About Charles Zhang
In September 2023, Charles Zhang discussed the launch of Xpeng's city-level Navigation Guided Pilot (NGP), which he described as capable of handling complex urban driving scenarios. He stated that this launch, combined with the company's existing highway NGP and valet parking assist, brought Xpeng "one step closer to the full autonomy."
Regarding U.S. restrictions on AI chip exports to China, Zhang said the export license requirement applied to chips used for deep learning training on the server and cloud side, not to chips used in the vehicles themselves. He stated that Xpeng had built a private cloud with Alibaba Cloud for deep learning training and that its computing power needs for the next two to three years were "basically secured." Zhang also noted that Xpeng had a long-term partnership with Nvidia and was launching the G9 model equipped with Nvidia's chip.
Source: AI-verified profile updated from Charles Zhang's recent appearances.
Browse all interviews →
Transcript (6 segments)
C
Charles Zhang0:00
Yeah, I think that was always a week. This weekend we launched the city-level NGP, which is able to handle the complex driving scenarios in the city. And this is actually the major milestone for our conductor roadmap. So last year we launched the highway NGP, which would allow us to cover the highway driving scenarios, and we also launched the valet parking assist, which allowed us to cover the underground parking lot driving scenarios. So I think having these technology capabilities to cover all these scenarios, I think we're one step closer to full autonomy.
I
Interviewer0:38
Charles, let's talk about the challenges and the geopolitics, if I may. If the U.S. is restricting the licensing and supply of AI chips to China, how much of an obstacle is that going to be for your AI program?
C
Charles Zhang0:59
So I think we heard that the U.S. government recently added one export license for one of the important AI chips for NVIDIA, which is designed to do the conductor deep learning training in the server and at the cloud side. So far, for us, we have built a private cloud for deep learning training purposes with Alibaba Cloud. Basically, our computing power required for deep learning training in the next two or three years is secured.
E
Emily Tan1:34
Charles, if I could just follow up on that, this is Emily in Hong Kong joining the conversation. I'm aware that your ADAS platform uses the NVIDIA chip now. With this directive by President Biden banning the supply of AI chips to Chinese companies for this such use, have you been in talks with NVIDIA and what is the backup? Would you be using chips from another company?
C
Charles Zhang1:59
So I think basically the export license required is for the chip on the server side for deep learning training, not on the car itself. So for the AI chip used in our EVs and other companies' EVs, it is for civilian use. So right now we don't see any risk of restriction. Obviously, we have a long-term partnership with NVIDIA. We have been one of the first users of NVIDIA's Orin, and on Wednesday we're launching our G9, which will be equipped with NVIDIA's Orin.
E
Emily Tan2:36
Hi, I'm Emily Tan and thanks for watching CNBC. You can check out more of our videos by clicking on the boxes on the screen, and don't forget to subscribe to the channel for more. Thanks for watching.
