Robotaxis are coming. Can Chinese companies take the lead?
On June 22, 2025, Tesla officially launched its driverless taxi (Robotaxi) service. Ten Model Y vehicles (instead of the previously exposed driverless taxis) were transformed into Robotaxis and put into use. The test area is a small area in the southern part of Austin, USA. The platform has a geofence, and it takes about 30 minutes to drive from one end to the other.
(Source: Internet)
Currently, the service is only available to invited users. The fare for each trip is tentatively set at a fixed $4.2, and the operating hours are from 6 a.m. to midnight. For safety reasons, a safety officer is arranged in the front passenger seat of the Robotaxi. In addition, the Robotaxi is also equipped with a remote monitoring system. Users can press the Support button on the rear - seat screen of the vehicle to communicate with remote service personnel and ask relevant questions.
After some bloggers and local users experienced it, they found that it is similar to an ordinary taxi service. Users need to log in to the Robotaxi App within the service area to confirm the destination, place an order, wait for the car, confirm the vehicle number upon arrival, and press "Start" on the APP after getting in the car to start the journey. Users can see the real - time route map in the car, and they can also watch movies, listen to music, and play games.
Tesla founder Elon Musk tweeted, "This is the achievement of ten years of hard work by everyone." According to Musk, Tesla's artificial intelligence chip and software team was founded from scratch within Tesla. The successful launch of the Robotaxi means the start of its commercial operation, "and it also represents that the visual + end - to - end model technology route adopted by Tesla's driverless technology has been fully proven."
Actually, Musk had long said that Tesla's driverless taxis would be launched in the United States in June, and by 2026, the fleet size would exceed 100,000. In addition to selling smart electric vehicles, driverless taxis and humanoid robots support Tesla's market value of over one trillion US dollars.
However, after watching the actual test of the Robotaxi, Blue Shark Financial Media believes that it is still in the early stage, and the operating scope is also very small (it will not leave Austin until the end of 2025).
In August 2022, Apollo Go achieved driverless autonomous driving travel services in Chongqing and Wuhan (with a remote operation team). As of May 2025, the number of Apollo Go's orders had exceeded 11 million, covering 15 cities around the world, more than the 10 million orders of Waymo, the driverless taxi service under Google.
Market Scramble
Judging from the number of current operating cities, the operating scope, the total number of orders, and even the cost per vehicle (less than $30,000 per vehicle), Apollo Go is undoubtedly the world - leading driverless taxi company.
However, just as Tesla plays the role of a key changer and leader in the smart electric vehicle track (it launched the Model 3 in 2016, with over 400,000 global orders at that time. The construction of Tesla's Shanghai Gigafactory in 2019 quickly detonated the global new - energy vehicle market and promoted the rapid evolution of Chinese smart car companies). Now, its official launch of driverless taxis means that this market will see explosive development (it is expected that the fleet size will reach 100,000 in 2026, while in 2025, Apollo Go and others only have a fleet of more than 1,000 vehicles). Can Chinese driverless taxi companies lead the way?
(Source: Apollo Go official website)
1. Both Baidu's Apollo Go and Google's Waymo adopt the technical route of multi - sensor fusion (lidar, millimeter - wave radar, etc.) + high - precision maps (taking advantage of their strengths in mapping) and prioritize safety. This is safer than Tesla's pure - vision route, but the cost is higher (Waymo's vehicles cost hundreds of thousands of dollars each).
Tesla and XPeng Motors dare to adopt the pure - vision driverless route. First, it is a cost consideration (lower cost). Second, they believe that just like humans, autonomous driving can solve the problem of safe driving with pure vision. With the explosion of large - scale AI model algorithms and the significant reduction of inference costs, their algorithms and computing power do have the opportunity to help pure - vision achieve good driverless results. China has the world's largest smart car industry chain (batteries, sensors, complete vehicles, etc.), which allows Apollo Go, even with the more expensive lidar, to further reduce the (per - vehicle) cost of driverless taxis. Baidu has an advantage in this regard.
2. Form alliances to collect data. Although Tesla's Robotaxi project has just been launched and the order volume is not large, its driverless system uses FSD (Full Self - Driving) - an autonomous driving system pre - installed in all Tesla models. By 2024, Tesla had sold a cumulative total of 8 million electric vehicles. These vehicles are constantly on the road, accumulating driving data (standard and real - time) from various countries and regions around the world. Data is the core for the accelerated evolution of large - scale models, which is Tesla's unique competitiveness.
To collect more data for better training of its driverless solutions, Waymo not only builds its own fleets in cities such as San Francisco and Los Angeles in the United States but also cooperates with Uber and others, using the massive orders on their platforms for testing.
Previously, Baidu Group founder Robin Li revealed that Apollo Go is seeking cooperation with partners such as mobile operators, local taxi companies, and third - party fleet operators (still aiming to expand the user base and collect more data). Recently, Apollo Go reached a strategic cooperation with CAR Inc. and will launch the world's first autonomous vehicle rental service. This year, Apollo Go has started public road verification tests in Dubai (Middle East), and the test area in Hong Kong has also been further expanded. On June 20, Bloomberg reported that Apollo Go is accelerating its layout in the Southeast Asian market. Previously, the media reported that Apollo Go was also exploring entering European markets such as Switzerland and Turkey.
On June 16, Chinese autonomous driving technology provider WeRide announced a tri - party cooperation agreement with Uber and the Dubai Roads and Transport Authority (RTA) to launch commercial Robotaxi services in Dubai.
3. Who will see the dawn of profitability first? The field of driverless (taxi) is considered by NVIDIA founder Jensen Huang to be one of the most important directions in the next 10 years, on par with robotics. Goldman Sachs predicts that the global Robotaxi market will explode in 2025, and the market size is expected to reach $40 - 45.7 billion by 2030, with a compound annual growth rate (CAGR) of over 60%. Since Google began exploring autonomous driving in 2009, Baidu was the earliest domestic player to heavily invest in autonomous driving. In 2024, Apollo Go started large - scale operations. In the first quarter of 2025, Apollo Go provided over 1.4 million travel services globally, a year - on - year increase of 75%. Based on Apollo Go's per - vehicle cost advantage, technological maturity, and operating scale, Robin Li said that he has "seen a clear path to profitability." Waymo, with a parent company (Google) having a market value of over $2 trillion, has a per - vehicle cost of hundreds of thousands of dollars, and profitability is still far away (it can afford to lose). Tesla's Robotaxi operating scale is too small and is still in the testing stage. How can it talk about profitability? Autonomous driving technology providers such as WeRide, TuSimple, and Pony.ai make money through 2B services, but as of 2024, they are all in a state of huge losses.
Driverless Rush
From Tesla to Apollo Go, they all depict a beautiful vision of driverless taxis - cars will no longer be a large depreciating item (cost). Owners of driverless taxis can let the car pick them up when going to or from work. They don't need to control the vehicle or worry about traffic jams and can handle work, watch movies, listen to music, etc. in the car. When working or resting, they can let the driverless taxi go on the road to pick up passengers and make money for the owners.
However, from Tesla to Apollo Go to Waymo, the most urgent problem to solve at present is to obtain permits (licenses) for driverless taxi testing and operation from various countries and cities.
Although data from Apollo Go in the past two years shows that in China's complex road conditions, the actual accident rate of Apollo Go's vehicles is only 1/14 of that of human drivers; in the face of this new thing of autonomous (driverless) driving, no country or local government dares to fully liberalize it easily.
The Chinese government clearly realizes that China has leading large - scale AI models (DS), a complete and large - scale new - energy vehicle upstream and downstream industrial chain, high - level intelligent new - energy vehicles, relatively standardized data (traffic rules and people), and easy data collection (just need to drive). It has the internal and external conditions to lead the world in driverless technology (one of the most important application scenarios of artificial intelligence).
On April 1, 2025, the first local regulation on autonomous driving in China, the "Regulations on Autonomous Vehicles in Beijing," officially came into effect. L3 - level vehicles can be piloted on main roads within the Fifth Ring Road, the airport expressway, etc. The liability subject is clearly transferred to the car manufacturers for the first time. If the system fails to give timely warnings or malfunctions in the autonomous driving state, the car manufacturer needs to bear the compensation liability, and the driver is only responsible when failing to respond to the takeover request.
(Source: Beijing Municipal People's Government)
Even earlier, on March 1, Wuhan began to implement the "Regulations on Promoting the Development of Intelligent Connected Vehicles in Wuhan," allowing L3 - level vehicles to be piloted in areas such as the Yangtze River Main Axis Smart Corridor.
The Chinese Ministry of Industry and Information Technology plans to issue certification standards for L3 - level autonomous vehicles in 2025, including hardware configurations such as lidar, high - precision maps, and computing power (needing to exceed 1000 TOPS), as well as test processes such as extreme weather simulation verification.
The policies introduced by Wuhan, Beijing, and the Ministry of Industry and Information Technology are obviously beneficial to driverless taxi companies such as Apollo Go.
Apollo Go released the world's first large - scale model for L4 - level autonomous driving, Apollo ADFM, in May 2024. It ensures the stability and reliability of the vehicle through a 10 - fold safety redundancy scheme and a 6 - fold MRC safety strategy, which is 10 times safer than human driving. Apollo Go has previously launched driverless taxis in super - first - tier and complex traffic environments such as Beijing, Shenzhen, and Wuhan. It can handle various complex situations, including diverse pedestrians, cars, bicycles, bus lanes, narrow and complex roads, construction sections, and narrow roads at night, and can quickly cover complex urban - level scenarios across the city. This accumulation of domestic experience enables Apollo Go to gain trust and quickly launch services in overseas markets such as Dubai and Hong Kong.
However, those targeting the autonomous driving (driverless taxi) market include not only Apollo Go but also autonomous driving technology providers such as WeRide and Pony.ai, as well as mobile travel giants such as Didi Chuxing, and automotive giants such as BYD and Geely (with Caocao Chuxing under its umbrella). If Apollo Go wants to maintain its leading position, in addition to continuing to increase investment in technological R & D, it also needs to get approval from more national and regional governments and obtain more testing licenses (permits); cooperate with more partners, such as local taxi companies, travel platforms like T3 Chuxing, ride - hailing companies that provide operational support for Didi and Gaode, and hotel and travel platforms like Ctrip and Qunar. This is a test of on - the - ground promotion and operational execution.
This article is from the WeChat official account "Blue Shark Financial Media". Author: Wei Qiang, Editor: Lu Xucheng. Republished by 36Kr with permission.