Autonomous driving firms are repositioning the capabilities accumulated from their vehicle fleets for new valuations in the physical world beyond automobiles.
Historically, Robotaxi company data primarily served vehicle R&D, with collection, annotation, simulation, and model training all ultimately aimed at one goal: enabling vehicles to operate safely on more roads. With the rise of embodied AI and physical intelligence, this data production capability is now engaging new clients, as robots, logistics equipment, and industrial agents require perception, decision-making, and action in real-world environments.
This shift is altering the competitive boundaries for autonomous driving companies. The competition is no longer just about whether vehicles can operate on roads, but also about whether the long-accumulated engineering systems can be transformed into infrastructure usable across more physical AI scenarios.
According to informed sources, WeRide Inc. has initiated the independent operation of its data business, which is being undertaken by its wholly-owned subsidiary, Jingshuo. One source indicated that Jingshuo's business focus is on embodied intelligence and data infrastructure software services, encompassing data generation, data collection, and proprietary data systems. Its current pool of real-world data is not solely derived from automotive scenarios.
In response to inquiries, WeRide Inc. stated on July 22nd that there was no information available for disclosure at this time.
This move by WeRide Inc. signifies that autonomous driving companies are cross-applying their capabilities, establishing the engineering systems formed from data collection, processing, simulation, evaluation, and model iteration.
For the broader Robotaxi industry, competition is transitioning from accumulating more road mileage to productizing the capabilities developed on the road into repeatably purchasable products. Only those who can generate sustained orders for this system beyond automobiles will have the opportunity to advance from competing on autonomous driving solutions to competing in physical AI infrastructure.
Operational Independence
The unit spun off by WeRide Inc. is a software service centered on data production and model iteration. This marks the first time WeRide Inc. has separated a business segment for independent operation.
Jingshuo was established in May 2024 as a wholly-owned subsidiary of WeRide Inc.. Leveraging experience from the autonomous driving field, it has been building an AI data closed-loop and solution system.
Jingshuo already possessed a foundation in data services and delivery, though it previously functioned more as a supporting operation within WeRide Inc.'s autonomous driving R&D framework.
Currently, Jingshuo's externally presented business has expanded further. Its AI platform covers management of raw data, training data, and annotated data, as well as model selection, training, evaluation, and deployment. Its embodied intelligence business encompasses data synthesis, teleoperation data collection, and data processing.
It is understood that Jingshuo is positioned as an embodied intelligence and data infrastructure software service business, covering data generation, data collection, and proprietary data systems. The real-world data it has accumulated is not limited to automotive sources.
The core product Jingshuo aims to sell is not merely the quantity of samples, but the capability to transform real-world data into training material. Furthermore, data collection is just one pathway; once real data enters the system, it can continue to generate simulated data.
For robotics companies, the value of this capability lies not in acquiring another batch of road videos, but in whether it can reduce data production costs and integrate data from diverse sources into a unified training and validation pipeline. What the robotics industry lacks is not samples from a specific scenario, but a system capable of consistently generating high-quality data.
The aforementioned source stated that the related business had undergone a considerable period of development, and the decision to begin independent operations and seek financing was made following a recent assessment.
In other words, Jingshuo is not starting from scratch to find a direction in embodied intelligence, but rather reorganizing existing capabilities into a business facing external clients.
Jingshuo's revenue target for 2026 is 300 million yuan. According to WeRide Inc.'s financial reports, its intelligent data service revenue for 2024 was approximately 55.8 million yuan, with a projected figure of 159.6 million yuan for 2025 based on report extrapolations.
It is understood that Jingshuo recently completed a Series A funding round. The internal assessment at WeRide Inc. is that there is significant demand for this business, and investors are relatively confident in its prospects. Previously, WeRide Inc. also facilitated an independent financing round for its Robovan unit, which achieved a valuation exceeding $400 million.
The revenue logic for Robovan and Jingshuo differs. Robovan focuses on vehicles, logistics operations, and technical services, where clients are concerned with vehicle delivery, operational efficiency, and per-unit cost reduction. Jingshuo, however, focuses on data generation, simulation, and model training services, where clients need to integrate with their own models and robot platforms.
Both, however, rely on the WeRide One foundational platform. WeRide One is WeRide Inc.'s universal autonomous driving technology platform, responsible for supporting the R&D, deployment, and operation across different vehicle types and scenarios.
This is precisely the foundation enabling Jingshuo's spin-off: WeRide Inc. already possesses an internal platform for cross-vehicle, cross-scenario technology reuse. Jingshuo's task is to further package the data production and engineering capabilities within this platform into services for external clients.
Valuing Capabilities
While the spin-off of the data business may spark new imagination in capital markets in the short term, it does not automatically translate into a realized new growth line.
For Robotaxi companies, the accumulated real-world road data, autonomous driving algorithms, and operational experience primarily serve their own vehicles and fleets. Spinning off the data business fundamentally changes how this capability is organized: it transitions from an internal R&D cost to a business that can be separately priced, financed, and accounted for.
The challenge facing Jingshuo is not whether it has data.
Road data cannot be directly converted into robot data. Autonomous driving and embodied intelligence involve different sensors, action spaces, and task objectives. Robots require extensive data on grasping, movement, manipulation, and human-robot interaction. What is potentially cross-applicable are the processes for data governance, scenario mining, simulation generation, model evaluation, and engineering delivery.
Therefore, the key for a Robotaxi company to enter the embodied intelligence data business lies not in simply transferring vehicle data to robots, but in reorganizing the engineering methodologies developed for autonomous driving into data products suitable for different robot platforms and tasks.
Similar moves are already emerging in this market. In June 2026, the data business unit under Ruqi Mobility released an embodied intelligence data platform, attempting to extend Robotaxi-related data capabilities to robotics scenarios.
The difference between companies does not lie in who first proposed physical AI, nor in whose promotional language aligns more closely with large language models. A more critical distinction is that some companies retain data, world models, and simulation capabilities internally to serve vehicle scaling and technological iteration, while others further separate these capabilities into independent businesses directly facing clients outside the automotive sector.
A March 2026 report on physical AI simulation and data platforms by Frost & Sullivan suggested the related market remains in an early growth stage, with platform value dependent on the technological maturity and commercialization progress of downstream applications. The report also noted that both intelligent vehicles and embodied intelligence are increasing demand for long-tail scenario reproduction, sensor simulation, and algorithm closed-loop optimization.
A technical representative from another embodied robotics company indicated that limited training data and fragmented simulation stacks are core issues in physical AI R&D. Although world models and simulation engines can compress training and evaluation cycles, they ultimately rely on deployment in real-world scenarios.
This places higher demands on data service providers like Jingshuo. Simply providing annotation and collection personnel can easily devolve into project-based human resource services. What truly has product attributes is the ability to combine real data, simulated data, scenario libraries, model evaluation, and delivery processes to reduce the time and cost for clients developing robots.
A Nomura report in July pointed out that Data-as-a-Service can be monetized quickly by the hour or project, but if suppliers lack model evaluation and application capabilities, they risk later integration by robotics enterprises themselves.
This assessment points to a reality of the data business: data collection is easily visible, but the capability to continuously improve model performance is harder to outsource.
This is also a new challenge facing the Robotaxi industry. Previously, the market primarily measured such companies by fleet size, operational mileage, orders, and per-vehicle economics. In the future, it will also assess whether companies can convert long-term R&D investments into orders beyond automobiles, and whether internal toolchains can become products that external clients are willing to purchase continuously.
The spin-off of Jingshuo by WeRide Inc. provides an observation case. It aims to validate not whether the concept of "embodied intelligence" can attract short-term attention, but whether the data and engineering systems accumulated by Robotaxi companies over years can transition from R&D investments serving the core business into capabilities sold across different scenarios.
If this calculation proves viable, competition among autonomous driving companies will extend from who has the larger fleet to who can sell the data and engineering processes accumulated within their fleets to more physical AI clients. Orders, repeat purchases, and gross margins would then truly grant them a new valuation.
Comments