Hesai Group launched Kosmo, a 3D world data engine built to solve one of the most critical bottlenecks in Physical AI: access to high-quality real-world data at scale. Kosmo combines Kosmo One, the world’s first AI spatial camera(1), with the Kosmo World cloud platform to transform physical environments into high-fidelity 3D digital spaces for robot training, world model development and immersive content creation. At the launch event, Hesai also announced strategic partnerships with embodied AI companies Galbot and Dexmal and showcased applications developed with Pixar’s RenderMan and Amap, demonstrating Kosmo’s potential across robotics, filmmaking and location services.
Digitizing the real world for Physical AI
Kosmo One, the world’s first 8K ultra-high-definition AI spatial camera(2), captures real-world environments as detailed, reusable 3D digital assets. It integrates high-resolution lidar, a one-inch imaging sensor and a high-precision inertial measurement unit (IMU) to capture geometry, color, texture and camera position with microsecond-level synchronization. Advanced 3D Gaussian Splatting (3DGS) reconstruction delivers millimeter-level geometric accuracy and lifelike detail, faithfully reproducing materials, lighting and shadows.
Designed for ease of use, Kosmo One offers instant scanning on power-up, intelligent guidance and one-click scan resumption. A single operator can scan a typical 100-square-meter indoor space in approximately 15 minutes(3). Kosmo World then processes the captured data into high-quality 3D models at speeds of up to approximately five times those of comparable products(4).
Turning real-world spaces into training grounds for Embodied AI
Kosmo World, a one-stop platform for 3D content creation and management, converts captured data into simulation-ready environments for embodied AI training. Powered by Hesai’s Kosmo Spatial Core (KSC), the platform generates 3DGS, mesh and depth data in a single reconstruction, combining visual realism, interactive objects and precise spatial information within one consistent scene. Generative AI fills unscanned areas and reconstructs materials and textures, while AI-powered semantic segmentation identifies individual objects. Views can also be adapted to a robot’s camera characteristics, helping narrow the gap between simulation and real-world operation.
A deployment at a Dairy Queen (DQ) ice cream store demonstrates this in practice. Creating a 3D training environment previously required approximately 300 hours of manual modeling by a professional team. With Kosmo, scanning and reconstruction took just 15 minutes. In this deployment, the Sharpa robot’s task success rate reached 99%, while the failure rate fell to one-tenth of their previous level, enabling the robot to begin operating in an actual store.
Grounding world models in reality
Kosmo also provides a real-world foundation for world model development. Its geometrically accurate 3D scenes bring together precise camera poses, depth information and consistent multi-angle views, giving developers a real-world reference for creating realistic virtual environments. From a single captured space, Kosmo can automatically generate more than 10,000 training trajectories, so developers can obtain more than 100 times as much high-quality training data for the same cost.
Beyond embodied AI and world models, Kosmo is bringing value to a growing range of fields, including film and media, cultural tourism and heritage, and commercial use of spatial data.
Film and media: faster production, greater creative freedom
Kosmo lets filmmakers capture a location once and reuse it at every stage, from previsualization and virtual production to postproduction. In established tools such as Unreal Engine, artists can explore new camera angles, adjust lighting and reuse individual scene elements. A 10,000-square-meter European streetscape that would take two months to build as a physical set can be scanned, reconstructed and edited in just one week, with potential cost savings of up to 80%.
Leading film and television companies worldwide are already using Kosmo, including in workflows built around Pixar’s production-grade renderer RenderMan, which gives artists precise control over the appearance and stylization of captured environments. To help creators adopt the workflow, Pixar’s RenderMan has also released an official tutorial along with downloadable resources.
Hesai is also partnering with Yuemo Studio on AI-assisted filmmaking, pairing captured real-world environments with AI-generated content to reduce token consumption while maintaining visual realism.
Cultural tourism and heritage: preserving history in 3D
Kosmo can preserve cultural heritage permanently in digital form, allowing people to experience it from anywhere in the world. Hesai is working with TRIC, a United Nations partner, to create an immersive digital artwork of Du Fu Thatched Cottage in Chengdu, the former home of Du Fu, the Tang Dynasty poet revered in China as the “Poet-Sage.”
Spatial commerce: navigating to the product, not just the store
Through its strategic partnership with Amap, a leading Chinese mapping and local discovery platform, Hesai is bringing immersive 3D environments into everyday location services. Reconstructions of the Gravity Art Museum and PIFO Gallery are already live in Amap’s “Flying Street View” feature. Consumers can preview destinations before visiting, while merchants gain digital spaces they can update and analyze to inform store layouts and product placement. Amap reaches nearly one billion monthly active users, so the partnership offers a broad platform for expanding 3D experiences. Ultimately, the aim is to extend navigation from finding a store to locating a product inside it.
An open ecosystem with two business models
Hesai is building Kosmo into an open spatial intelligence ecosystem, with open access to raw capture data and 3D assets, and integration with leading platforms including NVIDIA Isaac Sim, Isaac Lab, Unreal Engine, RenderMan and Houdini. The ecosystem supports two business models: integrated hardware and software solutions for customers across robotics, media and other industries; and services that produce high-quality real-world spatial data at scale for world model and Physical AI developers.
At the event, Hesai also announced new partnerships with Galbot and Dexmal to advance embodied AI development using Kosmo.
Hesai’s physical AI roadmap
Hesai is extending more than a decade of expertise in lidar, ASIC development and precision manufacturing into the broader Physical AI market. Its lidar deliveries have doubled for nine consecutive years, and it holds a leading global market position in passenger-vehicle ADAS and robotaxis. On this foundation, the Company is building its next phase of growth around 3D world data and robotic motion.
Alongside Kosmo, Hesai is expanding its robotic actuation business, designing hardware around the demands of real-world AI applications. Its actuation modules power Sharpa’s Wave dexterous hand and deliver three times the power density of comparable products in a package 37% smaller. The hand has been incorporated into NVIDIA’s Isaac GR00T humanoid robot reference design and has been adopted by leading research institutions including UC Berkeley and Stanford. Cumulative deliveries of Hesai’s actuation modules exceeded 10,000 units by the end of the second quarter of 2026. The portfolio is expanding from hand modules to joints throughout the robot’s body. Shipments are expected to reach six figures in 2027.
For more information, visit hesaitech.com.
- Based on publicly available product information as of October 2026, Kosmo One is the world’s first AI spatial camera with an integrated 8K imaging system.
- 8K refers to the maximum capture resolution supported by the camera. The maximum resolution of files exported by users is 6K.
- For a typical indoor room, scanning a 100-square-meter space takes approximately 15 minutes. Actual capture time varies with the layout and size of the space and the capture requirements.
- Compared with similar 3DGS reconstruction products in the industry, model generation can be up to approximately five times as fast under similar hardware, dataset, and quality requirements. Actual results vary with the scene, data volume, and parameter settings.
The post Hesai launches 3D data engine for physical AI appeared first on Engineering.com.