Credit Suisse quoted XPeng stating that robots and electric vehicles will share the same hardware to achieve cross-domain integration and scalability

XPeng-W (09868.HK) today (11th) saw its stock price continue to be strong in the afternoon, reaching a high of HKD 108.5, up 17.9% at one point, and the latest price reported at HKD 107.5, up 16.9%, with trading volume surging to HKD 7.6 billion. Jefferies issued a report stating that they recently added XPeng (XPEV.US) to their related workshops after the AI Day. XPeng's management shared deeper insights into the technology roadmap for VLA (Vision-Language-Action) model 2.0, Robotaxi, and humanoid robots. XPeng has begun operational planning for the launch of Robotaxi, including city selection in mainland China, site layout, and license applications. The mass production and commercial deployment target is set for the second half of 2026. To bridge the gap between L4 and L2 autonomous driving, XPeng focuses on specific scenario challenges, including passenger pick-up and drop-off, campus parking, emergency takeover, and road construction detection. At the same time, it aims to expand operational coverage to achieve flexible pick-up and drop-off in residential areas, office buildings, and parking garages, enhancing passenger comfort. The firm has given XPeng H shares a "Buy" rating with a target price of HKD 116.9.

Regarding humanoid robots, XPeng is developing full-sized humanoid robots to seamlessly integrate into human environments and accelerate the shift from hardware-driven to software-driven innovation. The company believes that humanoid robots are essential for achieving Artificial General Intelligence (AGI). The integration of hardware and software is a core strategy, such as a unique waist design that mimics the human spine rather than a rigid structure, significantly improving posture and movement. In motion control, the third-generation controller uses generative models and incorporates styles (such as catwalk), achieving natural walking and diverse movements, with a single model capable of completing multiple tasks. The fourth-generation controller will have stronger generative capabilities, supporting various styles and movements without the need to switch models. The company is actively exploring low-cost data acquisition methods to support continuous learning and iteration.

XPeng indicated that the humanoid robot is driven by the initially developed VLA model for autonomous driving. This model is trained using diverse real-world data, including road videos, hand movement videos, and 3D spatial video inputs. Notably, XPeng's robots and electric vehicles will share the same underlying hardware (such as chips), achieving deep integration and scalability across fields.

The company's goal is to develop a general-purpose humanoid robot with a wide range of capabilities. Once the core functions are established, an SDK interface will be opened to allow third-party developers to train and deploy exclusive skills. XPeng stated that the design of VLA 2.0 is more streamlined and straightforward, avoiding unnecessary system complexity, thereby enhancing scalability