Nvidia CEO Jensen Huang says its US AI chips are around “60 times” faster than Chinese counterparts
Table of Contents
With the launch of the Chinese startup DeepSeek, Nvidia’s stock took a big hit as companies saw the potential for greater AI performance at significantly lower infrastructure costs. However, instead of viewing this as a negative, Nvidia was among the few companies that saw DeepSeek as an “excellent AI advancement” and stated that they don't see it as a negative because “inference requires significant numbers of NVIDIA GPUs.”
The company has also recently teamed up with DeepSeek to bring DeepSeek R1 optimizations to their Blackwell architecture, showing that while DeepSeek did initially introduce some challenges for Nvidia, they ultimately managed to work things out. However, DeepSeek wasn’t the only challenge the Green Team faced from China. Due to export controls, Nvidia has been restricted from doing business in the region, causing its revenue percentage from China to drop massively.
Revenue isn't the only thing that's impacted
According to Jensen Huang, the company's revenue percentage in China has fallen by about half due to export restrictions. On top of that, competitive pressure from companies like Huawei has made it difficult for Nvidia to dominate the AI market in the region. These export controls have not only impacted revenue but have also caused a significant gap in performance between AI chips in the U.S. and China.
“GB200 is probably somewhere along the lines of 60 times the token generation rate of what is being shipped in China that is currently export-controlled, and so the seperation of performance is quite high.”
Nvidia CEO, Jensen Huang. Source: CNBC
Huang also added that next-generation AI will need 100 times more computing power than older models due to new reasoning approaches that think “about how best to answer” questions step by step. We recently saw this with Elon Musk’s latest AI model Grok 3, which now “thinks” before responding, and all of this is thanks to over 100,000 Nvidia GPUs in the colossus supercomputer (this number is now being doubled). So, while the US market is pushing more powerful hardware to train AI models, the AI situation in China still seems to be quite behind in terms of performance.
About the Author