Xpeng holds grand ambitions in the autonomous technology industry

Chinese automotive manufacturer Xpeng has reaffirmed its ambition to become an artificial intelligence (AI) company by allocating significant funds towards the development of autonomous driving technology. At the CVPR 2026 event in Denver, USA, Dr Xianming Liu, Head of Xpeng’s General Intelligence Centre, revealed that the company spends approximately 300 million yuan every month on training AI models. Annually, this investment amounts to roughly 427 million euros, a substantial figure considering Xpeng recorded deliveries of around 200,000 electric vehicles last year.

According to Dr Liu, these funds are dedicated to the development of the next-generation autonomous driving software known as VLA 2.0. This technology is a core component of Xpeng’s strategy to accelerate the implementation of AI-based vehicle systems in mass production. In the VLA 2.0 architecture, certain intermediate stages have been removed; Xpeng believes that using language as an intermediary adds unnecessary computational load and potential latency.

The company’s engineering team discovered that electric vehicles can generate approximately two billion visual tokens per second from their camera systems, whereas controlling steering, acceleration, and braking requires only about 10 to 20 command tokens. Consequently, VLA 2.0 is designed to process visual data directly without the need to first convert it into linguistic descriptions. This approach is believed to accelerate the vehicle’s decision-making capabilities on the road. Xpeng’s move reflects a new trend in the global automotive industry, where software development and artificial intelligence are becoming primary competitive factors, moving beyond a sole reliance on hardware innovation.