He Xiaopeng on Assisted Driving, Automotive Competition and Organizational Change

Zhang Xiaojun begins by asking He Xiaopeng about the Xiaomi SU7 launch. He recalls talking with other automaker founders there and seeing public interpretations of their seating and expressions.

He says founders can remain in conversation even while competing. Discussing Lei Jun’s car project, he places its investment and first product in the context of long-term industry competition.

He describes a software update at XPeng, presenting its end-to-end driving system and cabin language model as early steps. His account reflects the company’s plans and views at the time.

He traces that thinking to mobile internet: smartphones changed interactions and then the situations in which people used them. He sees a possible parallel for cars, while noting their safety, regulatory and adoption challenges are more complex.

He imagines that if future vehicles can safely handle more driving tasks, parking, cabin interaction and the passenger experience could change. This is a future scenario, not a claim that current systems drive without supervision.

Considering internet founders entering the auto industry, he says their different backgrounds shape choices about technology, operations and globalization.

He reflects on XPeng’s early market entry: with more experience in automotive branding, he might have chosen a different price range and product sequence. He notes he was initially an investor and the company already had a position when he joined.

On moving from his previous startup into carmaking, he says he wanted another difficult challenge and understood that the choice involved commitment and risk.

On whether assisted driving can help sell cars, he links customer value to capability, price and experience. His projected improvements and timetable are goals stated in the interview, not demonstrated outcomes.

He says driving and language models may draw on similar underlying methods, but driving has stricter safety requirements and cannot rely on model output alone; it also needs control and protective mechanisms.

He describes the team reading papers, studying model principles and exploring engineering applications. In his account, a driving system must also connect perception, localization, planning and control.

He distinguishes software demonstrations from hardware deployment: a technical improvement in one component still needs manufacturing, reliability, maintenance and scale before it becomes a sustainable product.

He says the team tried reorganizing data collection and training around model-based methods; early results could not simply be used in a product and required substantial further training and validation.

He sees different capabilities as necessary for general model research and industry applications. An automaker must understand models and combine them with data, vehicles and safety systems.

On research organizations, he emphasizes giving strong teams time and room to work, and says he increased direct discussions with employees about strategy and questions.

Asked whether XPeng should become a technology supplier amid competition from other manufacturers, He discusses his choice in terms of the relationship between vehicles, research spending and customer experience.

He argues model development can create financial and data barriers: research, training and fleet scale are connected, and smaller companies may struggle to fund continued investment.

Estimating the cost of replicating an advanced driving system, he stresses funding, talent, vehicles on the road and sustained execution. Heavy spending alone does not guarantee success.

On building an automaker, he describes the competitive pressure as “swimming in a sea of blood.” The metaphor expresses his experience, not a universal conclusion about every company.

He reflects on the G9 and organizational changes, saying management transformation involves processes and people and cannot be solved by a single announcement.

He discusses the division of work with president Wang Fengying, emphasizing collaboration across responsibilities rather than reducing it to a reporting hierarchy.

Facing intense competition, he says the company aims for a leading position while acknowledging that industry changes and partnerships will affect the outcome.

On product choices, he uses a card-playing metaphor: a company must time different models while considering resources, market demand and pace together.

He says he focuses on strategy and organizational management while asking teams to examine decisions through business performance, end-to-end customer experience and the bigger picture. In a long automotive value chain, local optimization may not improve the whole.