Written by Hao Xin, Wu Xianzhi | Edited by Wang Pan

On October 9, it was reported that Baidu's autonomous driving service platform "Luobo Kuaipao" was recently brewing a global layout, having conducted in-depth communication with multiple international companies and planning to enter overseas markets.

Meanwhile, the Apollo autonomous driving open platform 10.0 is about to be released, equipped with Baidu's latest autonomous driving large model ADFM (Autonomous Driving Foundation Model). This upgrade will significantly enhance the safety, intelligence, and ease of use of the autonomous driving open platform.

At the same time, the global Robotaxi market is about to welcome a heavyweight player - Tesla officially announced earlier that it will launch its long-planned Robotaxi service on October 11, Beijing time.

This race between "local and foreign radishes" ignited enthusiasm in the secondary market. On the first trading day after the holiday, the self-driving sector, represented by suppliers related to Luobo Kuaipao, showed unusually active performance, with stocks such as Zhonghaida, Tianmai Technology, Jingwei Hengrun, and Xingwang Yuda hitting their daily limit in the A-share market. The underlying logic is that autonomous driving is a typical application of large models reconstructing the physical world, and the development of self-driving is the general trend both domestically and internationally. The market is optimistic about self-driving driving the common development of upstream and downstream industries in the industrial chain.

Overall, this year's Robotaxi market exhibits two notable characteristics. First, autonomous driving enterprises from both China and the United States, represented by Luobo Kuaipao and Waymo, have entered the commercialization stage. Second, under the global leadership of China and the United States in autonomous driving, two distinct driving forces have emerged.

Resource endowments have enabled Chinese and American giants to forge multiple paths in technology and implementation.

The United States' advantage in integrated circuits drives the "Waymo faction" and "Tesla faction" in perception routes. Domestically, under policy promotion and continuous investment in new infrastructure, Luobo Kuaipao's "standalone vehicle intelligence + vehicle-road collaboration" solution has been implemented for a long time.

Waymo, Tesla, and Luobo Kuaipao, autonomous driving companies from both China and the United States, have converged on the core goal of achieving "unmanned" autonomous driving, despite taking different paths.

China and the United States lead the way, with the "Big Three" taking shape

In early October, the "Big Three" in autonomous driving coincidentally made moves.

First, on October 2, Waymo, a subsidiary of Google's parent company Alphabet, announced that its driverless taxi service in Austin, Texas, would soon be open to the public. It is reported that the service will be rolled out in phases, initially available to select users through the Waymo One app within the year and transferred to Uber early next year. Meanwhile, rumors circulated that Waymo was partnering with Geely Group subsidiary Zeekr to develop new vehicles specifically designed for driverless operation. Recently, the two parties have been advancing the testing of the sixth-generation Waymo Driver in San Francisco.

A week later, news of Luobo Kuaipao's upcoming global layout emerged, followed by the imminent release of Tesla's Robotaxi, marking the initial emergence of a three-way competition among Luobo Kuaipao, Tesla, and Waymo in the autonomous driving arena. Behind this "three-legged stool" reflects the fundamentally different internal driving forces of China and the United States as runners.

The United States holds a first-mover advantage in autonomous driving technology research and development, with its developed and comprehensive integrated circuit technology, especially its long-standing leadership in high-end chip design, laying a solid foundation for the development of high-performance automotive chips. Both Waymo's standalone vehicle intelligence route and Tesla's no-high-precision map + pure vision solution (FSD) rely on the country's capabilities in artificial intelligence algorithms and decision-making chips. However, as revealed in leaked spy photos of Tesla's Robotaxi, netizens noticed a protrusion above the roof, suspectedly indicating the use of LiDAR again.

Waymo currently faces numerous challenges, including persistently high costs for standalone vehicle intelligence, which have burned through $5 billion over three years. In July, its parent company Alphabet provided an additional $5 billion, citing it as an "important model for long-term investment." Upon hearing this, Elon Musk joked, "Waymo money."

Beyond cost issues, Waymo relies solely on autonomous vehicle fleet data and annotated data for feedback, limiting its data sources. Moreover, the presence of safety zones within geofenced test areas further restricts its business expansion speed.

If Waymo prioritizes hardware over software, Tesla does the opposite. Tesla's FSD is essentially an end-to-end system that requires continuous feedback accumulation in real-world scenarios. If one car hits a curb today, all Teslas will know to avoid it tomorrow, making the system better with each use. The challenge lies in no one wanting to be the first to hit a curb, and the ever-changing real-world road conditions, making continuous learning costly.

Despite their respective biases, Andrej Karpathy, former head of Tesla's AI and Autopilot division, believes that Tesla's autonomous driving technology is superior, and that FSD and Robotaxi may form a mutually reinforcing relationship - relying on Tesla's vast installed base and the millions of miles of driving data generated daily under various conditions, creating a flywheel effect.

While China's autonomous driving started later, it has shown a trend of catching up quickly. In terms of industrial supporting facilities and policy support, China has provided solid backing for domestic enterprises to achieve leapfrog development.

According to statistics from the Ministry of Public Security, as of July 2024, the number of motor vehicles in China reached 440 million, including 345 million automobiles. Large-scale deployment is a crucial means for Robotaxi to reduce costs, and the vast market not only meets this requirement but also provides an abundant source of data feedback, facilitating the commercialization of autonomous driving in certain sectors.

Another advantage of being a late starter is the comprehensive new infrastructure, represented by 5G, satellite internet, data centers, and intelligent transportation. Compared to foreign countries, China places more emphasis on the synchronized development of intelligence and networking, using networking capabilities to build an integrated solution of "person-vehicle-road-cloud," significantly reducing the development difficulty of standalone vehicle intelligence.

Additionally, policy documents such as the "National Internet of Vehicles Industry Standard System Construction Guide (Intelligent and Connected Vehicles)" and the "Development Action Plan for the Internet of Vehicles (Intelligent and Connected Vehicles) Industry" provide guidance for relevant enterprises. Technically, significant progress has been made in recent years in areas such as high-precision maps, LiDAR for autonomous driving perception, and in-vehicle computing chips for control layers.

These conditions have given rise to a cohort of companies capable of rivaling American giants.

A previous research report by Huajin Securities mentioned that domestic Robotaxi enterprises can be broadly categorized into autonomous driving technology companies and traditional manufacturers. Currently, technology companies are ahead in progress. In terms of landing cities and test mileage, technology companies represented by Luobo Kuaipao, Wenyuan Zhixing, and Pony.ai are leading the way, while traditional manufacturers started road testing later but are actively seeking partnerships with autonomous driving solution providers to advance the commercialization of Robotaxi.

The Robotaxi Business Model Forms a "Loop"

For autonomous driving to navigate the myriad of urban thoroughfares from the lab, it must contend with financial realities. Technology and commercialization are like the two ends of a balance, constantly restraining the forward progress of autonomous driving.

This year marks a turning point. As costs continue to decrease, Robotaxi is gradually forming a closed-loop business model, regardless of the perception route or map-based route. There are two reference dimensions for exploring the revenue-cost relationship of Robotaxi: one is the scaling effect that dilutes various cost expenses, and the other is how to reduce vehicle manufacturing costs, safety operating costs, and transportation operating costs. At present, vehicle manufacturing costs clearly dominate.

This year, domestic users have become acutely aware of the presence of Robotaxi, which is related to its expanding coverage and increased number of deployed vehicles. As scale increases, the marginal cost of Robotaxi decreases accordingly.

Data shows that the Luobo Kuaipao platform has accumulated over 7 million orders. Since 2021, the platform has opened passenger testing in 11 cities nationwide, including Beijing, Shanghai, Guangzhou, Shenzhen, Chongqing, Wuhan, Chengdu, Changsha, Hefei, Yangquan, and Wuzhen.

According to Frost & Sullivan's forecast, Robotaxi will achieve large-scale deployment around 2026, and it is expected to be widely adopted globally by 2030, with a penetration rate of 31.8% in China's smart travel by then and 69.3% by 2035. It is projected that by 2030, the Robotaxi market size in China and globally will reach 488.8 billion yuan and 834.9 billion yuan, respectively.

Focusing on Robotaxi itself, there has been a notable decrease in the cost of its full life cycle operations (vehicle manufacturing, safe operation, and transportation operation). Currently, vehicle manufacturing costs are the primary driver pushing up unit service costs.

Since the current mainstream Robotaxi models rely on high-precision maps and LiDAR for perception, they must be equipped with a full set of high-value LiDARs. According to data from Ruqi Chuxing, LiDARs alone account for more than 50% of hardware costs, while high-performance computing chips for autonomous driving also elevate the overall vehicle manufacturing cost. In May of this year, the cost-reducing effect on the hardware side was evident, with Luobo Kuaipao's sixth-generation driverless vehicle Yichi 06 priced at only 204,600 yuan, a 60% reduction from the cost of the fifth-generation vehicle.

Overseas self-driving companies are also seeking cost reductions. It is rumored that Tesla's Robotaxi plans to adopt new automotive assembly technology to achieve cost savings and efficiency gains. While traditional production lines typically add components sequentially, Tesla plans to produce Robotaxi in a manner similar to assembling Legos.

The development curve of technology follows a cyclical pattern, often experiencing phases of emergence, maturity, and large-scale deployment, accompanied by cost savings and efficiency gains at the consumer end, repeating the cycle.

Large models are reshaping the physical world, with autonomous driving being a crucial link. They significantly expand the dimensions and scope of autonomous driving data collection, enhancing safety performance through simulations and predictions of real-world roads and vehicle conditions, enabling accurate predictions during travel. Large models and autonomous driving move in step, with the former paving the way for a new round of cost reductions as it matures.

Behind Robotaxi lies a value chain, with upstream industries related to autonomous driving and vehicle manufacturing technologies, midstream automakers and internet-based automakers, and downstream Robotaxi operating platforms and related service providers.

A breakthrough in a single point translates into a full bloom across the entire value chain. At this stage, advancements in autonomous driving technology, policy, and commercialization also sow the seeds for future upgrades to the industrial chain.

The battle for Robotaxi is imminent, with Chinese and American giants each selecting their technical routes and leveraging cost reductions to gain an edge. However, the outcome remains uncertain. Nevertheless, as the global competitive landscape becomes increasingly clear, Chinese autonomous driving enterprises must maintain a sense of urgency, lest one misstep leads to a series of errors. Only by constantly striving to accelerate can they emerge victorious in the China-US competition for autonomous driving.