LiDAR Emerges as Physical AI’s Biggest Winner

Post LinkedIn

🐯Read original on 虎嗅

#adas #3d-perception #embodied-ai #sensor-fusionlidar-for-physical-aihesai robosense lidar physical-ai

💡LiDAR is becoming the critical data and sensing layer for real-world AI, robotics, and autonomous driving.

⚡ 30-Second TL;DR

What Changed

Physical AI requires real-world deployment data, making 3D sensing and LiDAR increasingly important.

Why It Matters

For AI builders, better and cheaper LiDAR could accelerate embodied AI, autonomous vehicles, and industrial perception systems. The main competitive factors are shifting from raw hardware cost toward sensor quality, integrated compute, deployment scale, and access to real-world data.

What To Do Next

Prototype a ROS 2 3D-perception pipeline with LiDAR point clouds and benchmark detection latency, power use, and performance against camera-only input.

Who should care:Developers & AI Engineers

Key Points

•Physical AI requires real-world deployment data, making 3D sensing and LiDAR increasingly important.
•Hesai reportedly leads global ADAS LiDAR shipments with a 43% share and reached 55% of China’s market in March 2026.
•RoboSense shipped 71.9万 LiDAR units in the first half of 2026, up 170% year over year.
•The industry is moving toward sensor-compute integration using SPAD-SoC designs to reduce size and power consumption.
•Hesai’s Picasso chip targets 4,320 lines and 6D color perception, while RoboSense’s Falcon chip supports 2,160 native lines and up to 600 meters of detection.

🧠 Deep Insight

AI-generated analysis for this event.

🔑 Enhanced Key Takeaways

•The integration of LiDAR with Physical AI is driving a shift toward 'software-defined LiDAR,' where perception algorithms are increasingly offloaded to centralized domain controllers rather than processed solely on the sensor.
•Recent supply chain data indicates that Chinese LiDAR manufacturers are aggressively expanding into the European and North American markets to circumvent domestic price wars, despite facing potential geopolitical trade barriers.
•The industry is seeing a transition from 905nm to 1550nm laser wavelengths for long-range highway autonomy, as 1550nm systems offer superior eye safety and higher power density for long-range detection.
•Major automotive OEMs are increasingly demanding 'sensor fusion-ready' LiDAR units that provide raw point cloud data with synchronized timestamps to simplify the integration with camera-based vision systems.
•The rise of Physical AI has spurred a new market segment for 'solid-state' LiDAR specifically designed for humanoid robots, emphasizing low power consumption and high-frequency scanning over the extreme long-range requirements of automotive ADAS.

📊 Competitor Analysis▸ Show

Feature	Hesai (ATX Series)	RoboSense (M-Platform)	Luminar (Iris)	Innovusion (Falcon)
Architecture	SPAD-SoC (Picasso)	ASIC-based	Fiber Laser (1550nm)	1550nm Hybrid Solid-State
Max Range	~300m	~250m	~600m	~500m
Primary Market	ADAS/Robotaxi	ADAS/Robotics	Premium ADAS	Highway ADAS
Cost Strategy	High-volume/Margin	Cost-sensitive/Scale	Premium/Performance	Performance/Integration

🛠️ Technical Deep Dive

SPAD-SoC Integration: Modern LiDAR architectures are moving from discrete components to System-on-Chip (SoC) designs, integrating Single-Photon Avalanche Diode (SPAD) arrays directly with processing logic to reduce latency and physical footprint.
6D Perception: Hesai's Picasso architecture utilizes 6D perception, which adds velocity and reflectivity data to traditional 3D spatial coordinates, enabling better classification of dynamic objects in complex urban environments.
Native Line Resolution: The shift toward 2,000+ native lines allows for higher point density at distance, which is critical for identifying small road debris or lane markings at highway speeds.
Power Efficiency: New generation chips are targeting sub-10W power consumption, a critical threshold for integration into passive-cooled automotive housings and battery-constrained robotic platforms.

🔮 Future ImplicationsAI analysis grounded in cited sources

LiDAR unit costs will drop below $300 for mass-market ADAS by 2027.

The transition to highly integrated SPAD-SoC architectures significantly reduces bill-of-materials (BOM) costs and assembly complexity.

Physical AI will force a consolidation of the LiDAR market to three major global players.

The high R&D costs required for custom SoC development and the scale needed for automotive-grade manufacturing create a high barrier to entry that favors established leaders.