Chinese electric vehicle manufacturer Xpeng has officially commenced mass production of its first purpose-built robotaxi, marking a significant escalation in the race to deploy fully autonomous vehicles on public roads. The rollout, which took place at the company’s manufacturing facility in Guangzhou on May 18, represents the first time a Chinese automaker has achieved mass production of a robotaxi using entirely in-house, full-stack development.
The deployment positions Xpeng as a direct challenger to Tesla’s Full Self-Driving (FSD) ambitions in the Chinese market, shifting the competitive landscape from software benchmarks to physical hardware deployment. The newly unveiled vehicle is built on the company’s GX platform and is engineered specifically to meet Level 4 (L4) autonomous driving standards, which allow a vehicle to navigate designated areas without any human intervention.
Turing AI Chips Power a Pure Vision Architecture
At the core of Xpeng’s new robotaxi is a fundamental shift in hardware strategy. The vehicle operates entirely without LiDAR sensors or high-definition maps, relying instead on a pure vision solution. This approach mirrors Tesla’s philosophy but is executed through Xpeng’s proprietary hardware and software ecosystem.
The computational heavy lifting is handled by four of Xpeng’s self-developed Turing AI chips. Together, this silicon cluster delivers an effective on-board computing power of 3,000 TOPS (Tera Operations Per Second). This massive processing capability is required to run the company’s VLA 2.0 (Vision-Language-Action) end-to-end large model directly on the vehicle.
According to Xpeng’s official announcement, the VLA 2.0 architecture eliminates the traditional language-translation step found in older three-stage autonomous systems. By processing visual data directly into driving actions, the system compresses response latency to under 80 milliseconds. The company claims this end-to-end approach provides superior urban generalization capabilities, allowing the robotaxi to operate across different cities without requiring extensive local mapping data.
The Timeline to Fully Driverless Operations
The mass production milestone follows a series of regulatory approvals. In January 2026, Xpeng secured a road testing permit for intelligent connected vehicles in Guangzhou, allowing it to begin routine L4 public road testing. In March, the company formalized its ambitions by establishing a dedicated Robotaxi business unit to oversee product definition, research and development, and commercial operations.
According to the South China Morning Post, Xpeng plans to initiate pilot robotaxi operations in the second half of this year. These initial trials will focus on validating technical viability, assessing user acceptance, and refining the commercial business model. The ultimate goal is to achieve fully autonomous operations—without an on-site safety officer present in the vehicle—by early 2027.
To support the commercial rollout, Xpeng is opening its Robotaxi Software Development Kit (SDK) to external partners. The first global ecosystem partner is Amap, the mapping and navigation service owned by Alibaba Group. This partnership will allow users to hail Xpeng’s self-driving taxis directly through the Amap application, instantly providing the service with access to millions of active users.
(Related: China’s Robotaxi Sector Reaches Inflection Point as Goldman Projects Fleet to Triple)
The Competitive Threat to Tesla FSD
Xpeng’s aggressive timeline and pure vision approach place it on a collision course with Tesla, which has been working to secure regulatory approval to deploy its FSD software in China. While Tesla has focused on selling software upgrades to individual car owners, Xpeng is simultaneously attacking the consumer market and the commercial fleet sector.
The rivalry is explicitly acknowledged by Xpeng’s leadership. During a previous presentation, Xpeng Chairman and CEO He Xiaopeng cited internal testing data comparing the VLA system to Tesla’s FSD. According to that study, Tesla’s system required seven human interventions during a 54-minute test route, while the Xpeng vehicle required only one intervention over the same route completed in 49 minutes.
By controlling the entire stack, from the Turing AI chips to the VLA 2.0 software model and the physical vehicle manufacturing, Xpeng believes it can iterate faster than competitors who rely on third-party hardware or software. The robotaxi also shares its foundational AI model with Xpeng’s other physical AI projects, including its humanoid robot and flying car initiatives, creating a unified development ecosystem.
As the robotaxi sector transitions from technical validation to large-scale commercialization, the ability to manufacture purpose-built vehicles at scale will become a critical bottleneck. With the first units now rolling off the Guangzhou assembly line, Xpeng has demonstrated that it possesses the industrial capacity to match its software ambitions.