The advancement of autonomous driving technology is poised for a significant leap, with projections suggesting 2028 could mark a crucial turning point for the large-scale deployment of Level 4 (L4) autonomous systems. This optimism is spearheaded by Momenta CEO Cao Xudong, who believes that the development of world models is accelerating an “intelligence driving Moore’s Law,” a trend he anticipates will culminate in widespread L4 adoption within the next few years.
The “Intelligence Driving Moore’s Law”
Cao Xudong has characterized the rapid progress in autonomous driving safety and performance by defining an “intelligence driving Moore’s Law.” This concept posits a tenfold improvement in key metrics every two years. He further predicts that leading companies in the field could achieve a tenfold increase in these metrics annually between 2026 and 2028.
The primary catalysts for this accelerated development, according to Cao Xudong, are the continuous expansion of world models, the sheer volume of data collected (Data Scaling), and the increasing complexity of the models themselves (Model Scaling). These factors are collectively driving the performance and capabilities of autonomous driving systems at an unprecedented rate.
Momenta’s Progress and Vision
Momenta, a key player in the autonomous driving space, has already achieved significant milestones. The company reports cumulative mass production deliveries exceeding 1.3 million units and has accumulated nearly 20 billion kilometers of driving data. Impressively, the time taken to deliver the latest 100,000 units has decreased to less than a month, down from approximately 24 months for the first 100,000 units, highlighting the accelerating pace of learning and experience accumulation.
Further bolstering Momenta’s capabilities is its affiliate company, XHEART. This venture focuses on designing chips specifically tailored for the demands of large-scale models. XHEART’s chips are engineered to deliver approximately a tenfold performance increase at equivalent cost and power consumption. The upcoming X9 chip is expected to support a hundredfold expansion of end-side model scale.
Synergy of Software and Hardware
The strategic collaboration between Momenta and XHEART exemplifies a powerful synergy between advanced software (large models) and specialized hardware (chips). By allowing model requirements to dictate chip design and, conversely, enabling chips to support model expansion, this soft-hard collaboration is a critical factor in realizing the “intelligence driving Moore’s Law.” Coupled with the ever-growing volume of mass production data and continuously evolving world models, this integrated approach is expected to accelerate the practical realization of L4 autonomous driving.
The convergence of these technological advancements, data accumulation, and strategic hardware development suggests that the timeline for widespread L4 autonomous driving adoption is rapidly approaching. The year 2028 is being eyed as a potential inflection point where the technology transitions from nascent development to large-scale, commercially viable deployment.
