This article is reprinted from Tencent News · Smart Driving Frontier, all rights reserved. Click to view the original link.

---

Has the LiDAR Battle Entered Its Second Half Centered on Chips?

Published in Jiangsu on 2026-06-24 08:40

April 2026 was incredibly busy for the LiDAR industry. Within just a few days, a string of new technologies were released: RoboSense unveiled its Genesis digital architecture along with two SPAD-SoC chips, Phoenix and Peacock; Hesai Technology released the world's first 6D full-color LiDAR super-sensing chip, Picasso SPAD-SoC; and Huawei introduced its 896-line dual-optical-path image-grade LiDAR in March. From these dense launch events, a trend in the LiDAR industry is visible: competitors are no longer comparing scanning structures or stacking line counts; the focus of competition has clearly shifted to chips.

From Analog Architecture to Digital Architecture

Over the past decade, the core debate in the LiDAR industry revolved around scanning structures—whether to use MEMS mirrors or rotating mirror scanning—which was once the primary yardstick for dividing technical camps. But since 2024, this dividing line has been rapidly disappearing, replaced by the distinction between analog and digital architectures.

Traditional analog LiDAR relies on seven discrete components to build the transmission, reception, and processing systems. Performance improvements depend on stacking components; the more lines, the more components, binding cost and performance linearly. The 128-line analog architecture has long been considered the performance endpoint by the industry, because the more components stacked, the higher the volume and power consumption. Digital architecture centers on SPAD-SoC (Single Photon Avalanche Diode System-on-Chip), integrating photon reception, signal processing, and large-scale digital computing units into a single chip. This achieves a fully digital signal processing chain where photons entering immediately become digital pulses, requiring only two core chips to achieve equivalent functionality for the entire system.

The difference between analog and digital architectures is structural. Analog follows a "stack hardware" logic: every step up in resolution requires corresponding increases in hardware cost. Digital stands within the standard semiconductor coordinate system: performance leaps can be achieved through chip process node improvements without proportional cost increases. RoboSense CEO Qiu Chunchao made a judgment: 128 lines is the endpoint of stacking materials for analog architecture, while 192 lines is merely the starting point of performance for digital architecture.

RoboSense compares this transition to the imaging industry's underlying migration from CCD to CMOS. Qiu Chunchao believes the choices facing the LiDAR industry today are surprisingly similar to those back then. Current mainstream analog solutions rely on piling up performance with discrete components like APDs and SiPMs, increasing hardware costs with every step up in resolution. Digitization gives LiDAR evolutionary capabilities akin to Moore's Law for the first time.

What Changes Do Chips Bring to LiDAR?

The first change SPAD-SoC brings to LiDAR is that the sensing link is digital from the start. SPADs can directly convert return signals from individual photons into digital signals, which are then processed by the on-chip SoC. Once in the digital domain, LiDAR can evolve along the semiconductor path: higher integration, higher resolution, lower cost.

Take the Phoenix chip released by RoboSense in April 2026 as an example. It is the world's first monolithically integrated native 2160-line automotive-grade SPAD-SoC, achieving over 4 million pixel resolution and ultra-long-range detection of 600 meters. The Phoenix chip uses a single-chip, single-optical-path design, with a point cloud resolution of 2160×1900, capable of identifying objects as small as 13×17 cm from 150 meters away. The Phoenix series offers five models, supporting LiDAR product designs ranging from 2160 lines down to 240 lines.

The Genesis digital architecture itself comprises a four-layer system. The basic process layer adopts a 28nm automotive-grade process, reducing core area by 40% and power consumption by 30%; the third-generation hypersensitive SPAD sensing layer achieves 45% photon detection efficiency; the second-generation 3D stacking process reduces noise; and the core computing layer features a 4320-Core heterogeneous computing array, supporting 495 billion point cloud sampling processes per second. The algorithm acceleration layer integrates an anti-interference engine, with sunlight noise resistance and crosstalk resistance reaching 99.7%; the safety and reliability layer embeds an ASIL-B functional safety architecture, ensuring stable operation in environments from -40°C to 125°C.

Another chip, Peacock, takes the large-area array route, integrating a 640×480 high-density SPAD array to achieve VGA-level resolution and approximately 300,000 pixels. Its field of view reaches up to 180°×135°, with a minimum detection distance of less than 5 cm and a frame rate of 10-30Hz, aligning with camera frame rates for the first time. Millimeter-level detection precision has improved sixfold compared to previous generations. The Peacock chip can be used in automotive solid-state blind-spot coverage, standardized visual modules for robots, and fused sensor scenarios. Both chips are scheduled for mass production landing within 2026.

Hesai Technology's path differs slightly but points equally toward chipification. In April 2026, Hesai released its fifth-generation self-developed chip platform, the Picasso SPAD-SoC, the world's first 6D full-color LiDAR super-sensing chip. Traditional LiDAR can only sense 3D spatial coordinates (XYZ), whereas Picasso synchronously acquires spatial coordinates and color information (RGB) on the same chip, natively outputting pixel-level spatiotemporally aligned colored point clouds. This is not simply stitching camera images with LiDAR point cloud data, but achieving native fusion of sensing information at the chip level. Every point output by the 6D full-color LiDAR natively carries color information, aligned at the pixel level, requiring no stitching or mental interpolation.

Picasso's Photon Detection Efficiency (PDE) can break through 40%, reaching international top-tier levels. PDE is the core metric measuring a SPAD chip's sensing capability, directly determining how far and how clearly a LiDAR can see. The ETX series LiDAR equipped with Picasso supports up to 4320-line full-color 4K ultra-high-definition sensing, with a maximum range of 600 meters, reaching 400 meters at 10% reflectivity. It can clearly identify tiny targets such as water-filled barriers (120×60cm) within 300 meters, small animals (60×40cm) within 280 meters, and wooden blocks (15×25cm) within 150 meters. The ETX is planned for mass production delivery in the second half of 2026.

Comparing RoboSense and Hesai, Hesai has a longer accumulation period in chipification. To date, Hesai is the only LiDAR company globally to achieve full-stack self-development of seven core components: lasers, detectors, laser drivers, TIA chips, ADC chips, digital signal processors, and controllers. Twenty-one self-developed chips have obtained AEC-Q automotive certification, with cumulative deliveries exceeding 230 million units. Cumulative shipments are expected to break 300 million units by the end of 2026.

RoboSense, meanwhile, has explored another dimension in chip integration. RoboSense has achieved full-stack self-development and mass production of four core chips: scanning, transmission, reception, and processing. Its SPAD-SoC and 2D VCSEL chips passed AEC-Q102 automotive certification in 2025. In October 2025, RoboSense became the first globally to integrate SPAD and SoC into a single chip via 3D stacking, using copper-copper bonding technology to give each SPAD independent processing circuits, achieving lossless and efficient signal transmission.

There is another strong competitor in the LiDAR field: Huawei. Huawei released its 896-line dual-optical-path image-grade LiDAR in March 2026, adopting industry-first dual-optical-path technology with two laser receiving units of different focal lengths: telephoto and wide-angle. The wide-angle provides a large field of view overlooking the front, while the telephoto focuses on long distances to discern details. The two units can be controlled independently, with algorithms automatically adapting to different scenarios. The vertical resolution of this radar is four times higher than 192-line radars, capable of stably detecting obstacles the size of a soda can from 120 meters away. Huawei defines this as a generational leap in automotive sensing capability from point-cloud level to image level.

Cost Changes Brought by Chipification

The most direct market impact of chipification on LiDAR is reflected in cost. Early LiDARs used generic chips with low functional utilization, causing cost waste. Self-developed chips, however, can eliminate unnecessary functions, significantly reducing costs. Integrating hundreds of discrete components into a few dedicated chips greatly reduces component count and simplifies structure, serving as a key path for cost reduction, miniaturization, and high reliability.

Early 64-line LiDARs sold for as high as $80,000. By 2025, the price of primary LiDARs for L2 assisted driving had dropped to about $200, down from $500 two years prior. Hesai integrated seven core components like laser emission and reception onto chips, reducing costs by 60%; RoboSense's ADAS LiDAR unit price dropped by approximately 90% from 2020 to 2025.

For automakers, falling LiDAR costs allow mid-to-low-end models to implement more combined driving assistance. In 2025, the price of models equipped with LiDAR dropped to the 150,000 yuan level. By 2026, with the launch of models like Leapmotor A10 and Changan Qiyuan Q05 Laser Ultimate Smart Edition, LiDAR has penetrated down to the 80,000 yuan market.

Cost reductions have also driven the penetration rate of new energy vehicles. In 2025, the LiDAR penetration rate for NEV passenger cars reached 21%, peaking at 28% in a single month, crossing the 16% "chasm critical point." Zhongshang Industry Research Institute predicts the LiDAR passenger car penetration rate may reach 27% in 2026. In Q1 2026, total LiDAR installations in the domestic passenger car market exceeded 985,000 units.

Market Landscape Being Redrawn by Chip Capabilities

In Q1 2026, total LiDAR installations in the domestic passenger car market exceeded 985,000 units. Looking at suppliers, Hesai Technology ranked first with a 34.9% market share, followed closely by Huawei at 32.3%, together occupying 67.2% of the market. Seyond and RoboSense constitute the second tier with shares of 16.8% and 11.0% respectively. The four companies combined occupy 92% of the market share, indicating a relatively high level of industry concentration.

Further segmented data shows that in March 2026, Hesai's installation share in the domestic passenger vehicle primary LiDAR market reached 55%, higher than the combined share of all other suppliers, maintaining the No. 1 position for domestic installations for 14 consecutive months. Yole Group's 2026 Global Automotive Market Report data indicates that Hesai ranked first in global ADAS primary LiDAR shipments in 2025 with a 43% global market share.

If looking only at the forward primary LiDAR market, 2025 data shows Hesai first (41%), Huawei second (28.3%), and RoboSense third (23.5%), with the three companies combining for a 92.8% market share. This is directly related to chip capabilities. RoboSense developed its SPAD-SoC chip in 2022, integrating reception and signal processing functions into a single chip. With mass production of self-developed chips, product costs dropped significantly. Huawei, meanwhile, connected the VCSEL emission, LDD driver, and SPAD reception links at the chip level.

In terms of market size, reports from Zhongshang Industry Research Institute show China's LiDAR market size was approximately 13.96 billion yuan in 2024, about 24.07 billion yuan in 2025, and is predicted to reach 43.18 billion yuan in 2026. Expanding the view to the global market, CIC Consulting predicts the global LiDAR overall solution market size will climb from $3.5 billion in 2025 to $41.3 billion in 2030.

It is worth noting that competition in LiDAR chip capabilities has transcended the level of existence versus absence, entering a battle of generations. RoboSense achieved large-scale mass production of its self-developed SPAD-SoC in 2024, and based on SPAD-SoC, realized mass production of 520-line LiDAR in 2025. 2