At the 2026 Physical AI and World Models Conference, Momenta founder and CEO Cao Xudong said world models are accelerating a "Moore's Law of autonomous driving," and predicted that the industry is likely to meet an important inflection point for scaled L4 deployment around 2028.
Cao argued that the combination of end-to-end architectures, large-scale real-world data and simulation powered by world models is compressing the timeline for safe, driverless operation. Momenta has been a leading supplier of driving-assistance systems to global joint ventures and Chinese brands, and its comments carry weight across the supply chain.
The 2028 thesis aligns with a wider industry shift from assisted driving toward higher levels of autonomy in defined operating domains such as robotaxis, ports and trunk logistics. Several Chinese robotaxi and trucking firms have already set similar commercialization horizons.
Whether the timeline holds will depend on regulatory acceptance, safety validation and the cost curve of the underlying compute and sensor stack.
Source: 2026 Physical AI and World Models Conference
