MotorClaw.news
Subscribe
Search releases, companies, topics...
Updated

Hesai releases Spatial Intelligence white paper, detailing Kosmo 3D world data engine for Physical AI

Hesai has published a global white paper arguing that the supply of high-quality 3D real-world data, not AI model capability, is the bottleneck holding back Physical AI.

Kosmo One imaging resolution
8K
Dimensions of 3D data requirements
2 (scale and quality)
Kosmo application categories
2 (sim-ready and production-ready)

What Happened

Hesai Group, which describes itself as a Physical AI infrastructure company, released its Spatial Intelligence White Paper globally on November 10, 2026. The paper examines the technological evolution, industry trends and application opportunities of spatial intelligence and highlights key developments in Hesai's 3D world data engine, Kosmo. It identifies a critical bottleneck facing the Physical AI industry: while AI models are advancing rapidly, the supply of real-world data is struggling to keep pace. The ability to generate and deliver high-quality, multimodal 3D spatial data at scale is becoming an essential component of the infrastructure needed to advance Physical AI.

The white paper traces a decade of AI evolution, from reading and generating text to an initial ability to understand and reason about the 3D world, moving beyond digital environments and into the physical world. To operate effectively in real-world environments, AI must go beyond identifying objects: it needs to understand where objects are, how they relate to one another, how the surrounding environments are structured and how they may change over time. The paper defines this capability as spatial intelligence, the ability of an intelligent system to establish a complete loop of perception, understanding, reasoning and interaction based on real-world spatial information, so that AI can not only understand the world but also interact with it.

What different industries need from 3D spatial data
  • World models and generative AI require extensive and diverse real-world scenarios, placing a premium on data scale.
  • Embodied AI requires a balance of data scale and rich geometric, semantic and physical attributes.
  • Film and entertainment prioritize visual realism, geometric consistency and multi-view stability.
  • Cultural tourism and heritage place greater emphasis on geometric accuracy, textural detail and authenticity.

Kosmo combines Kosmo One, which Hesai calls the world's first 8K ultra HD AI spatial camera, with the Kosmo World cloud platform. Kosmo One captures real-world environments as detailed, reusable 3D digital assets, while Kosmo World, a one-stop platform for 3D content creation and management, transforms the captured data into high-quality 3D models and simulation-ready environments. Together they provide the real-world data needed for embodied AI training and world model development, while also enabling creators to capture real-world spaces and develop immersive content.

Kosmo's two application categories
Sim-ready
Supports world models, generative AI and embodied AI by providing real-world 3D data for model training, robot simulation and validation
Production-ready
Serves film and entertainment and cultural tourism and heritage by streamlining 3D content creation and enabling reuse of digital assets
Kosmo One imaging resolution

8Kultra HD

Hesai describes Kosmo One as the world's first AI spatial camera with an integrated 8K imaging system, based on publicly available product information as of October 2026.

Hesai positions Kosmo in the upper-right quadrant of the spatial data demand matrix, where requirements for both data scale and quality are high, and says it is initially addressing one of the industry's most challenging data supply problems, targeting high-value applications with strong demand for 3D data and high barriers to entry. As spatial intelligence moves toward large-scale deployment, the white paper argues the industry needs more than standalone tools: it requires full-stack platforms that can produce high-quality data at scale in a standardized manner and connect different parts of the value chain. Looking ahead, Kosmo will continue to leverage integrated hardware and software capabilities to scale the supply of high-quality real-world 3D data and the applications that use it, working with ecosystem partners to accelerate adoption and provide a foundation for Physical AI.

Previously from Hesai Technology

Motorclaw's earlier coverage of Hesai focused on its lidar business in robotics and autonomous driving. Yole Group ranked Hesai the world's top robotaxi lidar supplier with a 62% share of the global market in 2025, more than all other suppliers combined. Hesai has also explained how lidar helps warehouse robots, farm machines, drones and humanoids read physical space, and has pitched its 6D full-color lidar, built on the Picasso ASIC platform unveiled in April, as one sensor delivering chip-aligned color and depth for autonomous driving and Physical AI.

Background drawn from MotorClaw's earlier coverage of Hesai Technology's official releases.

Why this matters

The white paper frames large-scale, high-quality 3D spatial data as essential infrastructure for Physical AI, naming world models and generative AI, embodied AI, film and entertainment, and cultural tourism and heritage as the industries affected. For Hesai, it is a pitch that its Kosmo hardware-and-cloud platform can supply that data at scale.

Terms in This Story

Physical AI
AI systems that operate in and act on the physical world rather than only in digital environments such as text or images.
Spatial intelligence
The ability of an intelligent system to perceive, understand, reason about and interact with real-world spatial information.
World model
An AI system that builds an internal representation of an environment so it can reason about how that environment is structured and how it may change.
Embodied AI
AI that is built into a physical machine, such as a robot, and must interpret and act within real surroundings.
Read Original: Hesai Technology

Summarised from the linked release; details can be imperfect — always verify against the original source.

How useful was this article?

Feedback goes to the MotorClaw Desk and does not subscribe you to email. Privacy.