Robotaxi's keyword of the year: vertical and horizontal.
Author | Liu Xiaoqian
Edit | Zhao Jian
In the last month of 2021, Driverless Taxis (Robotaxi) players in the autonomous driving space spoke out intensively:
On December 9, Xiangdao Robotaxi was officially put into operation in Shanghai, jointly developed by SAIC Passenger Vehicles, SAIC Artificial Intelligence Laboratory, autonomous driving solution provider Momenta, Xiangdao Travel and other enterprises.
On December 21, GAC Group made a strategic investment in Wenyuan Zhixing, which in turn strategically invested in GAC's Ruqi Travel, and the three parties will jointly promote the research and development, manufacturing, mass production and commercial operation of robotaxi front-loading mass production models;
On December 24, Xiaoma Zhixing announced that it has reached a strategic cooperation with FAW (Nanjing) to deeply integrate the resources and advantages of both parties in the fields of L4 autonomous driving technology research and development, vehicle research and development and manufacturing, and accelerate the large-scale application of Robotaxi.
These collaborations send a signal that Robotaxi is moving from the road testing stage to mass production, and has a common feature, which includes multi-party cooperation between autonomous driving companies, car companies, and travel operators – some companies call it the "Iron Triangle" model.
Robotaxi is the most imaginative landing scene in the field of autonomous driving, but it has been controversial because of the problem of delayed landing of commercialization. In order to find the key to breaking the situation, more and more Robotaxi companies chose to "combine vertical and horizontal", from fighting alone to the iron triangle era of group fighting.
Will this be the ultimate model for Robotaxi? Who will dominate? Koshi Lightyear interviewed some Robotaxi companies and got two different answers.
In the view of the proponents, Robotaxi's iron triangle model is a multi-win-win ecosystem, and everyone learns from each other's strengths; companies that are not optimistic believe that the iron triangle model is "a model imagined by some autonomous driving companies themselves, and it is not feasible in the market."
This article will serve as a year-end inventory of Robotaxi to explore the rationality and feasibility of the Iron Triangle model.
1. "Iron Triangle" mode: to B and to C
The term "Iron Triangle" was first proposed by Wen Yuanzhixing. In August 2019, Wenyuan Zhixing promoted the landing of Robotaxi with the model of "AI technology company, car company/Tier 1, and travel service platform" trilateral cooperation.
Prior to this, Robotaxi was generally "single-handedly" by the autonomous driving company, and at the same time undertook a complete set of processes such as the establishment and operation of the autonomous driving fleet. The disadvantage of this model is that the autonomous driving company brings together a group of algorithm engineers, the advantage is technology research and development, and is not good at "building the car" and "putting the car on the market", as well as the subsequent vehicle scheduling and operation.
In the iron triangle model, Robotaxi has done a meticulous division of labor: autonomous driving companies (with L4 level autonomous driving solutions) rely on OEMs to establish fleets (providing automotive platforms and parts), rely on travel companies (such as online car-hailing, time-sharing leasing, taxis, etc.) to obtain C-end users, everyone takes their own strengths, and unites the upstream and downstream of the industry to create an autonomous driving ecosystem.
Wen Yuanzhixing's "Iron Triangle" model
Li Maoxiang, a partner of Kaihui Fund, once commented that the iron triangle model is a kind of "borrowing power" to play, "If the automatic driving thing is completely carried on its shoulders, the cost may be magnified by 3 to 4 times, or even far more than that." How to use the strength of others to better complete their own goals is also a manifestation of competitive strength."
In addition to Wen Yuanzhixing, many self-driving companies have similar actions, although they do not use the name of "Iron Triangle".
In August 2020, Shenzhen Yuanrong Qixing and Cao Cao Travel jointly announced that the two sides will conduct a test operation cooperation on autonomous vehicles in Hangzhou, with the goal of providing autonomous driving travel services during the 2022 Hangzhou Asian Games. Yuanrong Qixing also cooperated with Dongfeng Motor to establish an autonomous driving fleet in Wuhan, which will gradually release no less than 200 autonomous vehicles, operated by Dongfeng's travel platform T3 Travel.
This trend reached a small climax in December last year, when several companies announced multi-party cooperation to strengthen the layout of Robotaxi: Momenta - SAIC - Xiangdao Travel, Wenyuan Zhixing - GAC - Ruqi Travel, Xiaoma Zhixing - FAW (Nanjing).
In the Iron Triangle model, the self-driving company stripped Robotaxi's "dirty work" - travel operation work.
Dong Fangliang, executive director of Wenyuan Zhixing, told Koshi Lightyear that the operation includes providing vehicle maintenance, operation and maintenance, safety officer training, as well as boarding and disembarking points, passenger feedback, and serving as customer service roles. Dong Fangliang said bluntly, "The process of realization is very difficult, and there will be many challenges in it. "For example, the current ride-hailing car only needs to manage the driver well, but there is no driver in smart travel, and technology and operation must be closely integrated."
Specifically, there are also subtle differences in the way of operation in the Iron Triangle.
The first model is to set up a separate operating company or platform. Wenyuan Zhixing and Guangzhou Baiyun Taxi Group established a joint venture company Wenyuan Yuexing to achieve the landing operation of self-driving taxis. Wenyuan Zhixing copied the joint venture company to a number of cities, and Wenyuan Yuexing later established Wenyuan Suxing, Wenyuan Yuxing, wenyuan Chuxing, and Baidu launched a special operation platform "Radish Run".
This model is a to C business, by the autonomous driving company to build its own fleet, to provide "unmanned online ride-hailing services", but can only be in a specific area of the road test and operation, such as Baidu Radish Run in Beijing's operating area for Haidian District, Yizhuang, Shijingshan Shougang Park; Wenyuan Zhixing in Guangzhou's operation area for Huangpu District, Guangzhou Development Zone, Haizhu District and so on.
The second model is in-depth cooperation with an independent third-party travel platform, which is operated by the third-party travel platform. It's more of a to-B service that provides "AI drivers" to travel operating companies.
For example, in addition to establishing a joint venture, Wenyuan Zhixing also invested in Gac's Ruqi Travel in December last year; Momenta cooperated with SAIC Xiangdao Travel in addition to building its own fleet in Suzhou; and Yuanrong Qixing cooperated with Cao Cao Travel.
Dong Fangliang, executive director of Wenyuan Zhixing, told "Jiazi Lightyear" that the two cooperation is under the framework of the "Iron Triangle", and the slight difference is that Wenyuan Zhixing and Guangzhou Baiyun Rental Group cooperate, within the scope of operation, users can place orders for Robotaxi through Wenyuan Zhixing's own platform WeRide Go, and at the same time, they can also call Robotaxi on AutoNavi Taxi; Ruqi Travel is a broader online ride-hailing platform, radiating the whole country with the Guangdong-Hong Kong-Macao Greater Bay Area as the center. If Qi Travel joins the driverless service, it can apply the form of hybrid dispatch to bring the driverless car to the mature travel market.
However, this more mature and deepening cooperation model has yet to be implemented, and there is no clear conclusion on whether "self-operation" and "third-party operation" will be unified and integrated.
In addition to the daily operation of "unmanned vehicles", autonomous driving companies should also focus on "technical operations", including data collection and long-tail scenarios. The above get-off point as an example, this detailed problem that is not very encountered in the automatic driving test, in the actual operation will encounter different traffic scenarios, such as temporary parking occupies the drop-off point, in this case how to give passengers a side-stop, which needs to feed back through operational practice.
In last year's article, "Koshi Lightyear" judged that the decisive game in the second half of Robotaxi lies in "operation", because the operator determines the core "nutrient" of autonomous driving - data, the more data the faster the iteration speed of the algorithm, and the cooperation with the travel company not only accelerates the data closed loop of Robotaxi, but also has the opportunity to expand Robotaxi to a more mature market.
The Iron Triangle model is a win-win ecosystem for all parties. However, there always seems to be a difference between ideal and reality.
Although the iron triangle model can achieve "data closed loop", can it achieve "commercial closed loop"? Some people have also questioned this.
2. Distant water is difficult to quench near thirst
The Iron Triangle model is an exploration of robotaxi's commercialization. As we all know, the problem with the commercialization of Robotaxi is that the high cost is difficult to erase. The iron triangle model has been split in the division of labor, so what about the cost?
From public information, we cannot know the details of further commercial cooperation. Several autonomous driving companies that have practiced the Iron Triangle have said to Koshi Lightyear that "the model of commercial cooperation is inconclusive" and "a case by case process".
But a practitioner who did not want to be named said bluntly about "Koshi Lightyear": "The iron triangle model is a model imagined by some autonomous driving companies themselves, and it is not feasible in the market." ”
The reason, he argues, is a simple business factor that constrains it: who pays for the cost.
Let's first look at the cost composition of Robotaxi's bicycle:
Data source: Chebai Think Tank, Soochow Securities Research Institute
The purchase cost of traditional taxis is 100,000 yuan, the driver's salary is 576,000 yuan, plus fuel, insurance and other costs, the cost totals 1.132 million yuan, and the cost per kilometer is 1.89 yuan;
Self-driving taxis with safety officers, the cost of car purchase is 500,000 yuan, the driver's salary is 576,000 yuan, plus fuel, insurance and other costs, the cost is a total of 1.368 million yuan, the cost per kilometer is 2.28 yuan, much higher than that of traditional taxis;
In driverless taxis without safety officers, this cost plummets. The purchase cost of driverless taxis is 200,000 yuan, the driver's salary is 0 yuan, plus fuel, insurance and other costs, a total of 492,000 yuan, the cost per kilometer is 0.82 yuan.
It can be seen that the presence or absence of security officers is a key point in measuring Robotaxi's commercialization model. Under the premise that the safety officer cannot be removed at this stage, Robotaxi is still a loss-making business compared to traditional taxis.
In the cost structure of Robotaxi, driver (safety officer) salary and car purchase cost accounted for the majority, accounting for 42.1% and 36.5% respectively. Among them, the driver's salary will be replaced by the "AI driver" developed by the autonomous driving company, and 36.5% of the "car purchase cost" should be borne by who in the Iron Triangle?
Let's start with car companies. Robotaxi has a market space of up to $2 trillion and is attracting more and more car companies to join in.
In September last year, Mobileye launched a new driverless service brand MoovitaV, the first model selected is the domestic brand Niolai ES8; on November 23, Xiaopeng Motors Chairman He Xiaopeng proposed that Xiaopeng Motors will begin to explore the Robotaxi business in Guangzhou in the second half of 2022; on December 29, Geely said that it reached a cooperation with Waymo, and Extreme Kr will provide exclusive vehicles for Waymo's driverless fleet and put them into commercial operation in the United States.
The above-mentioned industry insiders said that car companies participating in Robotaxi must be driven by short-term interests - selling cars. Because car companies will not directly do the business of service, they will also do it "indirectly" through the establishment of a travel operation company, such as Geely's investment in Cao Cao Travel, and T3 Travel, which is jointly invested by FAW, Dongfeng and Changan.
In some special Robotaxi projects, the cost of car purchase will even be paid by the government, such as Dongfeng's autonomous driving pilot project. In October 2020, Dongfeng established an autonomous driving fleet in the Wuhan Development Zone (Hannan District) and the core area of Wuhan City, with a total investment of 600 million yuan.
Therefore, the cost of buying a car is paid by either the autonomous driving company or the travel operation company.
For the travel operation company, it is a business with very tight funds, making hard money, so there is no incentive to pay. Taking Didi, the largest travel operation company, as an example, the latest Q3 financial report data shows that Didi achieved a net loss of 30.375 billion yuan (US$4.71 billion) and a gross profit margin of only 4.43%.
Therefore, the above-mentioned industry insiders believe that only in the future when Robotaxi is profitable, travel operators will pay money to buy a car, buy a machine to make money, become a similar car rental business, and now it will definitely buy a loss.
And until that day came, self-driving companies built their own fleets were the only option.
This brings us back to the beginning of the story. How expensive is it to build a self-built fleet? Meng Xing, COO of Didi Autonomous Driving Company, once introduced that Didi's autonomous vehicles are equipped with nearly 20 sensors, including 1 64-line lidar, 2 16-line lidar, 7 cameras, as well as millimeter wave radar and ultrasonic radar, and the cost of a car is more than 1 million yuan.
In this way, the autonomous driving company still has not gotten rid of the high cost of Robotaxi, the "far water" of the iron triangle model, and has not solved the short-term commercialization "near thirst" for the time being.
3. Commercialization Breaker: Open Source and Throttling
When will Robotaxi be available on a large scale? This time point will be around 2024 to 2025.
Before that, if the autonomous driving companies that do Robotaxi want to survive to the decisive moment, they have nothing more than two things to do open source and throttling.
The first open source is to raise as much money as possible in the capital market and hoard grain and grass. In 2021, robotaxi's self-driving companies will still be sought after by the capital market.
Among them, Momenta raises two rounds of financing a year, and the investment institutions bring together several major car companies such as Toyota, SAIC, General Motors, Daimler and Tier1 giant Bosch; Wenyuan Zhixing raises three rounds of financing a year, and the investment institutions include Yutong and GAC, two major car companies; Ali and Tencent invest in Yuanrong Qixing and Momenta respectively.
The second open source is to expand more commercialization scenarios.
In the previous article, "Koshi Lightyear" sorted out the trillion market, 100 billion market and 10 billion market of autonomous driving, and in addition to Robotaxi, there are Robobus, Robotruck, low-speed unmanned vehicles and so on.
Han Xu, founder and CEO of Wenyuan Zhixing, has publicly stated: "Whoever masters the real autonomous driving platform technology in the future will be the king." Autonomous driving companies with this powerful technical capability will be able to implement dimensionality reduction strikes. In his view, companies that do low-speed unmanned vehicles and parks today are likely to be hit by a powerful L4 level autonomous driving company that occupies a strategic market in the next 3 to 5 years.
Many Robotaxi self-driving companies are walking on multiple legs:
Around the urban scene, Wenyuan Zhixing has successively launched the manned Wenyuan minibus Robobus and the self-driving light passenger Robovan carrying goods; Yuanrong started to transport goods from the same city on the city road and launched an autonomous light truck (Robotruck); Xiaoma Zhixing set its sights on the trunk logistics scene and developed an autonomous driving heavy truck (Robotruck); Baidu launched an unmanned minibus "Apollon" and personally built a car.
It is worth mentioning that the number of players in Robotaxi is also gradually increasing, and some self-driving companies that originally did "low-dimensional" scenarios pursued the strategy of "rural encirclement of the city" and began to attack the strategic highlands of Robotaxi after the low-dimensional scenes established a foothold.
Through the automatic driving brain, Zhixing has successively landed unmanned sanitation, special vehicles and Robotaxi, and the first batch of Robotaxi restructured and produced by Zhixinger was officially delivered to T3 Travel in December last year; Yishi Technology will take passenger car automatic driving as one of its four major business lines; Qingzhou Zhihang will cut into the city's low-speed scene through Robotaxi technology and launch unmanned minibuses.
The autonomous driving pattern presents a state of "division between the two heroes".
In addition to open source, throttling is also an important part of driving commercialization. Throttling here refers to reducing the cost of autonomous driving solutions.
In June 2021, Baidu Apollo released the fifth generation of unmanned shared vehicles - Apollo Moon, and announced for the first time to the outside world the cost of unmanned vehicles that has been avoided: 480,000 yuan including vehicle costs, only one-third of the average cost of the industry's L4 level autonomous driving models, cutting the cost per kilometer by nearly 60%.
After Baidu, Yuanrong Qixing also disclosed the cost of its own autonomous driving solution. In December 2021, Yuanrong Qixing officially released DeepRoute-Driver 2.0, an autonomous driving solution for front-loading L4 levels. It is reported that the L4 level autonomous driving solution costs less than $10,000, but does not include the cost of the vehicle.
The decline in total costs stems from a decline in hardware costs. According to CICC's forecast, the unit price of lidar is expected to drop to about 500 US dollars / piece in 2023, and the cost of a single lidar will drop to less than 100 US dollars when L3 matures and L4/L5 is introduced into mass production of 10 million units. The forecast mentions that with reference to the famous "Moore's Law" in the development of computers: the performance of computers doubles every 18 months, that is, every year and a half, the price drops to one-half, the same is true for autonomous driving.
In the case of Yuan Rongqi, mechanical lidar was abandoned on the body of the car in favor of MEMS solid-state lidar, 8 cameras, and a lower-cost computing platform.
Does low cost lead to performance degradation? Liu Xuan, vice president of Yuanrong Qixing, told "Jiazi Lightyear": "Probably in the eyes of most people, the imaging effect of solid-state lidar is not as good as that of traditional mechanical lidar, and after some companies adopt this equipment, the performance of the algorithm will be greatly reduced, or even completely unable to run." However, Yuanrong Qixing has done a lot of optimization at the algorithm level, achieving the same level of algorithms while maintaining security and low power consumption and low cost, even beyond the previous one. ”
In addition to the self-exploration of self-driving companies, there is also good news in policy. Some new developments have been made in determining the core element of Robotaxi's commercialization, the security officer.
At the beginning of the new year in 2022, it is reported that Shanghai will formulate an action plan for the development of the intelligent and connected vehicle terminal industry, promote the coordinated development of policies, standards, brands, etc., support pudong's safety officer-free driving legislation, and make good use of industry organizations and third-party institutions such as the International Automobile City, Shanghai Inspection, and Intelligent Vehicle Innovation Platform.
Once the safety officer can be removed, the cost of the driver disappears, and there may be an essential impetus to the cooperation model of the "Iron Triangle". Robotaxi is on the eve of dawn, with wolves and tigers before commercialization, and the market is desperate for confidence to survive the difficult winter.