On October 24th, Tesla released its third quarter 2024 financial report after the US stock market closed. During the conference call, Tesla CEO Elon Musk revealed more about Tesla's Robotaxi business and progress and plans for the Cybercab model after a recent press conference.
This year, the Robotaxi industry has been booming, with "mass production" becoming the key word in the industry. At the beginning of the year, Baidu released a new generation of mass production models of Apollo Go. In October, Wenyuan Zhixing also released a new generation of Robotaxi mass production model GXR jointly built with Geely Remote. Xiaoma Zhixing has established a joint venture with Toyota China and GAC Toyota to jointly develop and produce the first Platinum 4X Robotaxi model. Production preparations have already begun, and it is expected to deploy over a thousand Robotaxis to first tier cities from 2025 to 2026.
In October, Tesla also released its long delayed Robotaxi product Cybercab. However, after the press conference, Tesla's stock plummeted overnight, dropping 10% at one point during trading, marking the largest decline in over two months. It was not until the release of the third quarter financial report that there was a significant rebound.
In the eyes of the industry, is Tesla's Robotaxi just a name for itself, and can it successfully make waves in the industry?
At the press conference, Musk showcased Tesla's long-awaited Robotaxi product Cybercab to the public. This model adopts a dual wing door design, which is different from the designs of other autonomous taxi companies. The car only has two seats and introduces induction charging technology, giving it a futuristic design.
Tesla Cybercab Image Source: Tesla
Musk introduced that, as Tesla had previously envisioned, Cybercab has eliminated the steering wheel and pedals and will be equipped with AI 5 computing hardware. According to previous information, AI 5 will use a 4nm process, which will increase energy consumption by 5 times and computing power by 10 times compared to the latest HW 4.0. According to product detail pictures released by Tesla, the Cybercab comes with a central control screen, no rearview mirrors, and a large hatchback trunk. There are still 20 Cybercabs available for the audience to test ride on site.
The Robotaxi project was put on the agenda by Musk as early as 2016 and further refined in 2019. At the press conference, Musk mentioned that thanks to autonomous driving technology, self driving taxis can reduce travel costs for users and also lower operating costs for operators. Robotaxis equipped with FSD (Full Self Driving) are about 10 times safer than human driving.
In terms of business model, Musk's vision is that Tesla will not only build its own fully automated Robotaxis fleet, but also allow Tesla owners' vehicles to join the shared fleet, providing travel services for other users as autonomous vehicle in their spare time. And, one person can actually manage a convoy, 'just like shepherds taking care of their flocks'.
Over time, Cybercab's operating cost per mile is only around 20 cents, including taxes and other fees, which is about 30-40 cents per mile. As a comparison, the current average cost of ride hailing travel in North America is $1 per mile. And Cybercab's estimated vehicle cost will be less than $30000.
In addition to Robotaxi, Musk also showcased the autonomous box car Robovan on site, which can carry 20 people and transport goods without a steering wheel or pedals, completely autonomous. The operating costs of Robovan can even be as low as 5 to 10 cents.
After the press conference, Tesla's stock price plummeted overnight, indicating that the capital market still has doubts about Musk's story. Vague plans and timelines are one of the reasons why the market remains cautious, and Musk refined this during the earnings call.
According to its latest disclosure, Cybercab will achieve its goal of producing at least 2 million vehicles per year by 2026, and may ultimately reach 4 million vehicles per year. The model will continue to use the "Unboxed" manufacturing strategy, breaking down the car into multiple independent modules such as front and rear parts, chassis, battery modules, etc. These modules will be assembled simultaneously in different areas of the factory, and finally assembled into a complete vehicle. By adopting this approach, the cost reduction is greater.
Robotaxi business will be launched first. Although Cybercab has not yet achieved mass production, Tesla has provided Robotaxi travel services to employees in the San Francisco Bay Area in the United States. Employees can take a taxi through the app, and the car is still equipped with safety personnel. The vehicles used are existing models such as Model 3/Y.
Musk expects Tesla to begin paid autonomous taxi services in California and Texas next year. Tesla will initially provide services using existing models and gradually shift towards Cybercab.
Tesla is not the only player on the Robotaxi track. In fact, in the United States, Waymo, the parent company of Google Alphabet, and Cruise, the autonomous driving company of General Motors have taken root in the Robotaxifield for many years. In China, Baidu Apollo Go, Xiaoma Zhixing, Wenyuan Zhixing and other companies also have long-term operating experience.
Compared to Tesla's uncertain timeline, these companies' Robotaxis have been on the road for many years and have entered commercial operation. The mass-produced vehicles have also undergone multiple iterations, far ahead of Tesla's schedule.
According to the data of Yole Group, a market research organization, in 2024, Waymo's paid rides experienced exponential growth, and now provides more than 100000 paid rides per week, while Baidu Apollo Go takes about 75000 paid rides per week.
The reason why Tesla is favored is because of its unique technological path and many advantages it possesses.
The core difference between Tesla and other autonomous driving solutions is that Tesla adopts a pure visual solution without high-precision maps, which eliminates the need for expensive LiDAR as a sensing device at the time and saves high high-precision map usage fees, thus greatly reducing costs.
Tesla's implementation path for L4 is also different from other autonomous driving companies, as it does not adopt a one-step approach to L4 research and development, but rather a gradual path from L2 level assisted driving to L4 level autonomous driving. FSD was first installed on Tesla cars as an assisted driving system, and through iterations, it eventually reached the L4 level.
In the past two years, Tesla has made breakthrough progress in FSD. The latest V12 version of FSD introduces "end-to-end neural network" technology, replacing rule-based code with neural networks. The end-to-end model integrates perception, prediction, and planning models into one, without the need for lengthy code to formulate rules. Instead, it trains the system with massive amounts of data, enabling machines to have the ability to learn, think, and analyze autonomously, and better handle complex driving tasks.
FSD Beta v12 upgrades the city street driving stack to an end-to-end neural network, which has been trained on millions of videos and replaced over 300000 lines of C++code. This is how Tesla introduced V12 in its update log.
Musk said in a conference call that the FSD V13 version is about to be released, and compared to the 12.5 version, it is expected that the intervention interval mileage will increase by about five to six times. From early 2023 to version 12.5, the intervention interval mileage of FSD has increased by 100 times. Tesla's internal expectation is to make the intervention interval of FSD longer than that of humans and make FSD safer than human drivers by the second quarter of 2025.
This brings hope for Tesla to achieve the ultimate L4 autonomous driving, and many domestic and foreign autonomous driving companies have also announced their investment in this new direction.
But with these advantages, Tesla's autonomous driving also faces many challenges to truly land.
For example, although hardware costs can be reduced, the requirement for software development will be correspondingly higher without high-precision maps and purely visual perception solutions. Xiaomo Zhixing told Interface News that the LiDAR solution can better improve the stability of perception and compensate for some issues with visual perception algorithms. The prerequisite investment for pure visual perception solutions will be greater, and these will need to be amortized with mass production in the future.
The purely visual route still faces controversy in terms of technology. Zhang Ning, Vice President of Xiaoma Zhixing and Head of Robotaxi's Autonomous Driving Travel Business, told Interface News that Xiaoma Zhixing currently uses a fusion sensor solution, including laser radar, cameras, millimeter wave radar, etc. For non motorized vehicle objects, pure visual solutions have inherent flaws in detection, especially in domestic urban road scenes with a high number of pedestrians and bicycles. If there is a situation where objects on the road cannot be seen, the autonomous driving mode will not slow down or avoid them in a timely manner, resulting in safety hazards.
For Tesla, the software capability of FSD, as a production advantage for new energy vehicle companies, coupled with its powerful technology, data, and computing power, are all essential factors that make its Robotaxi flywheel turn. The most crucial aspect is that if FSD itself cannot reach L4 level and be approved for use on the road, it will be impossible to achieve the landing of Robotaxi. At present, it remains to be verified whether end-to-end can lead Tesla to a commercially viable L4.
Bernstein analysts pointed out that Tesla's current FSD service is still at the L2+level. Zhang Ning also mentioned to Interface News. Tesla FSD performs well in L2 assisted driving, but there is still a significant gap in terms of technology and safety indicators required for L4 autonomous driving. According to the evaluation report by third-party testing agency AMCI Testing, Tesla's latest FSD v12.5 version requires human intervention on average once every 13 miles (approximately 21 kilometers). Xiaoma Zhixing also conducted road tests on this version in relatively busy Bay Area cities in the United States, and the test results showed that it actually takes over on average every less than 10 kilometers.
At the Cybercab press conference, the on-site audience can freely test ride the 20 actual vehicles provided on site. However, the closed roads and relatively simple routes on site are not sufficient to verify the actual level of FSD loaded on Cybercab.
Regulation is another challenge for Tesla. On October 19th, the US National Highway Traffic Safety Administration (NHTSA) launched an investigation into Tesla's FSD. NHTSA stated that it has identified four incidents of FSD system operation, including a fatal accident in 2023, all of which occurred in low visibility conditions on roads. Therefore, the NHTSA Office of Defects Investigation (ODI) will evaluate the engineering control testing and appropriate response capabilities of FSD to reduce road visibility.
Not only is the safety of FSD being questioned, but obtaining relevant regulatory permits is also required for public testing and operation of Robotaxi. Due to the varying laws regarding autonomous driving in different states of the United States, convincing local governments and obtaining testing licenses from various locations also requires a significant amount of time.
The advanced design of Cybercab may also bring potential trouble to Tesla. Due to the cancellation of manual control devices such as steering wheel, pedals, and rearview mirrors, regulatory authorities are unlikely to approve these vehicles on the road. In 2022, Cruise, a self driving company under General Motors, submitted a petition requesting permission for its self designed autonomous vehicle Origin to be put on the road without brake pedals and rearview mirrors, but it was ultimately not approved.
From this perspective, it is still questionable whether Cybercab can achieve mass production as Musk claimed in 2026.
Consulting firm Frost& Sullivan predicts that the global mobile travel market will reach $4.5 trillion by 2025 and further grow to $4.7 trillion by 2030, calculated by GTV (transaction volume). Driven by technological advancements, supportive policies, and reduced hardware costs, Robotaxi services are expected to achieve commercialization around 2026 and enter a mature commercialization stage by 2030.
From a global perspective, China and the United States are still the regions with the fastest development of Robotaxi, leading in terms of technological development level, service application, policy promotion, number of enterprises, and industrial chain development.
The United States started early, but China is moving faster, "said an industry insider to Interface News, comparing the progress of autonomous driving between China and the United States. At present, the leading level of autonomous driving is the all day normal operation of fully autonomous driving, which only a few companies in the world can achieve, with Chinese companies accounting for the majority.
Xiaoma Zhixing told the interface news that compared with other overseas companies, China has more abundant test scenarios and road conditions, which is an inherent advantage for technology iteration. The auto drive system may perform better in handling rare and complex scenarios. Moreover, China's relatively open policies on autonomous driving have also prompted the earlier development of Chinese Robotaxi companies. L4 autonomous driving is an industry that places great emphasis on research and development investment. Chinese companies have accumulated more experience and invested more resources, and have also accumulated better technological barriers.
According to the latest developments, Xiaoma Zhixing already has a fleet of over 250 autonomous taxis in operation, among which Beijing, Guangzhou, and Shenzhen have started commercial charging for fully autonomous driving services; Wenyuan Zhixing has accumulated over 1800 days of public operation experience and operates the largest Robotaxi fleet in the Middle East in five cities at home and abroad, including Guangzhou, Beijing, Abu Dhabi, Nanjing, and Ordos; At the beginning of the year, Baidu Apollo Go officially announced the sixth generation of mass production unmanned vehicles. The cost of these latest vehicles is only 200000 yuan, 60% lower than the cost of the previous generation of complete vehicles. It plans to expand the unmanned driving fleet in Wuhan to 1000 vehicles by the end of the year, more than three times the number at the end of last year. It is expected that Apollo Go will achieve a break even in Wuhan by then.
The new generation Robotaxi GXR from Wenyuan Zhixing
In addition to autonomous driving companies, ride hailing companies are another major participant in the Robotaxi industry. Didi Chuxing has been laying out its autonomous driving business for a long time and has also invested the most in autonomous driving among domestic travel platforms. At present, Didi Autonomous Driving has over 200 autonomous vehicles in cities such as Beijing, Shanghai, and Guangzhou, and has achieved mixed dispatching in the operating areas of Shanghai and Guangzhou. In April of this year, Didi Autonomous Driving and GAC Aion's joint venture company, Andi Technology, have been granted a business license. Both parties stated that they will integrate autonomous driving technology and new energy vehicle manufacturing experience to launch the first mass-produced L4 model by 2025.
How do these companies view Tesla as a "catfish"? In fact, most domestic companies have told Interface News that they welcome Tesla's entry into the Robotaxi market. Industry insiders generally believe that Tesla's entry can boost industry confidence and once again push Robotaxi in front of the public.
Xiaoma Zhixing also told Interface News that compared to other more mature industries or markets, autonomous driving, especially L4 level autonomous driving, is still a blue ocean market and is still in the cultivation period of the industry, far from reaching the stage of fierce competition. Faced with the vast global market, more players have the opportunity to promote the development of autonomous driving technology in more regions besides China and the United States, in response to issues such as aging populations and restricted travel.