Who Owns the Scene Owns AI: A Data Player Emerges in the Mobility Track

The AI industry is undergoing a fundamental shift. After large language models, world models and embodied intelligence are attracting capital and industrial attention. However, training these models requires vast amounts of interactive data from the real physical world—a resource that is extremely scarce. Industry insiders estimate a gap of nearly 100,000 times between demand and supply.

Traditional data sources such as static internet images, lab simulations, or crowdsourced collections cannot meet the need for complete cause-and-effect chains. This is where mobility platforms come in. Companies like Ruqi Mobility, a ride-hailing service under GAC Group, have turned their daily operations into a data goldmine. Their fleet of over 300 vehicles equipped with LiDAR, HD cameras, and other sensors continuously record the full interaction loop: driver decision, vehicle response, and environmental feedback.

Ruqi's data business, launched in 2023, has grown rapidly. In 2025, its technology services segment—primarily AI data services—reported revenue of 160 million RMB, a 487.4% increase year over year. The company now serves leading tech firms including Tencent, Pony.ai, Li Auto, Volcano Engine, Baidu AI Cloud, and GAC Group.

What sets Ruqi apart is its end-to-end capability: from compliant data collection and automated annotation (using proprietary OCC algorithm reducing manual work by 90%) to synthetic data generation for rare scenarios. Their datasets cover annotated data, behavior data, synthetic data, and multimodal training sets, all designed to help AI models not just recognize objects but understand causality and decision-making.

The article argues that in the age of spatial intelligence, owning a real-world scene is more critical than owning algorithms. Mobility platforms are perfect “meta-scenes” because they cover vast public road networks, involve constant human-machine co-driving, and generate billions of interactive events daily. This data production model—where data collection is a byproduct of normal operations—has inherently low marginal costs and high scalability.

Looking ahead, Ruqi is extending its data capabilities to other physical world applications such as robotaxi, embodied AI in car services (washing, charging, maintenance), and more. The belief is that companies with deep operational experience in physical scenes will see their data assets become foundational infrastructure for the next generation of AI.

In essence, the mobility platform is evolving from a people-moving service to a vital physical data gateway for AI to understand and enter the real world.