NVIDIA’s CEO Jensen Huang has recently made his third visit to China, where he praised the country’s performance in AI development during a media conference held in Beijing. Huang pointed out that China’s education system has produced some of the world’s top AI researchers, with around 50% of AI researchers based in China. He also commended the nation for its leading edge in AI modeling and engineering talent.
During this visit, Huang Renxun observed two key points: first, China’s supply chain is remarkably mature, with well-established infrastructure and a robust ecosystem; second, he believes that the interoperability of the global tech ecosystem has become a critical factor in industrial advancement. When discussing the CUDA-compatible platforms developed by Chinese manufacturers, he stated that this doesn’t affect his stance, even humorously mentioning If you are developing products compatible with CUDA, I totally don’t mind..
Jensen Huang also confirmed that the H20 chip has received re-approval for sale, and in the future, more products based on the Blackwell architecture will be introduced to the Chinese market. He emphasized that the memory bandwidth of the H20 chip is outstanding, making it particularly suitable for innovative applications like DeepSeek and Qianwen. However, he also reminded that the recovery of the supply chain takes time, with NVIDIA’s supply chain cycle extending up to 9 months.
When discussing China’s electric vehicle industry, Huang Renxun stated that it is Global Surprises, highlighting how Chinese car manufacturers like Xiaomi Motors, NIO, and Xpeng are reshaping the global competitive landscape. After meeting with Xiaomi founder Lei Jun, he expressed early foresight regarding Lei’s success and even indicated a willingness to purchase Xiaomi Motors, though it’s unfortunate that they haven’t yet launched in the U.S.
On the other hand, Jensen Huang revealed that he has gained comprehensive insights from using four personal AI tools: OpenAI, Gemini Pro, Claude, and Perplexity. He poses the same questions to these tools simultaneously to compare their responses.
Regarding the DeepSeek R1, Huang Renxun praised this revolutionary model, noting that it is the world’s first open-source inference model. He emphasized that openness allows participation from various industries and countries globally, asserting that AI should serve all humanity, not just a privileged few.
In the development of AGI, Huang Renxun believes that, based on current assessment standards, achieving it may not take too long and does not require a fundamental technological breakthrough. The current stage of AI development is divided into four phases: perception, generation, reasoning, and robotics, with the present being in the reasoning AI phase.
Addressing the issue of computing power anxiety, Huang Renxun pointed out that there is currently an evident shortage of computing power. He believes that a new era of computational revolution has ushered in a fresh phase of AI machine learning, and anticipates that this anxiety around computing power will not fade away for the next decade.
In the realm of robotics development, Jensen Huang expresses confidence in China’s related industries, noting that the country boasts top-tier AI technology reserves and robust mechatronic capabilities, along with significant advantages like the largest application scenarios and manufacturing base in the world. He emphasizes China’s unique strengths in technological innovation, which have birthed a range of innovative applications including WeChat, Xiaohongshu, and TikTok. He particularly mentions that mobile payment is nearly ubiquitous in China while he still relies on cash in the U.S., feeling like Antique.
Finally, regarding the attitude of the U.S. government, Huang Renxun revealed that he had communicated with Trump, who expressed pride in NVIDIA’s market value surpassing $400 billion, and while encouraging him to come to China, said, Make sure to have fun!.
