RoboSense Unveils the World’s First 2,160-Channel Automotive-Grade SPAD Chip, as LiDAR Enters the Image-Grade Perception Era
On April 21, 2026, RoboSense held a new product launch in Shenzhen, grandly releasing the new Genesis digital architecture and, based on it, two flagship SPAD-SoC chips: Phoenix and Peacock. Among them, the Phoenix chip, as the world’s first native monolithically integrated 2,160-channel automotive-grade SPAD-SoC, marks that automotive LiDAR has formally moved from point-cloud-grade perception into the new era of image-grade perception.
## Technical Breakthrough: From USD 90,000 Flowerpots to Chip-Level Integration
For a long time, the LiDAR industry has faced an unsolvable equation — products must keep chasing channel counts, but channel count scales linearly with cost. Analog-architecture solutions represented by SiPM, besides the core sensing devices, also require external analog chips for signal readout, amplification and conversion; the system is redundantly complex and costs stay stubbornly high.
RoboSense completely cracked this dilemma through its self-developed chip technology. As early as when the industry generally relied on purchased chips, RoboSense had already formed a self-developed chip team. In 2022, RoboSense developed its first-generation SPAD-SoC chip, integrating reception and signal processing into a single chip. Qiu Chunchao put it vividly: much of the anxiety came from a USD 90,000 flowerpot back then, whereas now a small chip packs several flowerpots.
The digital architecture stands in the standard semiconductor coordinate system and highly follows Moore’s Law. Under constraints of similar size and similar cost, performance keeps rising as long as chip processes keep advancing. Taking the EM4 platform as an example, it can reach 2,000 channels while keeping the same dimensions; converted by channel count, the cost drops by another order of magnitude.
## Phoenix Chip: Native 2,160 Channels, Detecting Small Targets at 600 m
As RoboSense’s flagship product, the Phoenix chip achieves comprehensive performance breakthroughs:
– Native 2,160 channels: point cloud fineness already surpasses a 4-megapixel camera, the highest channel-count spec in the industry today
– 600 m maximum detection range: capable of detecting distant targets, meeting high-speed autonomous driving scenario needs
– Small-target recognition: can clearly see a 13×17 cm paper box at 150 m, with small-target detection capability far above the industry mainstream
– Five model configurations: the Phoenix series offers five models, supporting LiDAR product designs from 2,160 channels down to 240 channels
The Phoenix chip has passed AEC-Q100 automotive qualification, and the 4-megapixel LiDAR solution based on it has already won a design win from a leading automaker, with volume production and vehicle installation within 2026.
## Peacock Chip: 180° Ultra-Wide View, Zero Blind Spots Up Close
The Peacock chip is the industry’s highest-spec mass-producible fully solid-state array SPAD-SoC:
– 640×480 high-density SPAD array: achieving VGA-grade resolution
– 180°×135° maximum field of view: an ultra-wide perspective
– Minimum detection distance under 5 cm: achieving zero blind spots at close range
The Peacock chip can output dense and fine 3D depth images, suitable for automotive solid-state blind-spot filling, standardized vision modules for robots, and brand-new fused sensor forms. The chip will ship at scale in Q3 2026, and products based on it have already been delivered to customers in small batches.
## Industry Impact: The Thousand-Channel Era Formally Arrives, Penetration Accelerates
Data from the Gaogong Intelligent Vehicle Research Institute (高工智能汽车研究院) show that in 2026 the penetration rate of LiDAR as standard front-mounted equipment on passenger vehicles in the Chinese market will break the 20% mark, and new vehicles equipped with LiDAR are expected to exceed 4.4 million units for the full year. LiDAR is rapidly sinking from a luxury item on high-end cars to a safety necessity on national models in the CNY 150,000 class.
L4 autonomous driving companies such as WeRide and DiDi Autonomous Driving have all adopted thousand-channel-class LiDAR. WeRide CEO Han Xu said the latest Robotaxi GXR will become one of the first driverless operation autonomous vehicles equipped with thousand-channel-class LiDAR, with 2,000 units expected to be delivered in 2026.
RoboSense CTO Liu Letian revealed that the first generation of mass-produced SPAD validation chips has already seen deliveries approaching 300,000-500,000 units, and close to 3 million chips are expected to be delivered in 2026. As chip costs keep falling and capacity scale keeps expanding, the adoption of LiDAR will accelerate further.
## RGBD Sensor Debuts in 2027 as SPAD Replicates CMOS Glory
RoboSense CEO Qiu Chunchao revealed that the true RGBD sensor will be officially released by the end of 2027. By then, SPAD is expected to replicate the glory of CMOS, giving birth to brand-new super-sensor forms and market applications.
An RGBD sensor can synchronously perceive 3D space (XYZ) and object color (RGB) on the same chip, directly generating color point clouds. This means LiDAR can see not only the position and shape of objects but also their colors, achieving pixel-level spatiotemporal alignment with no stitching and no fusing.
From USD 90,000 flowerpots to chip-level integration, from dozens of channels to 2,160 channels, from black-and-white point clouds to color depth images — Chinese LiDAR manufacturers are using self-innovated chip technology to redefine industry standards and lead the development direction of global intelligent driving perception technology.
—
Industry Watch: The Phoenix and Peacock dual chips released by RoboSense this time not only establish its global leadership in the SPAD chip field, but also open a new era of LiDAR imaging. As in-house chip capability becomes the core of competition, whoever occupies the high ground in chip technology will define performance, set the rules, and hold the core profits of the industry chain in the coming competition.



Leave a Reply