Shulou(Shulou.com)11/24 Report--

"Argo AI has been extensively developed and tested in more than eight cities in the United States and Germany. This gives Argo Drive access to a wide range of real-world data, terrain, climate, traffic patterns and driving behavior. Every mile it travels, it becomes safer, smarter and more scalable."

Argo AI founder Brian Salisky (Bryan Salesky) may not have imagined that less than a week after the tweet, the peak market capitalization was as high as $7 billion and Argo AI, with a team of more than 2000 people, would suddenly usher in the final moment.

Argo AI, a promising auto-driving technology development project that has raised billions of dollars from Ford and Volkswagen, will be shut down and disbanded, and its employees and some parts will be taken over by Ford and Volkswagen, Forbes reported.

Argo AI is a well-deserved star company in the self-driving industry. Founded in 2016 and headquartered in Pittsburgh, USA, it aims to develop, test and eventually commercialize its autopilot system.

In the second year of the company, Ford announced an investment of "1 billion US dollars over five years". In 2019, Argo AI won another $2.6 billion in strategic financing from Volkswagen, and Volkswagen and Ford each own about 39% of the shares in Argo AI.

Entrepreneurship is a survivor game, but the collapse of Argo AI still makes people see the ruthlessness and vividness of a surge into the industry. The reversal fate of Argo AI is also a microcosm of the self-driving industry: self-driving companies are suffering a serious valuation downturn.

For example, the valuation of Google's self-driving company Waymo has fallen from as high as $175 billion to $30 billion; recently, Intel's Mobileye went public with a market capitalization of about $16.7 billion at IPO price, more than 1/3 of Intel's initial estimate of $50 billion.

Crunchbase, a US enterprise services data company, tracked 14 companies with self-driving car-related technologies listed in recent years and found that they all fell by an average of more than 80 per cent after listing.

Around this incident, this article will try to answer three questions:

1. Why is Argo AI closed?

2. From the enthusiasm for capital to tide over the difficulties together, why are L4 self-driving companies not as expected collectively?

3. How should the commercialization of autopilot land?

The golden age of autopilot Argo AI was founded when it was the golden age of self-driving.

In 2016, Waymo became independent from Google and became a subsidiary of Alphabet; Mobileye became the world's largest supplier of advanced driving assistance systems, with 12 million cars worldwide at that time; NuTonomy, the world's first self-driving taxi, began to carry passengers in Singapore; and Cruise was acquired by General Motors.

It was at the end of the year that the ambitious Brian Salisky left Google's autopilot team to co-founded Argo AI with Peter Rander, the former chief engineer of Uber's advanced career technology division.

According to their vision, Argo AI will focus on providing L4-level autopilot systems and services such as shared travel and cargo delivery in crowded urban areas.

The establishment of Argo AI immediately attracted the attention of Mark Fields, the then CEO of Ford.

At the time, Mark Fields was working to lead Ford through its transformation from a car company to a mobile company, and "a class 4 self-driving car to be launched in 2021" was one of Mark Fields' strongest aspirations.

Previously, the leader with many advanced ideas on autopilot had invested a lot of resources and expectations on L4 autopilot, including Pivotal, a cloud computing company, and $75 million in multi-line lidar company Velodyne. In his eyes, Argo AI will be a key jigsaw puzzle for Ford to promote self-driving research and development.

At that time, there seemed to be endless hot money in the burning wilderness of autopilot. In February 2017, Mark Fields announced a $1 billion investment in three-month-old Argo AI, even though Argo AI had fewer than 12 employees and few decent hardware and software products.

This investment not only solves the urgent need of R & D funding for Argo AI, but also meets the strong demand of traditional car companies for self-driving technology. In 2019, Volkswagen also made a high-profile move, injecting $2.6 billion into Argo AI and co-controlling Argo AI with Ford, expanding its cooperation to electric vehicles and self-driving.

Under the blessing of a huge amount of funds, the development of Argo AI is drumming.

In 2021, Argo AI has been extensively developed and tested in more than 8 cities in the United States and Germany. In May, Argo AI launched a pilot program for the commercialization of self-driving in Miami and Austin, including a partnership with Lyft to promote automatic taxi services and a joint deployment of self-driving delivery services with Wal-Mart.

A spokesman for Argo AI has vowed that Argo AI's self-driving business will start commercial operation at an appropriate time.

However, in contrast to the high-profile advance of the Argo AI, autopilot is spending far more money than Ford and the public can imagine. At the same time, the process of commercial landing of L4 autopilot is far from what they expected.

Ford reported a net loss of $827 million in the third quarter of 2022, which Ford blamed on Argo AI, saying it recorded a $2.7 billion non-cash pre-tax impairment on its investment in Argo AI.

Jim Farley, CEO of Ford, said directly that while Ford was optimistic about the future of the L4 ADAS, the large-scale commercial landing of driverless cars would cost billions of dollars and would take at least five years.

In a statement, Ford admitted that it needs to invest in driving assistance technology that is easier to achieve in the short term, rather than Argo AI's goal of fully self-driving.

Oliver Blume, Volkswagen's chief executive, made a more euphemistic statement, but his position was equally clear. He believes that focus and speed are crucial in the development of future technologies. Volkswagen's goal is to provide customers with the most powerful functions in the shortest possible time, and to make the development of enterprises as cost-effective as possible.

If you look at it from an industry-wide point of view, Argo AI's bankruptcy is not difficult to understand. Against the backdrop of shrinking global liquidity, it is more difficult for start-ups to obtain external financing. L4 autopilot faces difficulties in technology, regulatory standards, supply chain and other aspects, and the speed of commercial landing is quite different from the early optimistic expectations of investors.

Obviously, Argo AI does not have the ability of self-hematopoiesis in a short time. Unable to get external support, Ford and Volkswagen finally chose to stop their losses with pain and put their limited funds into the development of L2 + and L3 intelligent drivers, which are easier to land.

A research report through the Valley of death once pointed out that 90% of the scientific research results are buried in the process from basic research to commercialization before they reach the market. Therefore, the process of an innovation from laboratory samples to market landing as a commodity is often referred to as the "valley of death".

Because in technology research and development, it is often a single point of breakthrough, but if we want to turn technology into a product, we should pay attention to the performance of the whole product, as well as the acceptability and experience of consumers. In addition to technical engineering, production, the company's operation also involves the market, financing, talent and other problems.

In fact, there have always been two technological routes of "gradual" development and "leapfrog" promotion in the research and development of autopilot.

The former is mainly based on the traditional mainframe factory and the new force of car construction, starting with the relatively basic and low-difficulty auxiliary driving, and through continuous iteration to realize the auxiliary driving functions of L1, L2 and L2 + to improve the automation level of the car; the latter, mainly based on science and technology enterprises, advocates the research and development of self-driving cars above L3 level and achieve high-level self-driving in one step.

Especially for L4 self-driving companies that advocate "leapfrog" technological route, the "valley of death" of technological innovation is particularly profound, and the process of commercializing high-level self-driving technology is much more difficult than they promised.

As one of the first companies to commercialize self-driving, Waymo is often the "stone" that many L4 self-driving companies explore for commercialization.

At present, Waymo has launched two major commercial projects, Waymo One and Waymo Via, the former providing unmanned taxi services and the latter providing unmanned freight services, but neither of them has been able to run through the complete L4 commercialization path so far.

Take Waymo One as an example. In December 2018, Waymo launched an unmanned taxi service, which was named Waymo One. Four years have passed, limited by multiple factors such as technology, cost, laws and regulations, Waymo One is only open to the public in the East Valley area of Phoenix, and has not made a profit.

The Bloomberg article hit the nail on the head: "Waymo's autopilot is 99% complete, but the remaining 1% is the hardest."

The reason is that, on the one hand, the research and development of high-level autopilot requires a large amount of money. McKinsey estimates that established automakers and startups have invested $106 billion in self-driving capabilities between 2010 and 2021.

However, compared with huge investments, L4 self-driving companies do not seem to be able to give investors a greater return in a short period of time. At present, most of the mainstream L4 self-driving companies are technology suppliers of self-driving taxis or logistics companies, there is still no timetable for large-scale commercial landing, and it is a long way off to make a profit.

On the other hand, the commercial landing of high-level autopilot is still faced with policies, regulations and other restrictions. In addition to the iteration of product technology, the improvement of the policy and standard system, the establishment of relevant infrastructure and the improvement of social acceptance are also the key factors restricting the landing of high-level autopilot.

The philosopher George George Santayana said: "those who refuse to learn from history are doomed to repeat its tragedy." For L4 self-driving companies, how to change their strategy to avoid repeating the mistakes of Argo AI?

Two paths to accelerate commercialization in 2020, China's "Intelligent Network Automotive Technology Roadmap 2.0" was released, which set three time nodes for the commercialization of self-driving in China.

By 2025, highly self-driving vehicles will be commercialized in limited areas and specific scenarios; by 2030, the market share of highly self-driving vehicles will reach 20%, and they will be widely used on highways and large-scale roads in some cities; by 2035, China's intelligent network-connected automobile technology and industrial system will be completed in an all-round way, the industrial ecology will be sound, and highly self-driving vehicles will be applied on a large scale.

Optimistically predicted that in China, high-level self-driving cars to achieve large-scale mass production, at least 8 years. For L4 self-driving companies, accelerating the landing of technology and realizing their own hematopoiesis will become the key to the development of enterprises. In the process, there are two paths worth paying attention to.

First, take the lead in autopilot landing in a closed or semi-enclosed area under a limited scene. The research report "Autopilot Application scenario and Commercialization path" released by che Bai think Tank shows that, from the point of view of the difficulty of technology application and the degree of being affected by laws and regulations, the commercial application path of autopilot will follow the principle of closing first and then opening up. The principle of loading goods first and then carrying people, in order to choose the commercial scene.

Autopilot can be first used in closed or semi-closed areas under limited scenes, such as automatic parking, closed logistics and transportation in the park, followed by trunk logistics, terminal distribution, fixed line sanitation, bus commuting, ride-hailing and so on.

Take the logistics transportation scene in the closed park as an example. The typical scenes of logistics transportation in the closed park include mining area, port, airport and so on. At present, the logistics industry in the closed park is facing problems such as reducing operating costs and difficulties in recruiting drivers, and is developing towards automation, intelligence and unmanned, and improves operational efficiency through high-quality transformation and upgrading.

Specifically, it is more feasible to use autopilot technology in the closed park. The maximum speed of self-driving vehicles in the closed park is no more than 30 km / h, which is lower than that of 40km / h on the road, which reduces the time and computing requirements of the whole processing flow from information collection to decision control. And the road condition inside the closed park is simple, there are no people and animals crossing the road, and the working vehicles and machinery all drive in the direction indicated by the road, which reduces the difficulty of the realization of the autopilot system as a whole.

Through the use of autopilot technology, closed park logistics can reduce personnel expenditure, fuel consumption and parts consumption. First, through the autopilot system to replace the driver, we can save the driver's labor cost and logistics cost; second, through accurate operation in the driving process, the system adopts the optimal driving strategy, which can effectively improve the driving efficiency, reduce fuel consumption and save fuel costs; third, using the optimal and highly consistent driving strategy, autopilot can effectively reduce the loss of vulnerable and consumable parts, including tires.

Second, reduce dimensionality and give priority to L2 / L3 auxiliary driving on production cars. At present, L2 intelligent driving is accelerating large-scale commercial applications. The latest data tracking report on China's self-driving car market released by IDC shows that the penetration of L2 self-driving technology in the passenger car market is increasing, reaching 26.6% in the second quarter of 2022.

Among them, Xiaopeng's NGP, Weilai's NOP and the ideal NOA have all realized automatic navigation-assisted driving on highways and some urban expressways. At present, the new forces of car building are competing for landing navigation assistance systems in cities, approaching L3 autopilot.

In this process, there are a large number of common key technologies that can be used to reduce dimensionality. Therefore, providing L2-level autopilot mass production solutions for mainframe factories is expected to become one of the important means for L4-level autopilot enterprises to commercialize.

In fact, many L4 self-driving companies at home and abroad have begun to reduce dimensionality in an attempt to grab the first "platform ticket" leading to the future.

While developing L4 self-driving technology, Cruise also undertakes the task of building a Ultra Cruise intelligent driving system for GM. It is reported that compared with the Super Cruise currently carried by GM models, the Ultra Cruise has new autopilot functions, such as ensuring front and rear intervals, supporting automatic and on-demand lane change, and supporting left and right turns.

In GM's words, Super Cruise will complement Ultra Cruise and extend driving assistance technology to GM's entire product spectrum, allowing more people to enjoy the technology dividend.

The end has a bright future and the road is tortuous. This sentence is the most appropriate to describe the development of self-driving.

In this adventurous field, auto companies, auto parts suppliers, Internet companies, start-up unicorns and other characters are "heroes everywhere" in an attempt to burn out a new world by virtue of the wall-breaking effect of technology.

The curtain call of Argo AI is more like a cold bath for the hot track of autopilot. People gradually realize the fact that the development of autopilot is a step-by-step process of "laying eggs along the way", which needs to follow the basic business logic. Whoever can find a better balance between technology, scale and cost will be more likely to bring innovation out of the valley of death.

The good news is that China is expected to accelerate the process.

On November 2, the Ministry of Industry and Information Technology and the Ministry of Public Security openly solicited opinions on the notice on the pilot work of Intelligent Network access and Road access (hereinafter referred to as "the notice").

The pilot content proposed in the notice is that on the basis of the road test and demonstration application work of the National Intelligent Network Association, the Ministry of Industry and Information Technology and the Ministry of Public Security select qualified road motor vehicle production enterprises and intelligent Internet connected automobile products with self-driving function with mass production conditions to carry out access pilot. For the intelligent network-connected automobile products that have passed the access pilot, pilot road access will be carried out within the limited public road areas of the pilot cities.

It is worth mentioning that the automatic driving functions carried by the intelligent network-connected vehicles in the notice include L3-level driving automation and L4-level driving automation functions. This means that after the implementation of the notice, L3 and L4 smart cars are expected to hit the road in specific areas, and the self-driving industry will enter a new stage of commercial pilot.

Full-text reference

[1] "burning up $3.6 billion, dragging Ford into a loss quagmire, autopilot unicorn Argo AI goes bankrupt", Economic Observer Network, Zhang Qian

[2] "autopilot costs more than money? Ford won this time. "car stuff, Origin."

[3] "what regulatory and regulatory problems do autopilot have to face before large-scale commercialization? Interface News, Wu Yangyu

[4] behind the sudden collapse of Argo AI: capital no longer believes in "Story" L4 players "reduced dimension"migration", set WeChat, Tusha

[5] "self-driving can be put on the card! Two departments solicit opinions ", Shanghai Securities News, Ruan Xiaoqin

This article comes from the official account of Wechat: che Bai think Tank (ID:EV100_Plus), by Cheng Honghe.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.