Shentucar original
Author | dawn
Edit | Wei Jia
The Robotaxi (driverless taxi) industry is buzzing again.
Since the beginning of this year, Baidu's "Radish Run" has landed in two cities in less than a month, Xiaoma Zhixing has completed a new round of financing, SAIC's Robotaxi team has accelerated its expansion, and Xiaopeng Motors is also eager to try. More and more new players are joining the battlefield, and the situation is starting to get complicated.
From a business point of view, Robotaxi doesn't seem to be "sexy" – replacing drivers with systems has not reduced costs, and it is not possible in the short term, let alone making money. The company has lost money and can't make a profit.
But this still can't stop the old players from continuing to invest, and the new players accelerate their entry. The new industry pattern has emerged in 2022.
Classified by origin and style of play, the current mainstream Chinese Robotaxi players can be divided into four categories:
Autonomous driving companies: Xiaoma Zhixing, Wenyuan Zhixing, Yuanrong Qixing, AutoXAI Company: Baidu ride-hailing platform: Didi Car Company: SAIC, GAC, Geely, Xiaopeng
These four types of players have their own tricks, and they have both competition and cooperation with each other, and the combination of vertical and horizontal makes up the Robotaxi jianghu.
In the next few years, before Robotaxi achieved unmanned operation, they were all heavy players in this jianghu. Their iteration speed will determine how often you can get into a real driverless car.
Four Gates
Robotaxi is one of many landing scenarios for autonomous driving. The same is an unmanned vehicle, which can be used to patrol, can also be used to pull goods, and can also be used to carry people, Robotaxi's manned scene, the technical difficulty is the highest.
Baidu was one of the earliest Robotaxi players in China. Its modified self-driving car, which was self-driving on Beijing's highways as early as 2015, reached a top speed of 100 km/h. Since 2018, Baidu has successively obtained the most road test licenses and tested them in Changsha, Cangzhou, Beijing and other cities.
Source / Network
Baidu has always wanted to define itself as an AI company. According to the origin, Baidu's Robotaxi project is considered to be a well-known and decent.
But just because a teacher is famous doesn't mean it can be invincible.
The year after Baidu's unmanned vehicle made its debut on the Beijing Expressway, two core executives left their jobs to start a business. They are Peng Jun, a former executive of Baidu's U.S. R&D center, and Lou Tiancheng, former chairman of Baidu's unmanned vehicle technical committee. They co-founded a company called Xiaoma Zhixing, which later became one of Baidu's biggest rivals in Robotaxi.
In 2017, Wang Jin, who was responsible for the high-speed incident on Baidu's unmanned vehicle mentioned above, and Wang Jin, who was the general manager of Baidu's autonomous driving business unit, also left Baidu, turned around and launched his own autonomous driving project Jingchi Technology in the United States, completing the first closed site unmanned test within five weeks.
If it were not for the fact that He was later sued by Baidu and Forced to leave, Jing Chi might have a duel with Baidu at the Robotaxi Circuit.
Founded at almost the same time as Jingchi, Roadstar, the once most popular self-driving startup, the three founders are Baidu backgrounds, they are from Baidu Silicon Valley unmanned car team, Baidu autonomous driving project team technical committee. The company got the highest valuation in the shortest time that year, but unfortunately it was later liquidated due to team infighting.
These three companies belong to the faction of autonomous driving companies. They all have deep roots with Baidu and have been direct competitors of Baidu.
The industry has always believed that Baidu is the company that has contributed the most to the entire Chinese autonomous driving industry, cultivated countless outstanding autonomous driving talents, and is a well-deserved Whampoa Military Academy, but the embarrassing problem is also that once a certain sector matures, core executives will gradually leave the self-reliance portal.
In addition to the above-mentioned Xiaoma Zhixing and Jingchi, there are also self-driving star companies such as Horizon Robot, Juefei Technology, Heduo Technology, Mainline Technology, Wide Stool Technology, Innovusion, etc., and the founders or core executives are from Baidu.
Liu Wenzhou, director of high-capital investment who focuses on automotive industry investment, invested in projects such as Xijing, Moshi, and Juefei, pointed out to Shentu: "Baidu's theoretical and technical research is undoubtedly excellent, but the in and out of talents is easy to make the accumulation of technology a fault, and it is difficult to maintain consistency. ”
It is precisely because of this that Baidu's Robotaxi project, although it started early, did not open a very obvious gap with other head players, and it had to face more and more enemies.
After Jingchi was divided into three, he was seriously injured, but he did not quit the Robotaxi circuit, and in October 2018, he changed his name to Wenyuan Zhixing, and is now the first camp of autonomous driving companies in China, and is also a radical player in the Robotaxi circuit.
After the liquidation of Roadstar, co-founder Zhou Guang pulled the team to set up Yuanrong Qixing, and then quickly completed nearly $50 million of Pre-A round of financing in 2019, and got an investment from Ali last year.
Baidu and these self-driving startups are the main force of Robotaxi, in addition, the online ride-hailing platform represented by Didi is another major force that cannot be ignored.
Wang Rongjin, a partner at Jinbang Capital, told Shentu that Robotaxi will have an impact on taxis and online ride-hailing cars in the future. "In about 10 years, if Robotaxi successfully lands, it will have a big impact on Didi's valuation, so Didi will definitely do Robotaxi."
Didi began to invest in autonomous driving research and development tests in 2016, upgraded its autonomous driving department to an autonomous driving company in 2019, and opened a small range of Robotaxi free test rides in 2020. However, after Didi went public in 2021, the action of automatic driving slowed down.
Automakers are another discipline. At present, SAIC's Robotaxi project has landed in Shanghai and Suzhou, GAC has made it clear that it wants to establish a Robotaxi team, and Xiaopeng said at the end of last year that it will also enter Robotaxi. These are all new players, they have the ability to manufacture the whole vehicle, and there are direct channels to consumers, and their strength should not be underestimated.
Horse racing
The four major factions conquered Robotaxi, and they all had to get licenses in Kaesong.
In this regard, the earliest starters were Baidu and Xiaoma Zhixing.
In March 2018, Beijing issued the first batch of autonomous driving road test licenses, and Baidu won 5 T3 licenses, which was the highest level of autonomous driving road test licenses in China at that time. Four months later, Pony Also received a T3 license, the first time a startup has been licensed in Beijing. At present, Baidu is the enterprise with the most licenses in China, with more than 300 manned test licenses.
Unlike Baidu, Xiaoma Zhixing has made great efforts in the US market very early, it obtained the California road test license in 2017, is the world's first company to launch Robotaxi business in China and the United States, and later obtained a California nobody license, but this license was revoked in October last year because of an accident, and the Chinese companies that still hold the license are AutoX, Wenyuan Zhixing, Baidu.
In the matter of getting a license, the policies of different places are different, and Robotaxi players are feeling the stones to cross the river.
Liu Wenzhou said to Shentu, "The significance of the number of robotaxi companies accumulating license plates is more at the level of publicity, and the automatic driving license is mainly divided into three levels: test license, there is a safety officer demonstration operation license, no one license, most companies only have a test license, no business license, even if it is a business license, the car must also be equipped with safety officers, and it is too early to achieve real commercialization."
In contrast, road test miles are more important to Robotaxi.
Source / unsplash
The industry regards road test mileage as an important indicator to measure the reliability of autonomous driving technology. The think tank RAND corporation once pointed out in the report that to prove that autonomous driving is safer than humans in a mathematical sense, it needs to test 11 billion miles (about 17.7 billion kilometers), and 100 cars traveling around the clock, taking hundreds of years to complete.
Information released by self-driving leader Waymo in August 2021 showed that its test vehicles had driven more than 20 million miles (about 32.19 million kilometers) on public roads. According to Baidu's data on Shentu, Baidu's actual road measurement mileage on urban roads exceeds 25 million kilometers, ranking first in China.
To accumulate more miles, more vehicles are needed on the road.
"The more vehicles you have, the more you run, the stronger the algorithm iteration, and the precision control will be better than others, which is equivalent to the ability to be fast in the eyes." Wang Rongjin said to Shentu, "It's like 'picking up darts', others throw darts, you keep picking up, you keep being trained, and your ability is stronger than others." ”
However, most players now have only a few hundred fleets, and even baidu, the test fleet is only 500 vehicles. According to data released by the Beijing Intelligent Vehicle Industry Innovation Center, 16 companies in Beijing conducted autonomous driving tests last year, testing 170 vehicles.
Wang Rongjin analyzed that compared with 1,000 vehicles and dozens of vehicles on the road, the difficulty level varies greatly, "the requirements for data processing capabilities and system scheduling capabilities are completely different, and this difference is like the regular army of the group to fight with the guerrillas." ”
Self-driving company AutoX announced in February that its Robotaxi fleet had exceeded 1,000 units. This magnitude has surpassed all Robotaxi companies in China and can even rank first in the world. AutoX said that the 1,000 unmanned vehicles are mainly concentrated in the four major cities of north, shanghai, guangzhou and shenzhen.
Some insiders have questioned this, "1,000 cars, depending on what state of the car, is really put into operation." He added, "AutoX might run a little faster on PR." ”
AutoX told Deepway: "Most of the [fleet] is operational. "Also," the test mileage (in) China is inconvenient to disclose. ”
In the next few years, it is believed that more and more Robotaxi enterprise fleets will exceed 1,000 vehicles. Baidu, for example, announced that it plans to deploy 3,000 unmanned vehicles in 30 cities within three years.
Vertical and horizontal
In the past, RoboTaxi players were immersed in research and development, and there was rarely direct confrontation or cooperation between them. But since the second half of last year, changes have emerged.
The biggest change is that the main engine factories have come down, and the entire industry has begun to engage in vertical and horizontal.
SAIC Motor has an online ride-hailing platform, Xiangdao Travel, which does not have a strong presence in China's ride-hailing market. SAIC also led Series C and C+ rounds of financing from autonomous driving technology company Momenta.
In December last year, SAIC Motor announced the launch of Xiangdao Robotaxi, which was officially launched in Jiading, Shanghai. A person close to SAIC told Shentu that Xiangdao Robotaxi is a combination of SAIC's multi-block business integration, Xiangdao Travel is responsible for operation, SAIC AI Lab and Momenta are responsible for technology, and SAIC can provide the whole vehicle. The industry uses the "iron triangle" to describe this cooperation model.
"When SAIC started the project, it called the SAIC Artificial Intelligence Laboratory, Momenta, and Xiangdao Travel to a meeting, and soon landed the project." The industry insider said.
This is completely different from the previous playing style of Baidu and Xiaoma Zhixing. Xiangdao Travel does not need to start research and development from scratch, and directly grafts Momenta's autonomous driving technology, "equivalent to putting other people's souls on their own bodies, so that they have the ability to drive intelligently." Wang Rongjin commented.
Source / Xiangdao Travel Weibo
Also using a similar style of play is GAC.
GAC Group also has an online ride-hailing platform called Ruqi Travel. In November last year, at the Guangzhou International Auto Show, Ruqi Travel and Wenyuan Zhixing announced a strategic cooperation. A few days later, Wenyuan Zhixing announced that it had obtained a strategic investment from GAC, and the two sides cooperated to upgrade, one of the most important of which was to jointly promote the research and development and manufacturing of Robotaxi front-loading mass production models, and build a robotaxi fleet of tens of thousands of vehicles in the next few years.
This is also the "iron triangle" cooperation model: a strong main engine factory, a ride-hailing operation company, and an autonomous driving technology company.
Two years ago, Geely Group also explored this model. Geely's Cao Cao Travel cooperated with Yuanrong Qixing to conduct the test operation of Robotaxi. However, this cooperation is not tied in multiple dimensions like SAIC and GAC today.
Stronger bundling means higher mobility. The entry of SAIC and GAC will have a great impact on the original Robotaxi industry pattern, and it remains to be verified. One thing is for sure, though, that this will greatly speed up the landing of robotaxi.
Baidu is also forging alliances, such as becoming more aggressive in its cooperation with automakers.
The fourth generation of Baidu Robotaxi models is cooperated with FAW Hongqi, and in the fifth generation of models, Baidu has expanded its cooperation objects to three - Jihu, Aian and Weima, and all of them use front-loading mass production.
Xiaopeng, who is about to enter Robotaxi, is eyeing the data.
According to He Xiaopeng, the founder of Xiaopeng Motors, Xiaopeng will seek cooperation with travel operators to install autonomous driving software and hardware on Xiaopeng mass-produced models to achieve Robotaxi operation in urban scenarios. Its core logic is not to do operations, but to further accumulate driving data and improve the stability and safety of autonomous driving algorithms, and the operation of Robotaxi will be mainly handed over to partners.
Profitability conundrum
Robotaxi has grown for so many years, and until now no company has been profitable.
There are many reasons, such as the technology is not so mature, the law is not perfect enough, the cost of the automatic driving system is too high, and so on. Mainstream players in the industry have launched services for the public, but these services operate in limited scenarios and have not been promoted on a large scale.
Baidu's commercial operation in Yizhuang, Beijing, is an important step, but it is still a restrictive scene, and the richness of the road is very limited, "This is actually a demo-level commercial operation, and it is difficult for enterprises to rely on this to achieve breakeven." Liu Wenzhou said to Shentu.
So now, no Robotaxi company is telling the story of making money, and everyone is expanding the team and running the mileage.
Xiao Jianxiong, founder of AutoX, said bluntly, "Asking a self-driving company 'does it make money' is like asking 'does a primary school student make money', which may force him to work at McDonald's." ”
An industry insider told Shentu that when the US stock market was hot last year, several Robotaxi companies in the domestic head also wanted to take the opportunity to IPO in the United States, but then stagnated due to policy reasons.
Feng Yanjiao, a partner at CIC Insight Consulting, told Shentu, "At least for now, Robotaxi is still a game of capital, and whoever has money can run a little more miles." Robotaxi is looking for diversified monetization paths, and Taxi's revenue alone is negligible. ”
Now, what companies can do is to reduce costs as much as possible.
Last year, the industry made two major progress in cost control, one is the fifth-generation model Apollo Moon cooperated by Baidu and Jihu, which controls the total cost of the body and automatic driving hardware at 480,000 yuan, and the second is that Yuanrong Qixing released a new set of L4 automatic driving solutions, facing front-loading mass production, claiming that the cost of the program does not exceed 10,000 US dollars. This makes mass production of autonomous driving possible.
Source / Apollo - Smart Driving
Baidu announces to the outside world that it will release a generation of autonomous driving platforms every two years, and each generation of products will be ten times higher in performance than the previous generation, and the cost will be reduced by half. Yuan Rong said that the cost of the program will be further reduced to less than 5,000 US dollars in the next two years. This is another form of arms race.
However, profitability and cost are only technical issues. Liu Wenzhou believes that The biggest problem in Robotaxi is ethics and regulations. "Because of the legal attribution issue, in the next five years, autonomous driving will be infinitely close to L2.99, but it will not be able to cross L3." Humans cannot yet accept that machines will decide life and death.
Even if the policy has liberalized demonstration operations and commercialization pilots, vehicles are still not allowed to remove safety officers. An industry insider said that this demonstration operation is more for the promotion of Robotaxi and the accumulation of technology, to show the advanced nature of Robotaxi technology, and to attract people to experience.
"Everyone thinks that robots are safer than people, but no one wants to die in the one-thousandth probability of death of robots, and driving by themselves may have an eight-nine-thousandth probability of death, but everyone will admit their fate, not recognize the machine's one-thousandth." Liu Wenzhou's analysis of the deep way.
Despite the difficulties, there are still investors who are optimistic. Li Wenbo, investment manager of Zhengqi Holdings, told Shentu: "From the perspective of our investment, Robotaxi is definitely the trend of future development, and there is no problem with the track, but who can run out, it is difficult to say, after all, it is an asset-heavy industry, everyone is burning money." ”
"Robotaxi, the track, will be more and more concentrated in the head, and it will be difficult to come up with new companies." Liu Wenzhou said.
*The caption is from @Apollo - Intelligent Driving.