Author | Wang Xiaojuan
Editor | Zhou Zhiyu
On March 24, Hesai released its unaudited financial results for the fourth quarter and full year of 2025.
Amid years of skepticism regarding &#34;burning cash for scale,&#34; Hesai boosted confidence in the LiDAR industry with record annual revenue of RMB 3.03 billion (up 45.8% year-on-year) and its first-ever full year of GAAP profitability.
In 2025, a year where the debate between pure vision and LiDAR technical routes in the intelligent driving industry has intensified, the release of these results is particularly intriguing.
On one hand, LiDAR is accelerating its adoption in vehicles alongside the increasing penetration of intelligent driving; on the other hand, in an era where automakers are aggressively cutting costs, LiDAR remains the expensive hardware most likely to be &#34;optimized out.&#34;
Is LiDAR a necessity for the era of intelligent driving or a luxury for a technological transition period? As the penetration rate of intelligent driving rises, the market is providing an answer; meanwhile, LiDAR hardware seems to have found more intelligent agents beyond automobiles, opening up new markets for itself.
<h2>01 The Profitability Singularity Under Scale Effects</h2>
The most noteworthy aspect of Hesai's financial report is not just the revenue growth, but the shift in its profits.
For hardware manufacturing, &#34;diseconomies of scale&#34; was once a problem that plagued all LiDAR manufacturers. In past years, high R&amp;D expenses, BOM costs, and the pressure of annual price reductions from automakers trapped many firms in a quagmire of &#34;selling more, losing more.&#34;
In 2023, Hesai lost 840 million yuan, and RoboSense lost 760 million yuan, as the industry fell into the trap of &#34;losing money just to gain market attention.&#34; However, Hesai's 2025 financial report proves that this business can be profitable.
The core logic behind Hesai becoming the world's first LiDAR company to achieve full-year GAAP profitability lies in the reconstruction of its underlying technical architecture. Traditional LiDAR relies on a large number of discrete components, making assembly complex and costs high.
Through proprietary chip-based (ASIC) development for both transmitting/receiving and processing ends, Hesai has not only significantly reduced physical size but also brought core component costs onto the trajectory of Moore's Law.
According to company disclosures, through continuous chip-based technological innovation, Hesai has driven down LiDAR costs by 99.5% in eight years. As of November 2025, the cumulative delivery of its self-developed chips reached 185 million units, ranking first in the world. In 2025, Hesai launched its proprietary main control chip, &#34;Fermi C500,&#34; completing the final piece of its full-stack self-development puzzle.
This ability to transform &#34;precision optical instruments&#34; into &#34;standardized semiconductor products&#34; is the key to maintaining gross margins during price wars. In 2025, against a backdrop of continuous ASP declines, the company's consolidated gross margin remained stable at 41.8%.
At the same time, industry differentiation is accelerating. Hesai's annual shipments climbed significantly, further squeezing the market share of small and medium-sized manufacturers.
In the current supply chain system, automakers have an extremely low tolerance for error from suppliers. Once leading enterprises cross the mass production threshold of one million units, their advantages in supply chain negotiation, yield control, and delivery stability will continue to widen. This also means that second- and third-tier LiDAR companies that have not yet achieved self-sustained growth will face a survival problem rather than a growth problem in 2026.
<h2>02 The Route Contest Continues</h2>
In contrast to the strong performance in Hesai's financial report, the automotive industry's attitude toward LiDAR in 2025 has become polarized.
Absolute shipment volumes are rising, but this is largely due to the increased penetration of intelligent driving features in mainstream models priced between 150,000 and 200,000 yuan. For example, in 2025, LiDAR penetration in new energy passenger vehicles reached 17%, crossing the &#34;chasm threshold&#34; of 16% for the first time (the book Diffusion of Innovations considers a 16% penetration rate as the &#34;critical point&#34; for whether a new technology can &#34;cross the chasm,&#34; truly explode, and become popular), rather than an absolute demand for the hardware itself from car companies.
Currently, the fierce competition in China's new energy vehicle market has permeated every component. Under the immense pressure of vehicle-wide cost reduction, LiDAR, whose unit price still hovers in the hundreds of dollars, is the first to be targeted.
LiDAR units that cost 5,000–6,000 yuan three years ago have now plummeted to 1,500–3,000 yuan, a drop of over 70%. Leading manufacturers are pushing prices down further toward the $200 (approx. RMB 1,400) range, aiming for application in models in the 150,000-yuan class.
Wall Street News has learned that several automakers that have long used intelligent driving as a core selling point, including Xpeng, have introduced &#34;light-LiDAR&#34; or even LiDAR-free intelligent driving solutions by strengthening pure vision algorithms and increasing vehicle-side computing power. Since 2024, models such as the Xpeng MONA M03, Xpeng P7+, and the new G6/G9 have all removed LiDAR.
Yuan Tingting, Xpeng's Director of Autonomous Driving, stated in an interview with CarNewsChina last year: &#34;Removing LiDAR is a clear choice.&#34; The underlying logic is that the new AI systems are built on large language models based on massive data, and LiDAR data cannot be effectively absorbed by these AI systems.
He Xiaopeng even predicted: &#34;The choice between the two paths might be a problem now, but by 2027 it likely won't be.&#34; On the other hand, brands affiliated with Huawei, Li Auto, and NIO continue to use multi-sensor fusion, viewing LiDAR as an indispensable configuration for flagship models.
However, many still regard LiDAR as a &#34;safety net,&#34; and public perception is shifting. Two years ago, a few &#34;horns&#34; on the roof were a symbol of high-end status and technology. Today, equipping LiDAR is for a set of &#34;safety redundancies&#34; that cannot afford to fail.
Richard Yu, Executive Director of Huawei, explained in a science video released in 2025 that cameras struggle in backlighting, at night, or in rain and fog, and millimeter-wave radar has low recognition precision. In contrast, LiDAR can form a 3D point cloud by emitting laser beams, providing sufficiently high recognition accuracy.
His conclusion is that only a perception solution integrating all three types of hardware can provide comprehensive safety assurance for assisted driving. Hesai CEO Li Yifan likened LiDAR to an &#34;invisible airbag,&#34; emphasizing its transition from a functional component to a safety component.
Zhang Chenghang, an analyst at Huachuang Securities, recently predicted in a research report that with the trend toward the democratization of intelligent driving, LiDAR is expected to reach a scale-driven turning point. The global automotive LiDAR market is projected to reach $9 billion by 2030 and $14.8 billion by 2035.
Currently, automakers still choose whether to equip LiDAR based on their own technical maturity and cost considerations. The controversy surrounding the LiDAR field has yet to fade.
<h2>03 End-Game Speculations Under the AI Leap</h2>
If we take a longer-term view alongside current autonomous driving trends, the long-term prospects for the LiDAR industry face a contradiction: the ultimate leap in intelligent driving technology may actually weaken the reliance on high-precision physical sensors like LiDAR.
2025 is the year of concentrated implementation for end-to-end autonomous driving models. When systems no longer rely on rule-based code written by engineers but are trained directly on massive video data to achieve a seamless &#34;perception-decision-control&#34; closed loop, the demand for 3D high-precision point cloud data diminishes.
If vision AI becomes powerful enough to infer 3D space and depth information from 2D images as humans do, the value of LiDAR will be diluted.
The logic behind Tesla and Xpeng’s shift to pure vision lies here—high-computing chips allow systems to process massive image data, whereas LiDAR’s point cloud data is difficult to integrate into end-to-end models.
Therefore, the more mature autonomous driving algorithms become, the weaker the reliance on LiDAR may get.
Even if the industry reaches a consensus that LiDAR must be retained for that 0.01% safety margin, its future growth logic will change. As products become commoditized, LiDAR may turn into a high-volume, standardized industrial product, leaving manufacturers with narrower hardware price premiums.
However, policy dividends for Level 3 (L3) autonomous driving are introducing new variables.
In December 2025, the Ministry of Industry and Information Technology announced the first batch of L3 autonomous driving vehicle access permits, with the Changan Deepal SL03 and BAIC Arcfox Alpha S (L3 version) successfully approved for pilots. This means that as responsibility shifts from the driver to the automaker under L3, LiDAR moves from optional to mandatory. The number of units per vehicle is expected to increase from the current one to between three and six, bringing new opportunities to the industry.
For leading companies like Hesai, the key to a breakthrough may lie outside the automotive sector. In 2025, Hesai's LiDAR deliveries in the robotics field reached approximately 240,000 units, a year-on-year increase of 425.8%. This growth rate far exceeds that of its automotive business. RoboSense's robotics business also showed strong performance, with a gross margin as high as 45%, far exceeding the 17.4% of its automotive business.
It is understood that Unitree Robotics' entire line of robots, equipped with Hesai's JT series LiDAR, appeared on the 2026 Spring Festival Gala, with cumulative deliveries of the series exceeding 200,000 units. Hesai also signed a major exclusive supply contract with the Dreame ecosystem for 10 million JT LiDAR units, a record in the consumer robotics field.
With the development of General Artificial Intelligence (GAI), fields such as embodied intelligence robots and autonomous logistics vehicles are on the verge of an explosion, and LiDAR will have a broad range of applications. Morgan Stanley predicts that by 2050, global demand for robotic LiDAR will increase nearly 300 times compared to 2025.
Unlike cars driving on structured roads, robots operate in complex, unstructured 3D physical worlds, requiring more stringent and precise spatial perception. This non-automotive grade market is becoming the next blue ocean for absorbing LiDAR production capacity.
Hesai has proven its victory in the first half of the hardware elimination round with a strong 2025 financial report—building a cost moat through proprietary chip development and diluting fixed costs through scale effects to become the first to achieve profitability in the LiDAR track.
But in an era where AI large models are evolving at lightning speed, the ultimate competitor for the LiDAR industry is no longer its peers, but the constantly self-evolving &#34;pure vision AI.&#34;
The hardware cycle, after all, depends on the software’s lead.

HSAI

02525

The "non-consensus" future of LiDAR

- Hesai Technology reported a record revenue of 3.03 billion RMB for 2025, marking a 45.8% year-on-year increase and achieving GAAP profitability for the first time.  
- This performance has revitalized confidence in the LiDAR industry amidst ongoing debates about the necessity of LiDAR in intelligent driving against the backdrop of a significant drop in prices and competition from visual technology.  
- Despite the strong showing, the long-term future of LiDAR may be challenged by advancements in AI and autonomous driving systems that increasingly rely on computer vision, potentially diminishing the reliance on high-precision sensors like LiDAR.

Robots "Feed" Hesai: Is Embodied Intelligence the Next Frontier for LiDAR?