XPENG is preparing for the 2026 Beijing Auto Show, opening April 24, with a presentation that goes well beyond new car models. The company has announced it will use the show to unveil what it calls its “Physical AI” ecosystem: a portfolio spanning intelligent driving, humanoid robotics, and aerial mobility, all built on a shared AI architecture. The framing is deliberate. XPENG is positioning itself not as a smart EV manufacturer that also does robotics, but as a physical AI company that happens to make cars.
The company has committed 7 billion yuan to physical AI research and development in 2026. That figure, disclosed alongside the auto show preview, is a significant signal of strategic intent for a company that has spent recent years fighting for market share in China’s intensely competitive EV market. The pivot to “physical AI” as an identity is also a competitive differentiation play: as Chinese EV makers converge on similar hardware and price points, the AI stack is increasingly where the battle is being fought. XPENG’s global footprint has grown substantially. The company now operates in over 60 countries and regions, having delivered 45,000 vehicles internationally in 2025, a 95.6 percent year-on-year increase.
VLA 2.0: Redefining Intelligent Driving
The centerpiece of XPENG’s auto show presentation will be the VLA 2.0 intelligent driving system. VLA, short for Vision-Language-Action, is the model architecture that powers XPENG’s advanced driver-assistance capabilities. The 2.0 iteration is designed to handle complex real-world driving scenarios with greater reliability than its predecessor, including autonomous campus roaming, fully automated parking, and advanced urban navigation without pre-mapped routes.
XPENG reports that user satisfaction scores for VLA 2.0 have risen, conversion rates from test drives to purchases have improved, and adoption of the Ultra version, the highest capability tier, is growing. To broaden access, the company recently launched the Mona UltraSE, which brings VLA 2.0 technology to a lower price point and a younger demographic. The data collection implications of wider deployment are significant: more vehicles on VLA 2.0 means more real-world driving data flowing back into model training, creating a flywheel that benefits both the driving system and the broader physical AI stack.
The Iron Robot and the Shared AI Architecture Thesis
XPENG’s Iron humanoid robot is targeted for mass production by the end of 2026. The company will present its latest development progress at the Beijing Auto Show. Iron is designed to leverage the same AI stack that powers XPENG’s vehicles — the company’s thesis being that the perception, planning, and action capabilities required for autonomous driving and for humanoid robotics are fundamentally similar problems, enabling shared model development and data infrastructure.
This shared-stack thesis is not unique to XPENG. Tesla has made similar arguments about the relationship between its Full Self-Driving system and the Optimus robot. However, XPENG is among the first Chinese companies to articulate it publicly and commit capital to it at scale. The Beijing Auto Show will serve as a platform to reinforce this international narrative, with media, partners, and consumers from across the world attending. For XPENG, the show is as much a statement about where the company is going as it is a product launch event. The company is also expected to showcase its aerial mobility vehicle at the show, completing the physical AI trifecta of ground vehicles, humanoid robots, and flying vehicles — all running on shared AI infrastructure.
The Beijing Auto Show as a Physical AI Showcase
The 2026 Beijing Auto Show, which opens April 24, is shaping up to be as much a physical AI exhibition as a traditional auto show. XPENG is not the only Chinese EV maker reframing itself around AI capabilities — BYD, Li Auto, and NIO are all expected to emphasize intelligent driving and AI integration in their presentations. The show provides a useful competitive benchmark: with dozens of companies simultaneously presenting their AI-driven systems, it becomes possible to assess relative progress in a way that individual press releases do not.
For XPENG, the stakes are high. The company has positioned the 7 billion yuan R&D commitment as evidence of long-term conviction, but investors will be watching for concrete product milestones — particularly on the Iron robot timeline and the VLA 2.0 adoption curve — that validate the physical AI thesis. The Beijing Auto Show is the first major public opportunity for XPENG to demonstrate that its physical AI ambitions are backed by tangible progress, and the company’s presentation will be scrutinized accordingly.
The broader significance of XPENG’s framing extends beyond the company itself. If a Chinese EV maker can credibly position itself as a physical AI company, with vehicles, robots, and aerial vehicles all running on shared AI infrastructure, it changes the competitive reference class from other EV makers to a broader set of technology platforms. That reframing, if it holds, has implications for how XPENG is valued, how it recruits talent, and how it negotiates partnerships. The Beijing Auto Show will be the first real test of whether the market accepts the new identity.