Text | Lao Yuer, Editor | Yang Xuran
All it takes is saying "I want to go somewhere" to your smartphone or other smart terminals, and the AI will automatically understand your travel intent, help organize your itinerary, plan the route, and directly book a ride-hailing service. When you step out of your home, a Robotaxi is already waiting by the roadside. Without needing to operate any APP, the vehicle automatically takes you to your destination.
Such AI-driven mobility, which once only existed in sci-fi movies, is now taking its initial shape on the streets of Chinese cities.
According to the "2025 China Mobile Mobility Market Data Report" by the E-Commerce Research Center of NetSun, the scale of China's ride-hailing market in 2025 was approximately CNY 411.9 billion, a year-on-year growth of 6.07%, with the growth rate continuing to narrow compared to 8.19% in 2024. The domestic ride-hailing user base remained at 539 million, flat with the previous year. User growth has peaked, and the traditional traffic dividend is basically exhausted.
As the entire industry begins to bid farewell to the golden age of rapid growth, the ride-hailing sector must seek new growth stories. AI Agent-driven intelligent mobility is the main thread of this new story.
Over the past three years, the AI wave brought by large models has triggered a series of changes in the mobility track. The industry has transitioned from the stage of technical experiments within closed parks to the formal stage of commercial implementation.
On a deeper level, this transformation brings about the disintegration and reconstruction of the entire mobility industry chain, as well as the rewriting of the power structure at every link in the chain. A new decisive force is already brewing quietly.
This article is an in-depth value piece from the content team of "Juchao WAVE". You are welcome to follow us on multiple platforms.
Closed Loop
In September 2026, Cao Cao Mobility densely completed multiple rounds of cooperation implementation with smart terminals. This may currently be the most active and aggressive mobility company in the AI field:
On September 14, the consumer version of the Doubao Phone Assistant was officially released, equipped on the Nubia NaviX Ultra smartphone. Cao Cao Mobility became one of the first mobility service providers to integrate with this system;
On September 15, Cao Cao Mobility's AI ride-hailing service was integrated into Honor's YOYO AI Agent. Users can directly tell YOYO their ride needs, and the AI Agent can automatically identify the start and end points, match capacity, plan vehicle types and routes, and complete the order;
At the OPPO Developer Conference on September 17, Cao Cao Mobility officially announced again that its AI ride-hailing service is integrated into OPPO's Xiaobu Next, completing a full-process demonstration on-site;
On September 22, at the Alibaba Apsara Conference, Cao Cao Mobility announced a partnership with Qwen AI Glasses. Both parties will jointly launch an AI ride-hailing service to achieve a hands-free near-eye interactive experience;
Prior to this, Cao Cao Mobility's ride-hailing capabilities had also been integrated into Huawei smartwatches and Xiaomi's miclaw phone assistant.
To many ordinary users, these cooperations may seem like just adding a small voice-activated ride-hailing feature. However, viewed from the dimension of the entire Chinese ride-hailing industry, their significance is much greater: these partnerships moving towards practical commercial implementation mark that the industry chain of AI mobility has initially formed a complete closed loop.
Looking back at previous AI transformation attempts by various institutions in the mobility field, most have stalled at certain stages of "semi-AI transformation."
Taking the pilot operations of Robotaxis in various regions as an example, the vehicles themselves have already achieved L4 autonomous driving, with the driving process handed over to AI. However, for users to hail a ride, they still need to open an APP, manually enter the address, and confirm the order.
Subsequently, Qwen's integration with Amap Ride and WeChat's AI Agent linking with DiDi both realized voice ordering functions. However, users still need to trigger operations within the APP or the WeChat ecosystem and cannot break away from the application carrier.
In contrast, the deep linkage between Cao Cao Mobility and smartphone manufacturers' system-level AI agents has substantively detached mobility services from standalone APPs and embedded them into the natural dialogue between users and AI agents. Users do not need to open any mobility software; they only need to converse with the AI on their phones to complete the entire ordering process.
This shift in interaction paradigm provides the foundation for a complete closed loop in AI mobility:
At the front end, terminal AI agents such as smartphone AI assistants and AI glasses serve as entry points, catering to users' mobility demands at any time.
AI large models have become the key to this. Taking the smartphone sector as an example, manufacturers such as Huawei, Xiaomi, OPPO, vivo, and Honor are all pursuing the route of self-developed large models. Meanwhile, smartphones like Nubia and Apple adopt an external cooperation model. Doubao, Alibaba Qwen, and Baidu Wenxin have become popular cooperation choices. This represents the closest distance for large models to serve as entry points for people to send commands;
In the midstream, operational platforms such as DiDi, Cao Cao Mobility, and Ruqi Mobility are responsible for order distribution, capacity scheduling, and completing order management. According to statistics from the ride-hailing regulatory information interaction system of the Ministry of Transport, the national monthly ride-hailing order volume in 2026 has been maintained at around 1 billion orders, and various ride-hailing platforms have accumulated rich operational experience;
At the back end, Robotaxis and unmanned minibuses in various regions complete the physical transportation of passengers, delivering the mobility results.
In this segment, domestic manufacturers are also clearly fully prepared and have rushed to the edge of commercialization:
As of August this year, Apollo Go has landed in 28 cities globally, completing over 23 million orders in total, with the total autonomous driving mileage exceeding 350 million kilometers;
Pony.ai's seventh-generation Robotaxi has commenced fully driverless commercial operations in Guangzhou, Shenzhen, and Beijing, and the fleets in Guangzhou and Shenzhen have even achieved revenue balance per vehicle;
WeRide's L4 fleet has been deployed in over 40 cities across 13 countries globally, holding autonomous driving licenses in 8 countries. Its global fleet size exceeds 3,000 units, covering various mobility carriers such as Robotaxis and unmanned minibuses;
Segway-Ninebot is also continuously implementing solutions in the fields of low-speed unmanned vehicles and automatic shuttle equipment, accumulating extensive commercial operation experience in short-distance unmanned mobility in scenarios such as parks and commercial districts.
After the entire chain is connected, AI mobility is no longer just a concept in the laboratory, but has transformed into a complete, truly operating industrial system.
Transformation
The formation of the new industry chain closed loop will inevitably bring about multiple transformations in the business logic of the mobility industry.
The first change is the shift in traffic entry points.
In the mobile internet era, aggregation platforms like DiDi and Amap Ride were the core entry points for mobility demands; users had to open the corresponding APPs to hail a ride. In the future, when users can simply say "Hail a ride for me" to their smartphones, smartwatches, or AR glasses to achieve mobility, the large models behind AI assistants will become the primary receivers of mobility demands.
Naturally, what follows is the shift of traffic entry points from aggregation platform APPs to various AI agents.
The second change is the thorough reconstruction of the industry's core competitiveness.
In the past, the underlying logic of competition in the ride-hailing industry was about who could offer more subsidies to attract users, and who could recruit and manage more drivers to seize a larger market share. In the era of AI mobility, the competitive logic shifts from competing on price and drivers to competing on vehicles and technology.
The core of the competition will become which ride-hailing platform can better understand users' complex mobility demands, possess a larger scale of Robotaxi capacity, and have an intelligent driving system that can more stably handle complex urban road conditions. The one that achieves this will be more easily prioritized and recommended by terminal AI agents, securing more orders.
The third change is the reshuffling of the positions in the upstream and downstream of the industry chain.
System-level AI agents built into smartphones and smart terminals are growing into new traffic entry points, with the opportunity to replace traditional aggregation platforms, take over users' mobility demands, and even participate in formulating new service rules for the ride-hailing industry.
In fact, not only in the mobility field, AI agents carry the output of an entire life service ecosystem, with mobility being just one of the infrastructure modules.
In this new industrial structure, traditional ride-hailing platforms and Robotaxi operators have all become capacity suppliers behind AI agents, or rather, capacity supply modules.
The shift in industrial logic means that the boundaries between platforms and autonomous driving companies are blurring.
In the past, the two basically stayed in their own lanes. Ride-hailing platforms operated traditional ride-hailing orders and drivers, with almost no involvement in autonomous driving; whereas autonomous driving companies focused on manufacturing autonomous vehicles and developing autonomous driving algorithms. Their own APPs only operated their limited Robotaxis and did not involve large-scale vehicle scheduling and order management.
In the AI era, the order entry point is controlled by the AI agents of front-end hardware. Unless capacity-side enterprises also manufacture smartphones and smartwatches, it is difficult for them to intrude. However, ride-hailing platforms and Robotaxi operators on the capacity side can replicate each other.
From the perspective of industrial competition and cooperation, possessing only order scheduling capabilities without unmanned vehicle capacity will leave one subject to the suppliers of autonomous vehicles in the future; conversely, possessing only autonomous vehicle technology without a mature mobility scheduling platform and compliant operation system will make it difficult to independently undertake large-scale user mobility orders.
It is extremely difficult to simultaneously master demand scheduling capabilities and unmanned vehicle fulfillment capabilities, but doing so can determine an enterprise's long-term industrial discourse power.
Therefore, from a longer-term perspective, it is highly probable that ride-hailing platforms and Robotaxi operators will ultimately move towards integration.
Integration
Regardless of who acquires whom between the two, there will ultimately be a showdown. Such a long-term trend already has short-term signs—leading ride-hailing platforms have long entered the arena to layout Robotaxi businesses.
DiDi Autonomous Driving, in conjunction with GAC Aion, has created the new-generation Robotaxi model R2, completing mass production and delivery in January 2026. The first batch of vehicles has obtained road test licenses for intelligent connected vehicles in Guangzhou and commenced normalized passenger-carrying tests in demonstration areas in Beijing and Guangzhou.
Cao Cao Mobility is also accelerating the layout of its autonomous driving business. In February 2025, Cao Cao Mobility launched the Cao Cao Smart Mobility autonomous driving platform in Suzhou and Hangzhou; in December 2025, it released the Robotaxi 2.0 full-stack solution, while proposing the long-term strategic goal of "Ten Years, Hundred Cities, CNY 100 Billion", planning to continuously expand the scale of Robotaxi implementation.
Obviously, while maintaining massive ride-hailing scheduling networks, these ride-hailing platforms are independently developing autonomous driving systems and implementing dedicated Robotaxi models, attempting to connect both ends of order scheduling and unmanned vehicle fulfillment to complete an integrated layout.
On the other hand, Robotaxi manufacturers are also either building or partnering with mobility platforms to make up for shortcomings in scheduling and operations.
Apollo Go has long relied on its own APP for operations, with almost no integration into other external aggregation ride-hailing platforms domestically. Order scheduling, fulfillment, and customer service are all handled by Apollo Go's proprietary system.
The cooperation between Pony.ai and Ruqi Mobility is a typical example of upstream integration.
In March 2026, Pony.ai delivered the first batch of over a hundred seventh-generation GAC Aion T-Rex Robotaxis to Ruqi Mobility. The vehicles are equipped with Pony.ai's self-developed autonomous driving system and integrated into the Ruqi Mobility platform for operation. This cooperation model features a clear division of labor: Pony.ai focuses on the research, development, and iteration of "virtual driver" autonomous driving technology, while Ruqi Mobility is responsible for fleet assets, safety assurance, vehicle scheduling, and platform operations. Both parties jointly build a scaled Robotaxi fleet.
Before the arrival of the era of complete integration, the mixed operation model has become a consensus in the industry's transition period.
For instance, after WeRide launched its paid Robotaxi service, early orders all came from its proprietary App, and it subsequently integrated into third-party aggregation platforms.
In March 2026, WeRide Robotaxi was officially integrated into the ride-hailing entry of WeChat's "Me - Services - Mobility Services", opening up in the Guangzhou region. Passengers place orders on Tencent Mobility Services, select "Autonomous Driving - WeRide Mobility", and the orders will automatically flow to WeRide's backend scheduling system to complete vehicle dispatch, fulfillment, and settlement. Overseas, WeRide integrates into platforms like Uber, where orders are distributed by the platforms. This is also the most core source of orders for WeRide's overseas Robotaxis.
After cooperating with Pony.ai, Ruqi Mobility intelligently matches human-driven ride-hailing services or Robotaxis based on factors such as user location, destination, real-time vehicle location, and waiting time, achieving unified order dispatch and unified fulfillment;
Similarly, by the end of 2025, T3 Mobility had integrated over 300 Robotaxis, conducting L4-level autonomous driving road tests in Nanjing and Suzhou, realizing the mixed operation of human-driven and unmanned ride-hailing services.
In the mixed operation stage, ride-hailing platforms continuously refine scheduling algorithms, autonomous driving enterprises continuously accumulate urban road data, and hardware manufacturers continuously reduce the costs of sensors, chips, and complete vehicles.
The mobility business model in the AI era has already taken its initial shape. All players in the industry chain will eventually face an ultimate choice. But before that, all enterprises participating in this transformation of mobility forms will enjoy a relatively long period of growth.