Jensen Huang's words were in vain.
On September 5th, according to multiple foreign media outlets such as Bloomberg and the Financial Times, Anthropic, the developer of Claude, one of the world's mainstream large models, issued an announcement stating that it would immediately cease providing services to "Chinese-controlled companies" due to "legal, regulatory, and security risks."
As soon as the news broke, it caught the widespread attention of the tech and media circles.
Anthropic was founded in 2021 by former OpenAI employees and is regarded as the most promising American artificial intelligence company after OpenAI.
However, this "strange move" by Anthropic undoubtedly exposes its narrow and naive perception.
Because in the end, the competition in AI is a competition for talent, not a competition for a certain technology. Jensen Huang, the founder of NVIDIA, who has suffered a lot from US bans, understands this best.
Jensen Huang has mentioned a shocking statistic in multiple interviews and public speeches: "Half of the world's AI researchers are of Chinese ethnicity." He uses this to wake up the United States.
AI Talent Is the Focus of Global Game
In the early summer of 2025, at the banquet hall of the Hilton Hotel in Washington, a summit called the Hill & Valley Forum was taking place. Politicians, business leaders, and tech giants gathered.
Jensen Huang, the founder of NVIDIA, took the stage wearing his iconic black leather jacket. When talking about the future of AI, he said, "Half of the world's AI researchers are of Chinese ethnicity."
He then emphasized that this reality must "become a core variable for us to re-examine the rules of this technological competition game."
Everyone knows that what Jensen Huang talked about is not just a simple statistic of AI talent. It also represents a complex issue of talent flow and cultural identity. This is one of the most core strategic anxieties of the United States in AI at present.
What's the background of this forum? The Hill & Valley Forum was established in 2023 and jointly initiated by the "US-China Economic and Security Review Commission" under the US Congress and several top venture capitalists in Silicon Valley. Its purpose is very clear: to specifically address the strategic challenges brought by China's technological rise and promote a closer cooperation mechanism between policymakers and the tech community. The forum is held in Washington every year, including closed-door roundtables and public summits. The topics discussed cover strategic directions such as national security, artificial intelligence, clean energy, and the reshoring of manufacturing. The guests almost include all the key figures in the US tech and political fields: members of Congress, senior officials of the Biden or Trump administrations, founders of Silicon Valley tech giants, and venture capital leaders. Some media believe that it has become one of the core platforms for the United States to formulate technological competition strategies.
It's no coincidence that Jensen Huang chose to emphasize the ethnic structure of AI talent in such an occasion where politics and technology intersect and the global perspective is very high.
What he implied is that AI talent is no longer just a matter of corporate recruitment. It has risen to the level of national strategic resources. The flow and identity of talent have become the focus of the game between countries.
This signal is very clear. Not long before the forum, the US government passed a series of policies aimed at restricting the flow of high-end technological talent, especially in the fields of AI and semiconductors. The US government strengthened the visa review of Chinese researchers and students who have connections with certain Chinese institutions or enterprises. This made it more difficult for them to obtain or renew F-1 student visas and H-1B work visas. It also prohibited or restricted some Chinese companies (such as Huawei and SMIC) that are considered a threat to national security from coming to the United States for technological cooperation and exchanges.
Chinese Talent Stands Out in the Global AI Talent Supply Pattern
To understand the profound meaning of this statistic, we must conduct an in-depth analysis of the global AI talent supply pattern.
According to the report "DeepSeek and the New Geopolitics of AI: China’s ascent to research pre-eminence in AI" released by Digital Science in 2024, Chinese AI talent is remarkable globally. The report points out that China's output of AI papers is approaching half of the global total, making it the engine of global AI research.
This is not only reflected in quantity but also in quality. According to the latest survey report "Country/Region AI Activity Metrics" by the ETO data analysis platform of George Washington University, from 2017 to 2022, in terms of the number of times papers were cited, the United States ranked first with 34,036 highly cited papers, followed closely by China with 29,229. Notably, in the rankings of academic and research institutions, the number of times AI papers of the Chinese Academy of Sciences and Tsinghua University were cited has surpassed that of world-renowned universities such as MIT and Stanford University, ranking first and third respectively.
In AI basic research, China is undoubtedly one of the global leaders. Chinese universities and research institutions are continuously cultivating AI talent with solid theoretical foundations and innovation potential. This large-scale production of talent provides a solid talent reserve for the global AI industry.
However, there is a problem we need to pay attention to. We have cultivated talent, but where do these talents go? Where do they ultimately create value?
Here, a significant geographical gap appears. Although many Chinese AI elites graduated from top Chinese universities, their breakthrough research results were achieved in the laboratories of Silicon Valley tech giants (such as OpenAI, Google DeepMind, and Meta). This phenomenon is called the geographical dislocation between "talent supply" and "result transformation."
For example, Fei-Fei Li, one of the most well-known Chinese scientists in the field of artificial intelligence, graduated from a Chinese university for her undergraduate studies, but her main achievements, the ImageNet project and the establishment of the Stanford AI Laboratory, were all completed in the United States. Another example is the top Chinese scientists at OpenAI. Their contributions directly promoted the birth of ChatGPT. These AI talents cultivated in China were ultimately transformed by the US innovation ecosystem.
Talent, Sovereignty, and Technological Hegemony
It's not hard to see that talent does not equal sovereignty.
Take OpenAI as an example. Behind the success of ChatGPT, Chinese scientists such as Lilian Weng and Chen Zeqing played a key role. Their research results provided important theoretical support for the model. However, the final ownership and commercial interests of these breakthrough research still firmly lie in the hands of US enterprises. The contribution of talent is huge, but it ultimately serves US capital and technological hegemony.
From a national perspective, the game between China and the United States is becoming more and more delicate.
The United States is worried about the outflow of talent. It is concerned that the intellectual property rights and innovation capabilities created by top talent in the United States may lead to technological "spillover" through talent return. This is directly related to US national security and technological monopoly status. This is why the US government has passed various policies in recent years to try to restrict high-end talent, especially Chinese scientists in the AI field, from having technological exchanges and cooperation with China. The US Department of Justice launched the "China Initiative" to investigate and prosecute Chinese scientists suspected of technology theft. All these are actually to cut off the possible leakage in the talent chain and ensure the exclusivity of technological achievements.
China is worried about the outsourcing of innovation. The top talent cultivated at great expense in China ultimately contributes to US technological hegemony. This is undoubtedly a loss of national strategic resources. China has made huge investments in talent cultivation but is facing the dilemma of talent loss and the inability to fully retain technological achievements. The outsourcing of talent's intellectual power allows China to have an advantage in the "upstream" of AI research but be in a passive position in the "downstream" commercial transformation and technological hegemony competition.
AI Geopolitics May Lead to the Differentiation of the Future AI Ecosystem
In the geopolitical game, enterprises cannot stay out of it either. The traditional dimensions of competition are changing. In the past, enterprises may only focus on who has the most powerful GPU and who has the strongest computing power. Now, the more core competition is: who can establish a global top AI team that can attract, retain, and collaborate efficiently.
So, if an enterprise relies too much on the talent supply of a single country or cannot freely allocate talent globally due to geopolitical factors, its long-term innovation ability will be severely restricted. This is the same for both China and the United States. For example, a US company that relies on Chinese students as its main research force may face a talent shortage due to tightened visa policies. Similarly, a multinational company that sets up a research and development center in China may also face the risk of data and technology spillover due to geopolitical tensions.
This also brings an open question: When geopolitics locks AI talent in different regions, will the future AI business models and technological development roadmaps be differentiated, forming different ecosystems?
In the near future, we may see that the AI models dominated by the United States will pay more attention to Western values and commercial applications. While the AI models dominated by China may be closer to local data and market needs.
This differentiation is not only technological but also cultural and strategic. It means the fragmentation of the global AI ecosystem.
For enterprises, future strategic layout will become extremely complex. They need to consider not only technology and the market but also geopolitical risks. Investment and research and development may need to be dispersed to different countries to establish a more flexible global network. For example, Microsoft and Google have set up AI research laboratories around the world to disperse geopolitical risks. Microsoft has established AI research laboratories in Montreal, Canada, Cambridge, UK, etc.; Google has established AI research centers in Zurich, Switzerland, Paris, France, Tel Aviv, Israel, etc.
On the other hand, some Chinese companies have also started to set up research and development centers overseas to attract global talent. Tencent established its Global AI Lab in 2017 and set up AI laboratories in Seattle, California, etc. Currently, it "gathers dozens of artificial intelligence scientists and 50 world-class AI doctors globally, focusing on machine learning, computer vision, and speech recognition." In the same year, Alibaba established the Alibaba DAMO Academy, a global research institute, focusing on cutting-edge fields such as machine intelligence and chip technology. It has set up research centers in eight cities around the world, gathering nearly 10 scientists at the IEEE Fellow level.
Summary
After this analysis, the final trend is very clear. In the future AI technology game, computing power (GPU) and funds will become more and more homogeneous over time. That is to say, the performance improvement of GPUs will slow down, and the investment boom will eventually become rational. What really makes a difference between countries or enterprises is the control of the talent chain. Whoever can attract, cultivate, and retain top talent will have an advantage in the long-term AI competition.
For a country, just "grabbing talent" is not enough. It's not just about treatment. More importantly, it is necessary to establish an environment where top talent can give full play to their value. This includes an open scientific research culture, an efficient transformation mechanism, and guaranteed intellectual property protection.
For enterprises, they need to have a global perspective and preferably establish a flexible organizational structure that can absorb and allocate talent globally. This requires enterprises to be more inclusive in culture and more flexible in management to build a real talent "moat."
The future technological war may not be a decisive battle of chips but a battle for talent. Whoever wins the talent can write the future rules of the game.
Data Sources
[1]Hook, Daniel (2025). Deepseek and the New Geopolitics of AI. figshare. Online resource. https://doi.org/10.6084/m9.figshare.29336588.v1
[2]Melot, J., Arnold, Z., Abdulla, S., & Chalal, H. (2024). Country AI Activity Metrics (1.3.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.14522783
This article is from the WeChat official account "Hundun University" (ID: hundun-university). The author is Hundun Academy. It is published by 36Kr with authorization.