Looking back at the various players in the autonomous driving race, those who have survived are either those who have a strong financial backing or those who have focused on specific scenarios and technology, gradually developing their capabilities. However, the commonality among those who are thriving is their ability to adapt to the demands of the industry and ensure that their technology meets practical needs.
Author|Dou Dou
Produced by|Industry Family
“When it comes to mass-produced models, Haomo is definitely the leader,” Frank said confidently, with no hesitation in his voice.
Walking into the Beijing Aobei Technology Park, passing rows of standalone buildings, the reporter saw the signs for Haomo Intelligent Mobility. Following the signs, a few steps later, they arrived at the company’s office and saw the autonomous delivery vehicle parked on the roadside – the “Little Magic Camel”.
Here, we met Haomo Intelligent Mobility’s COO, Hou Jun, but everyone prefers to call him Frank. He is articulate, down-to-earth, and looks like a typical tech-savvy individual.
In Frank’s words, Haomo Intelligent Mobility is dissected and broken down, presented more concretely and clearly.
When it comes to autonomous driving, the common impression is that it is expensive, and the massive market potential and space for growth after widespread adoption have attracted numerous investors and entrepreneurs over the years. However, as mentioned earlier, the high costs and slow commercialization have led to many startups failing halfway, and investors are becoming more cautious.
Autonomous driving requires a balance between obtaining funding and achieving commercialization, a process of self-sustenance. As a result, autonomous driving companies have taken different paths.
Among these different paths, Haomo Intelligent Mobility’s presence has become increasingly prominent. As one of the few autonomous driving companies that have survived and thrived, what is their secret?
1. In 2019, “a prediction”
From 2018 to 2019, there was a wave of financing in the autonomous driving industry. During this period, Shanghai issued the first batch of intelligent connected vehicle test plates nationwide, marking the official opening of China’s autonomous driving road testing.
Technology companies represented by Baidu, Alibaba, and Tencent accelerated their exploration and layout in the autonomous driving field.
In addition, Waymo launched the first commercial autonomous driving taxi service, Waymo One, in Arizona, USA, marking the beginning of commercial operation of autonomous driving technology.
The appearance of this industry “benchmark” became a driving force for the wave of autonomous driving financing. Morgan Stanley, based on its “leading position” in autonomous driving technology and the potential for significant new revenue from autonomous driving, issued a research report valuing Waymo at $175 billion.
“At that time, the market value of the largest car manufacturer in China was estimated to be less than 100 billion RMB.” In Frank’s memory, Waymo’s Robotaxi became the pinnacle of technology sought after by many autonomous driving companies.
The financing frenzy in the autonomous driving industry officially kicked off.
Data shows that in 2018, despite the overall economic downturn and capital winter, the total financing in the autonomous driving field did not decrease but rather increased significantly. The financing amount for autonomous driving components and solutions suppliers increased from 5.369 billion yuan in 2017 to 16.231 billion yuan in 2018.
In response to the trend, Haomo immediately responded. “Our goal is for artificial intelligence technology to be widely applied in the field of autonomous driving.” In Frank’s view, Haomo’s pursuit is not just about absolute technological leadership.
However, in reality, achieving widespread application of artificial intelligence technology in autonomous driving, especially through the Robotaxi route, is challenging in a market that is difficult to promote in the long term. Compared to other routes, surviving and thriving through this path is even more difficult, let alone reaching the day of widespread adoption.
Indeed, commercializing autonomous driving cannot be achieved overnight, which is not realistic. A gradual development path is essential to ensure steady technological progress and effective market integration. Therefore, Haomo Intelligent Mobility decided to follow a progressive development path from low-speed to high-speed, from cargo to passenger, and from commercial to civilian.
Looking back now, this decision has proven to be wise.
In 2021, the one-step-to-market approach represented by Waymo has struggled to achieve commercialization, with its valuation dropping to $30 billion, a decrease of over $140 billion from three years ago. In contrast, the gradual development route has gained more attention and recognition over time.
Looking at the companies that have stuck to the Robotaxi route, apart from giants like Baidu, very few have survived.
“We are also researching the Robotaxi route, but it is not our main focus. Our goal is a closed-loop for data, a closed-loop for business, and a closed-loop for cash flow and profits, otherwise, we may not survive until the day when Robotaxi is widely adopted,” Frank told Industry Family.
2. The “Golden Loop” of widespread deployment
So, how exactly did Haomo overcome the challenges of achieving a closed-loop for data, business, and profits over the years?
“Feasible, reliable, and commercially viable” are the three key stages of autonomous driving technology development proposed by Haomo Intelligent Mobility. In Frank’s view, this framework has guided the research and commercialization path of Haomo’s autonomous driving technology.
Haomo Intelligent Mobility has invested heavily in basic research on autonomous driving technology, including perception systems, decision algorithms, and vehicle control technology. By testing prototype vehicles in controlled environments, Haomo Intelligent Mobility has further validated the basic functions and performance of autonomous driving technology.
As the technology advances, it is crucial to ensure that the system can operate stably in a wider range of environments and conditions, meeting the strict standards of the automotive industry. The safety of the system, vehicle-level requirements, the ability to operate around the clock, and adaptability to all terrains have been enhanced.
Haomo Intelligent Mobility leverages a large amount of real-world operational data to continuously optimize and adjust autonomous driving algorithms, improving the system’s stability and safety. Through continuous technological iterations, the adaptability of the autonomous driving system to various weather, road conditions, and traffic situations is enhanced. This ensures that the autonomous driving system meets the strict standards and regulatory requirements of the automotive industry.
In the commercial application stage of autonomous driving, technology must not only be mature and reliable but also commercially viable, meeting the diverse needs of different scenarios such as high-speed passenger transport, low-speed cargo transport, while controlling costs and complying with policy regulations. At this stage, autonomous driving technology needs to gain market acceptance and create value in practical commercial environments.
Haomo Intelligent Mobility has conducted in-depth research and development of highly adaptable autonomous driving solutions for different application scenarios such as high-speed passenger transport and low-speed cargo transport. By innovating in technology and mass production, the cost of autonomous driving systems has been reduced, creating high-value products and enhancing market competitiveness.In the process, Momenta has accumulated a large amount of actual driving data by deploying autonomous driving systems on various vehicle models, including sensor data, vehicle status, and driving behavior. This data is used to train and optimize machine learning models, thereby improving the performance and safety of autonomous driving systems.
Momenta continuously iterates its autonomous driving algorithms to adapt to changing driving conditions and scenarios. The algorithms are validated through simulation and real-world testing, and necessary adjustments are made. Autonomous driving products tailored for different application scenarios, such as passenger car assistance systems and low-speed unmanned logistics vehicles, have been developed.
The driving data generated during the operation of these products is collected again and used for training and optimizing machine learning models, forming a data loop. Momenta has built a complete business loop from research and development to mass production to services, ensuring the continuous development and commercial application of the technology.
Currently, Momenta has launched two generations of seven intelligent driving products in the field of assisted driving, which can meet the mass production needs of different vehicle models at high, medium, and low price ranges.
Among them, the first-generation product achieved driving and parking on highways and expressways, characterized by high precision maps and good stability but slightly higher costs. The second-generation product, without high precision maps, has lower costs and can be priced at around one thousand yuan. The HP170, HP370, and HP570, three assisted driving products priced at around one thousand yuan, have been delivered successively.
In the commercial vehicle sector, Momenta’s L4 level unmanned logistics vehicles have been widely adopted since 2020. Multiple generations of products have been produced, and strategic partnerships have been established with companies such as Meituan, Alibaba, Wumart, Dada, Jitu, and Tongda.
It is worth noting that the cost of products in this scenario has dropped from 5-10 million yuan per unit in 2020 to less than 100,000 yuan per unit.
Thus, the “leader” mentioned by Frank in the opening sentence of the article has become more concrete.
III. Momenta, Taking a Different Path
“As the saying goes, everything has its three elements: big models, big data, and big computing power. Without the ‘three,’ there can be no advancement. However, the ‘three’ is not controlled by one or a few enterprises, but by two enterprises and multiple enterprises,” Frank told industry professionals.
It is well known that big data, big computing power, and big models are the three elements necessary for the large-scale implementation of autonomous driving technology. However, both automakers and technology companies, as the two major camps in the autonomous driving race, have their own unreachable areas.
From the beginning, Momenta has taken a unique path, forming a data loop through its collaboration with automakers. This is the unique route chosen by Momenta.
In the eyes of many, the relationship between automakers and technology companies in the autonomous driving race is not simply a cooperative relationship but a “coopetition” relationship.
Compared to autonomous driving technology companies, traditional automakers usually choose to cooperate with technology companies if the self-research cost of autonomous driving technology is too high. However, technology companies rely on car companies for massive data acquisition. While they can obtain data in the short term, cooperation with multiple car companies may lead to unstable cooperation, making it difficult to ensure data reliability and stability. However, new forces in the direction of self-research, such as WeRide, can do so.
This inevitably brings some hidden dangers for autonomous driving technology companies in the future competitive landscape.
“The large-scale implementation of autonomous driving is not just about the technology itself but also about the mechanism of the enterprise,” Frank believes that this phenomenon stems from the conflict of corporate culture, cooperation, and interest distribution mechanisms.
Traditional enterprise work models and incentive mechanisms are derived from the industrial age, emphasizing planning and clarity. The internet age, on the other hand, emphasizes innovation and data-driven approaches, tending towards adjusting while practicing. These two concepts can lead to significant conflicts in specific work and strategic planning.
One fact is that this mechanism is also being verified. Many automakers, after setting up autonomous driving technology subsidiaries and departments, have experienced layoffs and departmental dismantling, and have been defeated. From 2022 to 2023, the presence of automakers can be seen among the financing parties of many autonomous driving technology companies.
Alliances between automakers and autonomous driving manufacturers are gradually becoming mainstream.
By June 2024, Momenta has deployed over 20 mass-produced vehicles, with users driving over 170 million kilometers with assisted driving. Currently, Momenta has achieved a two-pronged approach – continuous funding from financing and self-sustaining capabilities through large-scale implementation.
In summary, Momenta’s technological implementation path not only involves predicting the development path, technological innovation, and data-driven commercial closed loop, but also considerations in terms of enterprise mechanisms.
The underlying logic of Momenta is also driving it towards the era of autonomous driving 3.0, represented by big models, big computing power, and big data.
IV. The AI Era, Heading towards Autonomous Driving 3.0
The wave of big model technology is sweeping in, propelling autonomous driving into a new era.
Even today, there are still many challenges to be solved in the development of autonomous driving.
In traditional autonomous driving systems, there are complex dependencies between modules such as perception, localization, planning, and control, requiring a lot of engineering effort to design and optimize each module.
Simply put, current autonomous driving systems are essentially based on rules written by engineers. Therefore, in autonomous driving, there are long-tail problems, which are rare but crucial driving scenarios where the system struggles to make the correct decisions.
The eruption of big model technology brings new possibilities.
End-to-end models can learn directly from data, simplifying the development process of autonomous driving systems and making the process from sensor input to decision output more direct. With no need for manually designed features, the models can adapt to various driving scenarios and conditions, while the integrated model reduces the dependency between modules. This enables rapid learning from a large amount of data on how to handle these rare events and driving scenarios and make decisions accordingly.
In 2023, Momenta released the industry’s first autonomous driving generative big model, DriveGPT (Xuehu Hai Ruo), achieving significant breakthroughs and innovations in data screening and mining, automatic annotation, simulation generation, and cognitive interpretability.
From this perspective, big model technology will further improve the intelligence level of autonomous driving systems in the future. Through continuous learning and optimization, generative big models like DriveGPT will be able to accurately identify and understand complex traffic scenarios, enabling more precise and safe decision-making. This will greatly enhance the capability of autonomous driving systems to respond to various emergencies and complex environments.
Driven by technology, Momenta is gradually moving from traditional modular design towards a more intelligent direction, advancing towards the era of true autonomous driving 3.0.
Looking at the various players in the autonomous driving race at present, those who survive either have “wealth at home” or focus on specific scenarios and technological reduction, gradually developing. However, the commonality among those who thrive is that they adapt to demand and ensure that technology is applied to the industry.
“In autonomous driving, the deeper the water, the bigger the fish. When the field is still in its early stages, even leaders will find it difficult to demonstrate their value,” Frank told industry professionals.
According to reports, Momenta also received orders for mass production from major global automakers at the beginning of 2024. The value of Momenta, this “big fish,” is becoming increasingly clear.
