“Can you help me build a self-driving car that can drive freely on any road in California?”
Twelve years ago, Google founder Larry Page threw an idea to Sebastian Thrun, Google’s “father of self-driving”. It was this idea that seemed unrealistic at the time that opened the road to Google’s automated driving, and also opened the prelude to the global autonomous driving industry.
After years of development, more and more self-driving startups are popping up in Silicon Valley. In China, on the other side of the ocean, the autonomous driving track is becoming more and more lively. Not only Internet companies such as Baidu and Didi have become pioneers; independent autonomous driving companies such as Pony.ai, TuSimple, and WeRide have also emerged one after another.
In addition, autonomous driving is an important part of the implementation of electric vehicles, and neither the new car manufacturers nor traditional car companies have given up this cake.
However, since its birth in 2009, the autonomous driving industry has been plagued by problems such as difficulty in ensuring safety, difficult to implement technology, and difficult to realize commercialization. From a domestic point of view, recently, Pony.ai, which previously wanted to go public in the United States, has had new developments.
According to a LatePost report, about a month and a half ago, Pony.ai has merged the truck autonomous driving research and development team into the passenger car research and development team. Before the merger, the truck division totaled about 100 people. Accompanied by the business contraction, there are also the departure of relevant core personnel.
Photo source Pony.ai official Weibo
When both capital and opponents are advancing at a high speed, Pony.ai seems to be pressing the adjustment button. Before that, the truck segment was a business that Pony.ai highly valued.
On the other hand, the commercialization of autonomous driving is also an industry-wide problem. Even the industry leaders Google, Baidu and other companies have not found a better way out.
Can Pony.ai, whose vitality is severely damaged, survive until it is profitable? Where is the way out for independent self-driving companies?
Pony.ai’s vitality is seriously damaged
Before the employees could react, the helm of Pony.ai has announced the completion of important adjustments.
According to a LatePost report, at a team communication meeting in mid-September, the two founders of Pony.ai announced that they would merge the truck autonomous driving R&D team into the passenger car R&D team. At that time, they said that the company had to concentrate resources to do important things, and the efficiency of independent research and development of the truck team was low.
At the same time, business adjustments have also caused turmoil in the management team. According to a LatePost report, after the business adjustment, several generals have resigned one after another. Pan Zhenhao, head of the autonomous driving technology of Pony.ai’s truck division, and Sun Haowen, head of the domestic autonomous driving PnC (planning and control) group, have resigned after adjustment.
By advancing the timeline, the adjustment of Pony.ai’s business will be more frequent. According to a LatePost report, before this structural adjustment, the R&D team of Pony.ai’s truck division tightened the recruitment quota. In September, the road test of trucks in the United States was suspended, and in August, Pony.ai shelved the previously advanced plan to go public in the United States.
Regarding this adjustment, Pony.ai responded to LatePost, saying that this adjustment takes into account the commonality of passenger car autonomous driving technology and heavy truck autonomous driving technology, and it is more efficient to research and develop together.
However, according to Wired, the truck business was once an important development direction of Pony.ai.
Last year, the Pony.ai Truck Division was established. In March of this year, Pony.ai officially released its independent business brand “Pony.ai Smart Card”. At that time, Pony.ai CEO Peng Jun said: “Our strategy is very clear. Pony.ai will focus on the two business segments of passenger vehicles and commercial vehicles at the same time.”
In April of this year, Pony Smart Card also signed a cooperation framework with ZF, a supplier of automotive brake control systems, to cooperate in the field of commercial vehicle logistics; in May, Pony Smart Card obtained the first batch of test licenses for autonomous trucks in Guangdong Province , can carry out logistics business operations. In July, Li Hengyu, head of the Pony Truck Business Department, said that Pony Trucks had carried out more than 200 days of road tests, completed 13,600 tons of cargo transportation, and ran 37,000 kilometers of commercial operation mileage.
Pony Smart Card, the source of the Pony.ai official account
Since then, the truck business has been in a fast-moving phase, with new developments almost every month. At the same time, Li Hengyu, head of the Pony.ai truck department, has publicly stated that Pony.ai can apply 80% of the autonomous driving technology of passenger cars to trucks.
It can be seen that in the new energy smart truck track, Pony.ai, which has spent a lot of manpower and financial resources, also ran on the highway. But now a series of changes are pointing to the truck business that was once promising is not progressing as smoothly as imagined.
The problem facing Pony.ai is not only the shrinking of internal business, but also the resignation of core personnel will undoubtedly make Pony.ai worse.
Specifically, Pan Zhenhao used to be the core technician of Pony.ai and communicated closely with Lou Tiancheng, the co-founder and CTO of Pony.ai. The team led by Sun Haowen was also trusted by the employees because of the good atmosphere.
According to the information obtained from the online travel, a company called “Qianhang Technology” was established on July 29 this year. At the same time, it can be seen from the official website of Tianyancha that one of the directors of this company is Sun Haowen, who used to be Pony Smart Card Technology Co., Ltd. At the same time, Qianhang Technology received strategic financing from Xiaopeng Motors and SF Express a week ago. The company is positioned as a “L4 driverless freight truck technology developer company”.
In addition, according to Auto Heart, in July, Zhao Ruixuan, the former head of Pony.ai’s investment and financing, founded a company called “Xingang Technology”. Tianyancha shows that the company’s entrepreneurial direction is to build electric trucks covering L2+ to L4 autonomous driving.
One thing in common is that the entrepreneurial direction of the above-mentioned resigners chose the field of autonomous driving invariably, forming a direct competitive relationship with Pony.ai. It used to attract talent intensively to the current outflow of talents. For a time, Pony.ai seems to have become the “Whampoa Military Academy” in the field of autonomous driving, and has cultivated many competitors for itself in the field of autonomous driving with many players.
Lou Tiancheng once described the way to attract and retain talents: “What top talents need is actually very simple, that is, success. We may not have other things, but we still have a real possibility of success.”
Today, this “possibility of success” may be a little vague. With the adjustment of the truck business and the departure of core personnel one after another, Pony.ai has also faced unprecedented challenges.
Where is Pony.ai’s imagination?
As the self-driving company with the highest valuation in China, Pony.ai has many stories to tell.
In fact, the founding team of Pony.ai, established at the end of 2016, is very luxurious. Pony.ai was co-founded by Peng Jun, the former chief architect of Baidu’s North American R&D, and Lou Tiancheng, the Chinese programming “Master Lou”, who served as the CEO and CTO of Pony.ai respectively. It is worth mentioning that both Lou Tiancheng and Peng Jun had professional experience in Google and Baidu, which can be regarded as “Google department” and “Baidu department”.
The two founders of Pony.ai, Lou Tiancheng on the left, Peng Jun on the right, picture source Pony.ai official account
It was rumored that when Pony.ai was founded, whenever Peng Jun or Lou Tiancheng and Waymo, Uber employees appeared in a coffee shop in Silicon Valley, Silicon Valley companies received resignation letters the next day. Although this is a rumor, it is enough to show the attractiveness of the two founders of Pony.ai in the industry.
Perhaps thanks to the dazzling founding team, Pony.ai has been the darling of capital since its inception. According to Tianyancha, Pony.ai has obtained eight rounds of financing so far, with Sequoia China, IDG Capital, Toyota Motor, China FAW and other well-known capitals successively entering the market, with a total financing amount of more than 1.1 billion US dollars.
Under a series of good capital news, the news of Pony.ai’s sprint for listing this year has been repeated. According to 36Kr report, the listing valuation of Pony.ai is expected to reach 10 billion US dollars.
There is no doubt that technology will always come first for self-driving companies. “Hopefully one day, we can stand together with Google to compete in technology,” Lou Tiancheng once said in an interview with LatePost.
In order to achieve this goal, Pony.ai has made many technical attempts.
An important milestone occurred when Pony.ai landed in Guangzhou in 2018. At that time, Pony.ai encountered a rainy day that had not occurred during the test in California before. Due to the hardware limitations of lidar, how to filter out the noise of the rain sensor and solve the problem of splashing water to block the camera’s field of view is a common problem in the industry.
To this end, Pony.ai added an automatic driving sensor cleaning system to automatically solve the problem of vision when it rains; upgraded multi-sensor deep fusion technology to avoid the shortcomings of a single sensor. WeRide, which is also an L4-level independent autonomous driving company, has recently launched an automatic sensor cleaning system to provide vision guarantees for adapting to multiple weather scenarios.
It should be noted that the general guideline of the industry is to classify autonomous driving into 6 levels according to the definition of the American Society of Automotive Engineers (SAE), from L0 (fully manual) to L5 (fully automatic). According to the latest definition, L0-L2 systems are named “driver assistance systems”, while L3 to L5 are considered “autonomous driving systems”.
Driving automation classification, source SAE International public account
In addition to technical upgrades, Pony.ai has also accumulated a certain amount of test mileage. According to the official introduction of Pony.ai, Pony.ai’s L4 autonomous vehicles have accumulated more than 8 million kilometers of test mileage around the world. WeRide Zhixing has accumulated more than 6.5 million kilometers of automatic road test mileage. According to the prospectus of TuSimple, the “world’s first autonomous driving stock”, its cumulative test mileage has reached 4.5 million kilometers.
On the road test license plate, Pony.ai recently obtained the Beijing “main driver without safety officer” automatic driving license. This is Pony.ai’s fifth road test city location after Guangzhou, Fremont, California, Irvine, and Milpitas. In the field of autonomous driving, such a multi-city road test is not an exception. TuSimple will obtain road test licenses in Shanghai, Tangshan, and California in the United States in the future, while WeRide Zhixing and Yuanrong Qixing will choose Shenzhen, Hangzhou, Wuhan, Zhengzhou, etc.
In fact, apart from Baidu, Pony.ai is the first independent self-driving company to obtain an unmanned road test license in Beijing. In June this year, Pony.ai’s fully driverless vehicles joined the Guangzhou Robotaxi fleet. The so-called fully driverless, that is, there is no one in the driver’s seat to control it. According to Wired Travel, the unmanned road test is divided into three test stages, namely “co-pilot”, “rear seat” and “remote outside the car”.
As can be seen from the road test video released by Pony.ai, Pony.ai’s fully unmanned autonomous driving system can enable passenger cars to reach the level of L4 autonomous driving. In the face of sudden road conditions such as temporary road closures, the system can also use the remote assistance platform (Remote Assistance) to use the fusion results of perception information to execute remote decision-making instructions, such as reversing, detouring, and pulling over, so as to get out of trouble.
Of course, the road test scene of the day is sunny, so if it is in bad weather such as rain, snow, hail, etc., the requirements for the L4 level will be higher. According to the estimates of the Ministry of Industry and Information Technology, by 2025, the sales of L2 and L3 intelligent connected vehicles in my country will account for 50% of all car sales, and the L4 intelligent connected vehicles will begin to enter the market. Therefore, it is not easy for autonomous driving technology to achieve full-scene L4 level.
To a certain extent, the industry consensus is that “bicycle intelligence + vehicle-road coordination” is the key to the implementation of autonomous driving technology. The two complement each other, which has both policy dividends and business considerations for companies. The more licenses you get in cities, the more complex road conditions and rules you can test, the more accurate and safer the test of your own technology.
It can be seen that Pony.ai is at the forefront of the industry in terms of technology, and it has also been expanding its commercialization scenarios. There is a saying in the autonomous driving industry, “Climb Mount Everest and lay eggs along the way.” Before this business adjustment, Pony.ai had two “eggs”, Robotaxi and autonomous trucks.
As early as December 2018, Pony.ai launched the Robotaxi project PonyPilot, which is the first company to launch Robotaxi services in China. As of last year, it completed more than 80,000 taxi orders across the country.
It is worth mentioning that in May this year, Pony.ai upgraded the PonyPilot+ service. According to the official information, these Robotaxis that provide PonyPilot+ services are all built on the basis of Lexus RX450h and equipped with the fifth-generation L4-level autonomous driving software and hardware integration system developed by Pony.ai.
Figure source Pony.ai official account
According to the public information query of Connected Travel, Pony.ai will also cooperate with T3 Travel to deploy Robotaxi’s operations in the Yangtze River Delta region. This also means that Pony.ai’s autonomous driving technology has entered the multi-city passenger vehicle operation test stage.
At the same time, as an operation service provider, Pony.ai will also cooperate with Toyota, BYD, GAC Group, China FAW and other auto companies to apply autonomous driving technology to electric vehicles of various brands.
From a commercial point of view, Robotaxi is the most promising field for autonomous driving. However, at present, the Robotaxi fleet still generally adopts the “safety officer + unmanned driving system” system, and the operating cost is still high. Removing the safety officer is the key to commercialization. The logistics and transportation market space for highway trunk lines is very broad, and the combination of driverless field and heavy trucks is still a blue ocean market.
Because of this, Pony.ai did not give up the layout of driverless trucks due to this business adjustment. According to a LatePost report, Li Hengyu, vice president of Pony.ai and head of the truck department, is still in charge of the business, and will continue to increase investment in the truck business in the future.
Before this business adjustment, Pony.ai’s autonomous driving technology has been extended to the unmanned truck business. During the pandemic last year, Pony.ai conducted self-driving, contactless cargo transportation in Irvine and Fremont, California.
It is worth mentioning that Pony.ai has also joined the field of vehicle manufacturing. According to 36Kr report, Pony.ai has set up a vehicle manufacturing team of more than ten people in Jiading, Shanghai. In the direction of car manufacturing, Pony.ai’s initial plan is to manufacture passenger cars, and on the path of car-making, Pony.ai does not rule out cooperation with car companies.
As Lou Tiancheng, co-founder of Pony.ai said in an interview with LatePost, Pony.ai’s goal is to be a virtual driver.
On the whole, Pony.ai, which has L4-level autonomous driving technology, has implemented autonomous driving technology on multi-brand vehicles, and is trying to find multiple commercial scenarios, has been at the forefront of many autonomous driving companies.
But even as a leading autonomous driving company, Pony.ai has not yet embarked on a road to commercialization that can sustain profits. This is the common dilemma of all independent autonomous driving companies today.
Where is the way out for “Pony.ai”?
If the competition in the first half of autonomous driving is about core technologies, then the competition in the second half is about commercialization.
“The progress of a company depends on the distance between products and commercial applications. This is the only measure of the second half of autonomous driving.” Lou Tiancheng once told Travel Yike.
Han Xu, founder of WeRide Zhixing, once described to Tencent Technology the importance of capital to commercialization. “Before investing in you, you will be asked how to operate, what your business model is, and how to achieve commercialization. Such questions will be asked more and more in 2019.”
After the valuation bubble period in 2017-2018 and the industry cooling-off period in 2019, capital’s attitude towards autonomous driving has become more rational. Focusing on the larger electric vehicle market, intelligence is also an eternal proposition for car companies. As an important part of the intelligentization of electric vehicles, startups with autonomous driving technology are favored by electric vehicle companies.
According to incomplete statistics of online travel, the investors behind Pony.ai are Toyota Motor and China FAW Group, and behind Wenyuan Zhixing is the support of Xiaopeng Motors founder He Xiaopeng. Behind it, there are four car companies including Toyota, SAIC, GM and Weilai. Another new car-making force, Ideal Auto, chose to invest in the unmanned delivery start-up Neolithic.
At this stage, the layout of intelligent driving by car companies is mainly divided into three ways. The self-research group represented by “Weixiaoli”, the cooperation group between traditional car companies and technology companies, and ARCFOX, the technology company provides full-stack solutions for autonomous driving.
Most traditional car companies and new car manufacturers are trying to develop self-driving technology, which means that self-driving startups are just an alternative for car companies. Taking Lixiang Auto as an example, Li Xiang, the founder of Lixiang Auto, publicly stated in 2018 that he would abandon the research and development of L3 autonomous driving, and divide autonomous driving into two products, L2.5 and L4, for simultaneous development. Among them, Li Auto has handed over the L2.5 autonomous driving to Yihang, a startup company invested and incubated by Chehejia to complete the research and development, and the L4 autonomous driving system is completely developed by Chehejia’s own team.
On the other hand, new car-making forces including Tesla, Weilai, and Xiaopeng, as well as car companies such as Volkswagen and Geely, are also focusing on the research and development of L2 and L3 technologies, as well as Robotaxi technology.
In other words, when an independent self-driving company is only the “Plan B” of car companies, it is already in a passive position. With the progress of self-driving research and development of car companies, the commercialization prospects of self-driving companies are worrying.
More importantly, as Chen Mo, the future CEO of TuSimple, once said to the media, “To do this industry, you need patient people.” But commercializing autonomous driving is like a marathon that requires more than determination. In addition to the “learning from each other’s strengths” of autonomous driving companies and car companies, the common problem is that this is a game that burns money.
According to data from research institute PitchBook, autonomous driving startups spend an average of $1.6 million (equivalent to 105.1 billion yuan) per month. Han Xu, CEO of WeRide, also told the media in an interview that the company maintains a spending rate of US$300 million in three years and US$500 million in five years.
In the internal letter of Apollo’s third anniversary, Li Zhenyu, vice president of Baidu, once said bluntly, “Baidu has been in the field of autonomous driving for 7 years and has not made money yet.”
The characteristics of heavy investment in autonomous driving can be seen more intuitively in the performance of the world’s first share of autonomous vehicles, “Tucson Future”. The financial report for the third quarter of this year clearly stated that Tucson’s net loss in the third quarter of this year will be US$115 million, a year-on-year increase of 28.56%. The reason is that although revenue is growing, losses are mainly due to a substantial increase in research and development expenses. According to the financial report, Tucson’s future R&D expenses in the third quarter were US$84.51 million, compared with US$60.04 million in the same period last year, a year-on-year increase of 41%.
This is not only a money-burning game, but also factors such as technology and policy are the main reasons for limiting the commercialization of automated driving.
On the one hand, the implementation of autonomous driving technology is very difficult. As Lou Tiancheng said at a sharing meeting, autonomous driving has not yet reached L4 product-level applications, and the main difficulty lies in technology, including safety, complex scenarios, deep integration of vehicles with complex weather, and technical requirements for vehicle public safety.
On the other hand, policy is also an important factor that makes it difficult for autonomous driving to land. 2018 was the year that self-driving was put on hold when a Volvo XC90 equipped with Uber’s self-driving system struck and killed a middle-aged woman.
After this traffic accident, countries began to formulate autonomous driving laws and policies. At present, countries around the world will still follow the relevant laws of their own countries to assign responsibility for L2-L3 autonomous vehicles, but for L4 and above, including China and other countries in the world are still exploring. This also means that companies with Robotaxi services such as Pony.ai and Weyuanzhixing, even if the feasibility of the technology has been verified, will not be able to implement, mass-produce, and then commercialize until the policy is implemented.
At present, technology, safety, policy, and commercialization have become the mountains in front of autonomous driving companies. Although they have been looking for commercialization possibilities in multiple paths, such as getting involved in vehicle manufacturing, deploying autonomous taxis, minibuses, trucks, etc., the road to commercialization is still long and obstructed.
At last year’s Baidu Conference, Baidu founder Robin Li predicted that “within five years, driverless technology will enter the stage of large-scale commercial use.”
Now it seems that this bold prediction seems a bit radical.
The autopilot scene once imagined in science fiction movies has become the pursuit of many companies, but the autopilot companies that have stumbled and stumbled for many years still have a lot of “money road”. When the dawn comes, who can finally live to that time?