Source of this article: Times Finance Author: Zhang Xu
Tesla CEO Elon Musk's "lightning" visit to China at the end of last month once again sparked discussions and speculation about Tesla's FSD entry into China.
According to China Daily, Tesla has proposed to land a robotaxi in China. In this regard, the Chinese government may first support its domestic test as a demonstration, but has not yet fully approved the full implementation of FSD in China. Times Finance interviewed Tesla China about this, and the other party said that there was no response for the time being.
Tesla Model S.
What conditions does FSD need to land in China, and if it can be landed, will it become a "catfish" again to stir up the intelligent driving industry? And the one that came before FSD is Tesla's driverless taxi.
Robotaxi has been in recent news, and at the first quarter of 2024 earnings call, Musk said that the company will release Tesla Robotaxi on August 8. In China, the two major travel platforms that are sprinting for Hong Kong stock IPOs, such as Qi Travel and Cao Cao Travel, have coincidentally taken autonomous driving and Robotaxi as an important part of the prospectus.
What are the requirements for FSD to land in China?
Tesla FSD is an autonomous driving system, and data collection is an indispensable link.
An autonomous driving engineer told Times Finance that although Tesla's FSD is an end-to-end large model, when it is applied in China, it still needs to supplement the training data of domestic scenarios, mainly information related to traffic rules and traffic signs, and complex intersection data also needs to be re-collected and labeled for adjustment.
As we all know, road data is important because it is related to safety to a certain extent. It can be seen that in order for FSD to land in China, data collection and data storage need to be compliant and localized.
Tesla FSD V12 upgraded to an end-to-end neural network.
It is worth noting that on April 28, the China Association of Automobile Manufacturers announced the "Notice on the Testing of 4 Safety Requirements for Automotive Data Processing (First Batch)", which showed that all the models produced by Tesla's Shanghai Gigafactory met the compliance requirements and passed the four compliance requirements of anonymization of face information outside the car, no collection of cockpit data by default, in-vehicle processing of cockpit data, and significant notification of processing personal information.
After passing four safety compliance requirements, Tesla also said that the ban on driving and parking for Tesla has been lifted in various parts of the country.
In terms of data storage, in August 2023, Tesla responded to the "Sentinel Mode" and clearly pointed out that Tesla has established a data center in China to realize the localization of data storage, and all the data produced by vehicles sold in the Chinese mainland market will be stored in China.
In addition, according to media reports such as the Wall Street Times, Tesla has also partnered with Baidu to deeply customize lane-level maps for use in the FSD system. In the eyes of industry insiders, the use of lane-level map data provided by Baidu can solve the problem of Tesla's surveying and mapping qualifications on the one hand, and accelerate the implementation of FSD in China on the other hand. Regarding the cooperation with Baidu, Tesla China also did not reply.
Since the data is obtained and stored locally, model training must also be carried out locally, and building a local computing center is an indispensable part.
In this regard, Yu Liutian, founder of Yuanzhou Zhixing, told Times Finance: "The main difficulty of Tesla's FSD landing in China lies in the construction of the data center. The cost of rebuilding a data center is extremely high, and these factors can hinder training deployment. ”
Become an industry "catfish" again?
Tesla once led the global electric vehicle industry, and as the second half of the competition, intelligence, especially intelligent driving technology, has once again become the focus of attention.
A number of engineers engaged in the development of autonomous driving fields told Times Finance that Tesla's autonomous driving technology is in the leading position in the industry. Although technologies such as end-to-end neural networks and BEV+Tranforemer have long been available, Tesla is the first car company to realize them and has played a leading role.
"I think Tesla will have a ChatGPT moment, if not this year, I don't think it will be later than next year." Musk's confidence in saying this sentence comes from Tesla's major breakthrough in self-driving technology.
With the update of 2024.3.10, Tesla's FSD has removed the (Beta) logo and changed it to FSD (Supervised), and the concept of FSD V12 has been officially proposed. Tesla's FSD V12 version uses an end-to-end neural network, which is the first end-to-end autonomous driving solution to land, and many decisions are closer to humans than the previous version.
In April this year, Lin Bin, co-founder and vice chairman of Xiaomi, experienced Tesla's FSD V12 version in the United States. He posted: "For the operation of vehicles, FSD is more like people driving. For example, if you turn right at this intersection, the car will stop like a person, and then slowly move the front of the car forward, see clearly that there is no oncoming car on the left, and then accelerate and turn."
Full self-driving is Tesla's ultimate goal. On social media, Musk used a metaphor to express his admiration for self-driving technology: "Betting on self-driving technology is an obvious choice, everything else is just some variation of a horse-drawn carriage." ”
Screenshot from Musk's social media accounts.
At present, Tesla's FSD has been iterated to V12.3 version, and for the upcoming FSD V12.4, Musk announced that drivers will no longer be required to turn the steering wheel. He also revealed that testing of FSD V12.5 and 12.6 versions is underway, capable of coping with more complex scenarios, including navigating narrow one-way streets, as well as dealing with sudden road closures, requiring vehicles to reverse and look for alternative routes.
However, if it is to be implemented in China, it will take time from data collection to retraining, and Chinese car companies are also continuing to increase their investment in intelligent driving.
At the recent Beijing Auto Show, self-owned brand new cars have taken intelligence as one of the cores of publicity. At the launch of Huawei's new intelligent vehicle solution brand, Jin Yuzhi, CEO of Huawei's intelligent vehicle solution BU, pointed out that intelligence has become a key factor for users to buy a car. For example, more than 70% of users choose the high-end intelligent driving version.
In addition, the WEY brand Blue Mountain intelligent driving version was unveiled at the Beijing Auto Show, and the car is equipped with a global intelligent solution for the coffee vehicle, among which, the Coffee Pilot Ultra intelligent driving system is claimed to support the NOA function of no high-precision full-scene NOA; Looking up at the third model, the million-level flagship sedan, looking up at the U7, also ushered in its offline debut, combined with the "Eye of the Gods" high-end intelligent driving assistance system, it is announced that the city navigation function will be opened in the third quarter of this year.
Xiaomi Car SU7.
Xiaomi SU7's urban NOA will also start to be pushed in May. Jiyue Automobile, which also adopts a pure vision solution, plans to achieve urban NOA coverage of the whole country this year, and announced that by the end of 2024, Baidu Maps can use the intelligent driving system wherever it can be navigated. It can be seen that "NOA can be opened nationwide" has become a new focus of competition.
At the same time, the price of high-end intelligent driving continues to fall. On both the two fronts of technology and price, China's smart cars will become a strong opponent of Tesla's FSD in the Chinese market.
Robotaxi news goes on and on
Compared with Tesla's FSD, Tesla's Robotaxi may be one step ahead of Chinese users.
According to China Daily, China may first support Tesla's test of driverless taxis in China, and has not yet fully approved the full implementation of FSD in China. Regarding the entry of FSD into China and the testing of Robotaxi, Times Finance asked Tesla China for verification, and the other party said that there was no response for the time being.
In the first quarter of 2024 earnings call, Musk confirmed that Tesla will release a Robotaxi product in August this year, or named it "Cybercab"; He also said that the company is developing a robotaxi robotaxi, a self-driving taxi, which hopes to achieve mass production in 2024, which will be a "huge boost" for Tesla's future growth.
The project was put on the agenda by Musk as early as 2016, and Musk said in 2019 that in the first phase, the business model of Robotaxi will be similar to online car-hailing and car sharing, and the average cost per mile will be reduced to less than $0.18. In 2022, Musk again said that Robotaxi is expected to be mass-produced in 2024 and remove the steering wheel or pedals.
Morgan Stanley maintained its optimistic outlook for Tesla in a research report in April this year, saying that the launch of Robotaxi symbolizes the transformation of Tesla's business model to hardcore AI and robotics technology. They predict that Tesla's mobility business revenue will reach $17 billion by 2030, directly driving the growth of total revenue.
In addition to Tesla, there has been a lot of news about the Robotaxi industry recently.
On April 29, Cao Cao submitted a prospectus to the Hong Kong Stock Exchange. Cao Cao said in the prospectus that part of the funds raised will be used to invest in autonomous driving, and will cooperate with Geely Group to deploy vehicles with a front-loaded L4 autonomous driving system in the future, and is working with business partners to launch a driverless taxi program.
In September 2022, Cao Cao and Geely Automobile Innovation Research Institute's Intelligent Driving Center launched a cooperation on the Robotaxi project.
In Cao Cao's view, in the long run, autonomous driving is expected to drive the structural transformation of China's shared mobility market, and the advancement of autonomous driving can increase the supply of compliant vehicles and create huge room for cost reduction.
Ruqi Travel, which submitted the form to the Hong Kong Stock Exchange for the second time in March this year, also put self-driving taxis in an important position.
At present, Ruqi Travel has realized the hybrid dispatch mode of Robotaxi and manned online car-hailing, and the service will be launched in Nansha, Guangzhou and Shenzhen in 2022 and early 2024 respectively, and it is predicted that the commercialization of Robotaxi will be realized around 2026 and enter the mature commercialization stage by 2030.
According to the prospectus of Ruqi Travel, Robotaxi and ride-sharing services, marketing and promotion services are classified as "other" segments. From 2021 to 2023, the revenue of this sector will be 7.276 million yuan, 4.812 million yuan, and 2.000 million yuan respectively, accounting for 0.7%, 0.3%, and 0.1% of the total revenue. Although the sector is not losing money, Ruqi's Robotaxi business is still far from being independent.
At present, the domestic Robotaxi competition landscape is still changing. Huafu Securities Research Report said that the development of the Robotaxi industry needs to be jointly promoted by all parties. The "Golden Triangle" cooperation model between automakers, autonomous driving technology companies, and mobility service operators is currently the mainstream. Among them, autonomous driving technology companies drive the commercialization process of Robotaxi, and having mature L4+ technical support is the key to the implementation of Robotaxi.
Screenshot of the promotional video of Ziruqi travel.
On April 7, Didi Autonomous Driving Co., Ltd. and GAC Aion jointly invested in the establishment of Andi Technology. Andi Technology has become the first joint venture established in China between a L4 autonomous driving company and a car company to build Robotaxi production vehicles, and its first commercial L4 model is scheduled to be launched in 2025.
On April 26, Pony.ai and Toyota jointly established a new company, Lifeng Intelligent Technology, which plans to launch 1,000 units of 4X L4 driverless Robotaxi in the Chinese market in the first phase, and provide driverless Robotaxi travel services in major first-tier cities through Pony.ai's operating platform.
Zhou Xiaoyun, director of Toyota China's foreign strategic cooperation business, said that Toyota and Pony.ai decided to establish a joint venture to promote the commercialization of highly autonomous vehicles in the future. "We believe that the current technology has reached a state where it can be used for commercialization."
Policies have also helped Robotaxi. On May 1, the "Hangzhou Intelligent Networked Vehicle Testing and Application Promotion Regulations" was officially implemented, and the main central urban areas of Hangzhou are used as intelligent networked vehicle testing and application areas. Hangzhou has also become the first city in the country to clarify the specific process of autonomous vehicles on the road with local legislation, and it is also the first city in the country to legislate for low-speed unmanned vehicles.
If Tesla Robotaxi can be tested in China, it will undoubtedly play a role in verifying the reliability and feasibility of Tesla's FSD technology. An expert in the field of transportation told Times Finance that one of the major effects of Robotaxi is that "compared with non-commercial vehicles, commercial vehicles accumulate data much faster, and more data can be collected in the same time for model optimization." ”
With the rapid iteration of FSD software, the answer to whether Tesla Robotaxi has the ability to fully drive will be revealed on August 8.