The talent flows, the performance metrics, and the institutional investments all point in the same direction
For decades, the United States served as the gravitational center of global artificial intelligence research, drawing the world's brightest minds toward its universities and laboratories. Now, with Stanford's 2026 AI Index documenting an 89 percent decline in AI scholar migration to America since 2017, and with researchers like former Microsoft generative AI leader Li Hongzhi returning to institutions like Shanghai's Tongji University, the currents of talent and ambition are visibly reversing. The performance gap between American and Chinese AI systems has narrowed from 31.6 percentage points in 2023 to just 2.7 percent today — a compression that suggests not merely a catching-up, but a fundamental reordering of where the future of artificial intelligence is being built.
- Li Hongzhi, who spent over a decade leading generative AI at Microsoft AI Asia, has returned to China as a distinguished professor at Tongji University's new Institute of AI for Engineering — a high-profile signal of a deepening talent reversal.
- Stanford's 2026 AI Index reveals a staggering 89% drop in AI researchers migrating to the US since 2017, with 80% of that collapse happening in just the past twelve months.
- The US-China AI performance gap has shrunk from 31.6 percentage points in May 2023 to just 2.7% in 2026, suggesting Chinese AI capabilities are advancing at a pace few anticipated.
- Tongji's Institute of AI for Engineering is actively recruiting overseas Chinese researchers, reflecting a deliberate national strategy to repatriate talent and accelerate homegrown AI infrastructure.
- The convergence of talent outflows, narrowing technical gaps, and rising Chinese institutional investment points toward a potential end to the era of unchallenged American dominance in artificial intelligence.
Li Hongzhi spent more than a decade at Microsoft Research, eventually leading the generative AI group at Microsoft AI Asia. Last month, he returned to China to become a distinguished tenured professor at Tongji University in Shanghai, joining its newly established Institute of AI for Engineering. His credentials are formidable — a computer science degree from Zhejiang University, a PhD from Columbia, a first-place finish at ACM Multimedia 2013, and years of work at Microsoft's Redmond headquarters as a principal researcher and architect. In 2020, he co-authored a widely cited paper proposing a "double-head" neural network architecture for object detection.
But Li's homecoming is far from an isolated decision. Stanford University's 2026 AI Index, released in April, documents a dramatic and accelerating reversal in the flow of AI talent to the United States — an 89 percent decline since 2017, with 80 percent of that drop occurring in just the past year alone. Alongside this exodus, the technical gap between leading American and Chinese AI models has collapsed from 31.6 percentage points in May 2023 to just 2.7 percent today.
At Tongji, Li joins an institute led by Hua Xiansheng, whose own career traces a similar arc — 14 years at Microsoft, senior roles at Alibaba's DAMO Academy and Terminus Group, before arriving at Tongji in 2025. The institute is explicitly focused on applying foundation models and intelligent agents to real-world engineering problems, and the university has been deliberately recruiting overseas Chinese researchers as part of a broader national strategy.
What Li's move ultimately represents is not one researcher's career choice, but a convergence of signals — talent reversals, narrowing performance gaps, and rising institutional ambition — that together suggest the center of gravity in global AI research may be shifting in ways that will be difficult to reverse.
Li Hongzhi spent more than a decade at Microsoft Research, rising to lead the generative AI group at Microsoft AI Asia. Last month, he returned to China and accepted a position as a distinguished tenured professor at Tongji University in Shanghai, joining the school's newly established Institute of AI for Engineering. His departure marks a visible instance of a much larger shift: the United States is losing its grip on global AI talent.
Li's credentials run deep. He earned his bachelor's degree in computer science from Zhejiang University, then moved to Columbia University for graduate work. There, as part of a five-person team led by Brendan Jou, he won first place in the Grand Challenge category at ACM Multimedia 2013 in Barcelona for research on structured exploration of heterogeneous multimedia news sources. After completing his PhD, he joined Microsoft Research's headquarters in Redmond, Washington, where he held roles as principal researcher, principal architect, and principal applied science manager. His work centered on machine intelligence, multimodal content analysis, and cloud computing. In 2020, he co-authored a paper on object detection that proposed a "double-head" neural network architecture to separate classification from localization—a contribution that became widely cited in computer vision research.
But Li's move is not an isolated career choice. It arrives against the backdrop of a dramatic reversal in the flow of AI researchers to the United States. Stanford University's Institute for Human-Centered Artificial Intelligence released its 2026 AI Index in April, documenting what the report calls a "precipitous" decline in AI scholar migration northward. Since 2017, the number of AI researchers moving to the U.S. has fallen by 89 percent. More striking still: 80 percent of that decline has occurred in just the past year.
The numbers reflect a shifting competitive landscape. In May 2023, the performance gap between leading American and Chinese AI models stood at 31.6 percentage points. By 2026, that gap had narrowed to 2.7 percent. China's AI capabilities are closing in on American ones at a pace that would have seemed unlikely just three years ago. When talent stops flowing toward a country, and when the technical gap shrinks, the geopolitical implications ripple outward.
At Tongji, Li reports to Hua Xiansheng, the executive dean of the Institute of AI for Engineering. Hua's own trajectory mirrors a pattern: he spent roughly 14 years at Microsoft, working at Microsoft Research Asia starting in 2001, then moving through the Bing search engineering organization and Microsoft Research in the United States. In 2015, he joined Alibaba Group as a group vice president and head of DAMO Academy's City Brain Lab. He later became chief technology officer at Terminus Group, a Chinese AI infrastructure company, before taking the Tongji position in 2025. The institute itself is newly built, focused on what it calls "AI for Engineering"—adapting foundation models and intelligent agents for real-world engineering applications. The school has been actively recruiting overseas Chinese researchers, signaling an intentional strategy to build capacity.
What makes Li's move significant is not that one talented researcher chose to go home. It is that he is one of thousands no longer coming, and that the gap between American and Chinese AI performance is narrowing at a speed that suggests the era of American technological dominance in artificial intelligence may be entering a new phase. The talent flows, the performance metrics, and the institutional investments all point in the same direction: the center of gravity in AI research is shifting.
Notable Quotes
The number of AI scholars moving to the U.S. has fallen 89% since 2017, with 80% of that decline occurring in the past year alone.— Stanford University's 2026 AI Index
The Hearth Conversation Another angle on the story
Why does it matter that one researcher left Microsoft for a Chinese university?
On its own, it doesn't. But Li Hongzhi is one data point in a much larger pattern. Stanford's new report shows that 89% fewer AI researchers are moving to the U.S. than were a decade ago. When talent stops flowing toward a place, it usually means something has changed about that place's competitive position.
What has changed?
The performance gap between American and Chinese AI models has collapsed from 31.6% down to 2.7% in just three years. China's AI capabilities are catching up faster than anyone expected. Researchers follow opportunity and capability. If the best work is happening in China now, or if it's becoming competitive, the incentive to relocate to America weakens.
Is Li's move to Tongji a sign that China's AI is now better?
Not necessarily better—but competitive enough that a senior researcher with a decade at Microsoft sees more opportunity there than staying put. Tongji is building a new institute specifically for AI applications in engineering. They're recruiting aggressively. That kind of institutional investment, combined with narrowing technical gaps, creates pull.
What does this mean for American AI development?
It means the U.S. can no longer assume it will attract the world's best talent by default. If 80% of the decline in AI migration happened in just the past year, the trend is accelerating. That affects what kind of research gets done, where it gets done, and who controls the outcomes.
Could this reverse?
Possibly. But reversing a talent flow requires more than nostalgia. It requires competitive advantage—better funding, better institutions, clearer paths forward. Right now, the momentum is moving the other direction.