. The report is based on data from NeurIPS 2024, ICML 2024, and ICLR 2025 conference proceedings.
Ultimately, the MacroPolo findings show that while chips, data centres, and advanced AI models remain critical battlegrounds, the real competition may be over the people building them.
A new report suggests that attracting and retaining top researchers may be a more decisive factor in the global AI race than chips, computing power, or infrastructure. (Image for representation/Magnific) 
Share
