Hesai Expands into Robotics Training Data with Kosmo Platform
TREE NEWS reports: On October 14, Hesai Technology unveiled its Kosmo spatial camera and cloud platform, a new product line designed to scan real-world environments and reconstruct them in 3D for robot simulation training. The company also signed strategic partnerships with Galaxy General and Yuanli Lingji. This marks a significant pivot for a firm primarily known for its lidar sensors, extending its business into data and cloud services for the burgeoning embodied AI industry.
The move addresses a critical bottleneck in robotics: the lack of diverse training data. While autonomous vehicles can rely on large fleets to accumulate real-world driving data, robots have yet to be widely deployed, creating a chicken-and-egg problem. Developers need vast amounts of environmental and operational experience to make robots capable of performing real-world tasks, but without deployment, data accumulation is limited.
Market Implications: A New Revenue Stream for Lidar and Robotics
Hesai’s entry into training environment construction could reshape the economics of the robotics supply chain. By leveraging its expertise in 3D perception, chip design, and precision manufacturing from the automotive sector, Hesai aims to reduce the cost of building simulation environments. The company envisions two revenue models: selling integrated hardware-software products to industry clients, and providing spatial data production services to world model and physical AI companies.
This shift from a pure component supplier to a service provider could smooth revenue cyclicality. Component orders are tied to robot production volumes, while data services are linked to ongoing development and operational activities. As new stores open or layouts change, clients will need to re-scan and update environments, generating recurring service demand. Hesai’s Q2 2026 results showed total deliveries up 78.4% year-over-year, with revenue up 21.9%, though average selling prices declined. The Kosmo platform could help offset pricing pressure in the core lidar business.
Competitive Landscape and Broader Trends
Hesai is not alone in recognizing the data bottleneck. In August, Lightwheel partnered with Hugging Face to release a first-person human activity dataset, aiming to open 100,000 hours of data. Nvidia introduced a physical AI data factory blueprint to streamline data generation, augmentation, and evaluation. These initiatives highlight a growing ecosystem focused on improving the efficiency of robot learning.
Hesai’s Kosmo differentiates by focusing on the training environment itself. The system scans a 100-square-meter indoor space in about 15 minutes, then reconstructs it into a 3D model where objects can be manipulated and layouts altered. This allows developers to practice tasks under varied conditions without repeatedly setting up physical spaces. The platform integrates with Isaac Sim, Isaac Lab, and film production tools, aiming to fit into existing workflows.
Key Takeaways for Investors
- Diversification beyond lidar: Hesai is moving up the value chain, potentially creating recurring revenue streams that are less dependent on robot unit sales.
- Addressing a critical pain point: The robotics industry’s data scarcity is a major hurdle. Solutions that reduce training costs and accelerate deployment could see strong demand.
- Ecosystem play: Partnerships with robot companies like Galaxy General and Yuanli Lingji provide real-world feedback and validation, strengthening Hesai’s product-market fit.
- Risk factors: Execution risk in a new business model, competition from Nvidia and others, and the need to build a broader user base beyond enterprise R&D teams.
As robots move from prototypes to production, the tools and services that standardize training environments will become increasingly valuable. Hesai’s bet on Kosmo positions it at the intersection of lidar, AI, and cloud services, a space that could define the next phase of embodied intelligence.
