From Concept Cars to Level 3 Roadmaps
At Auto China 2026 in Beijing, autonomous driving stole the spotlight from traditional car launches. With a theme of “Future of Intelligence,” the show featured intelligent vehicles, including electric models approved for “hands-off, eyes-off” Level 3 autonomous driving on certain roads. Huawei self driving executive Qiu Chao argued that China’s autonomous driving industry is on the verge of a breakthrough, predicting that Level 3 autonomous driving will likely enter a real deployment era around 2027, once several hurdles are cleared. These include proving safety through mass real-world trials, clarifying legal liability as responsibility shifts from drivers to automakers and system suppliers, and giving the public time to accept cars that appear to drive themselves. The message from Beijing is clear: auto shows are no longer just about styling and horsepower, but about software, sensors and policy readiness for the next wave of automation.
What Level 3 Autonomy Really Changes for Drivers
Most advanced systems on the road today are Level 2, where the car can steer and control speed, but the human driver must constantly supervise and remain fully responsible. Level 3 autonomous driving, also called “conditionally automated driving,” goes a step further. In defined conditions, the vehicle can perform dynamic driving tasks on its own, allowing the driver to take hands off the wheel and eyes off the road, but still remain available to take back control when the system requests. For everyday motorists, that could mean stress-free cruising in traffic jams or on mapped highways, with the car handling most of the workload. However, as Huawei’s Qiu Chao emphasised, this also raises the bar for safety, demanding benchmarks far stricter than those applied to human drivers, along with clear rules for when the system, not the person, is at fault.
Inside Pony.ai’s NVIDIA-Powered Robotaxi Technology
While Huawei is talking timelines, Pony.ai is showing the hardware that could make large-scale robotaxi technology viable. The company has unveiled a next-generation autonomous driving domain controller built with NVIDIA, based on the DRIVE Hyperion architecture and powered by the DRIVE AGX Thor platform. This Pony ai NVIDIA platform is designed for Level 4 autonomy, where the system can handle all driving in specific areas without human intervention. The controller fuses data from multiple sensors, builds a detailed view of the environment, and interprets complex scenarios in real time. NVLink connects two DRIVE Thor system-on-chips, delivering up to 4,000 FP4 TFLOPS of computing power in multi-chip configurations, with built-in redundancy for safety. Pony.ai sees this as the backbone for scaling robotaxis, shuttles and other autonomous services, aiming to operate thousands of vehicles across more than 20 cities worldwide by the end of 2026.
Auto Shows Become Platforms for Autonomous Ecosystems
Events like Auto China 2026 illustrate how autonomous car shows are evolving into ecosystem showcases. The Beijing exhibition brought together global automakers, technology suppliers and component manufacturers across a record 380,000 square metres, with 1,451 vehicles on display. Beyond concept cars, visitors encountered production-ready models featuring advanced driver assistance, domain controllers from players like Pony.ai, and integrated software platforms. Announcements around Level 3 autonomous driving roadmaps and robotaxi-ready hardware now share the main stage with new vehicle launches. For tech companies, these events are a chance to prove that their systems can move from pilot projects to mass deployment. For automakers, partnering with firms such as Huawei and Pony.ai signals that future models will be as much about computing and connectivity as engines and batteries. The result is a new kind of mobility expo where code, chips and regulations matter as much as chassis design.
What It Could Mean for Markets Like Malaysia
Although these breakthroughs are centred in China and other early-adopter regions, they hint at what markets like Malaysia may soon face. As Level 3 and eventually Level 4 systems mature, regulators will need to define where and when autonomous functions are allowed, how to certify safety benchmarks, and how liability is shared between human drivers, car brands and software providers. Robotaxi technology tested abroad could later support autonomous shuttles or logistics services in Asian cities, especially in controlled environments like campuses and industrial parks. Yet consumer readiness will be just as important as legal frameworks. People will need time and clear communication to trust vehicles that can drive with minimal human oversight. Watching how Huawei self driving initiatives and Pony.ai deployments roll out in China, Europe and North Asia will offer useful lessons for policymakers and industry players across Southeast Asia.