Robots "feed" Hesai; is embodied intelligence the next stop for lidar?
```
Author | Wang Xiaojuan
Editor | Zhou Zhiyu
On March 24, Hesai Technology released its unaudited financial statements for the fourth quarter and full year of 2025.
Amid years of doubts about the “burn-money-for-scale” strategy, Hesai boosted confidence in the lidar industry with a record annual revenue of 3.03 billion RMB (up 45.8% year-over-year) and its first-ever full-year GAAP profit.
In 2025, as debates in the smart driving sector between pure vision and lidar technology pathways intensify, the release of these results is thought-provoking.
On one hand, lidar is accelerating onboard adoption as intelligent driving becomes more pervasive; on the other, as carmakers aggressively squeeze costs, lidar is often the expensive hardware most likely to be optimized out.
Is lidar a necessity in the era of smart driving, or a luxury during a technological transition? As smart driving penetration increases, the market is giving its answer; meanwhile, this hardware seems to have found more intelligent applications beyond automobiles, opening new markets.
01 The Tipping Point of Profitability Under Scale Effect
The most noteworthy aspect of Hesai’s financial report is not only the growth in revenue, but how its profits are changing.
For hardware manufacturing, “non-economies of scale” once plagued all lidar manufacturers. In recent years, high R&D costs, BOM expenses, and annual price reductions demanded by carmakers sent many manufacturers into a trap where “the more you sell, the more you lose.”
In 2023, Hesai lost 840 million RMB, Suteng Innovation lost 760 million RMB, and the industry was stuck in a dilemma of “losing money to gain attention.” However, Hesai’s 2025 financials have proven that this business can be profitable.
Hesai became the world’s first lidar company to achieve full-year GAAP profitability due to a fundamental restructuring of its technical architecture. Traditional lidar relied on many discrete components, complex assembly, and high costs.
Through self-developed (ASIC) chipsets for transmit/receive and processing ends, Hesai not only greatly reduced the physical size, but also pushed core component cost onto Moore’s Law trajectory.
According to company disclosures, through continuous chip innovation over eight years, Hesai drove lidar costs down by 99.5%. As of November 2025, its cumulative self-developed chip shipments have reached 185 million units, ranking first globally. In 2025, Hesai also launched its self-developed main control chip, the "Fermi C500," completing the last piece of its full-stack self-development puzzle.
This capability of turning “precision optical instruments” into “standardized semiconductor products” is key to maintaining gross margins in price wars. In 2025, despite a declining ASP, the company’s overall gross margin stayed steady at 41.8%.
Meanwhile, industry polarization is accelerating, with Hesai’s annual shipment volume surging, further squeezing the shares of smaller manufacturers.
In the current supply chain system, carmakers have minimal tolerance towards suppliers. Once leading firms cross the one-million-unit mass production threshold, their advantages in supply chain bargaining, yield control, and delivery stability grow ever larger. For second- and third-tier lidar companies that have not achieved self-sufficient cash flows, the issue in 2026 will shift from growth to survival.
02 The Route Battle Continues
In contrast to Hesai’s strong financial performance, 2025 shows carmakers diverging in their attitudes toward lidar.
The absolute shipment numbers are rising, mainly due to increased penetration of intelligent driving features in mainstream models priced at 150,000–200,000 RMB. For example, in 2025, lidar penetration in new energy passenger vehicles reached 17%, crossing for the first time the “chasm” critical threshold of 16%. (According to "Diffusion of Innovations," 16% penetration is the tipping point at which a new technology crosses the chasm and truly explodes into broad adoption.) This reflects market penetration rather than absolute hardware demand by automakers.
Currently, intense competition in China’s new energy vehicle market has reached every component. Under the heavy pressure of overall vehicle cost reductions, lidars, still priced in the hundreds of dollars, are the first to be targeted.
Three years ago, a single lidar sold for 5,000–6,000 RMB; now, prices have plummeted to 1,500–3,000 RMB, a drop of more than 70%. Leading manufacturers are pushing prices down further, to the $200 (about 1,400 RMB) range, aiming to install on models priced at 150,000 RMB.
Wall Street News has learned that companies such as XPeng, which have long marketed smart driving as a core selling point, have launched "light radar" or even radar-free smart driving solutions by strengthening pure vision algorithms and increasing in-vehicle computing power. Since 2024, models such as XPeng MONA M03, XPeng P7+, and the new G6/G9 have all eliminated lidar.
XPeng’s Autonomous Driving Director Yuan Tingting said in an interview with CarNewsChina last year: "Removing lidar is a clear choice." The logic is that new AI systems are built around large language models trained on massive data, and lidar's data cannot be effectively absorbed by this AI system.
He Xiaopeng even forecasted: “Right now choosing between two paths may be an issue, but by 2027, it might not be.” But on the other hand, companies such as Huawei, Li Auto, and NIO still use multi-sensor fusion, considering lidar an indispensable configuration for flagship models.
However, many still view lidar as a “safety net,” and perceptions are shifting. Two years ago, having several “horns” on the roof signaled high-end tech. Now, lidar is for the sake of an uncompromising “redundant safety system.”
Huawei executive director Yu Chengdong explained in a 2025 science video: Cameras struggle in backlight, nighttime, or rain/fog conditions; millimeter-wave radar doesn’t have high precision; but lidar forms three-dimensional point clouds via laser beams, providing sufficient recognition accuracy.
His conclusion is that only integrating the three hardware systems can provide comprehensive safety for assisted driving. Hesai CEO Li Yifan likens lidar to a “hidden airbag,” emphasizing its shift from functional to safety equipment.
HuaChuang Securities analyst Zhang Chenghang recently predicted in a research report: As smart driving becomes more accessible, lidar is likely to reach a scaling inflection point. By 2030, the global automotive lidar market will reach $9 billion, and $14.8 billion by 2035.
Currently, automakers continue to weigh their own technology maturity and cost considerations in choosing whether to equip lidars, and controversies around lidar remain unresolved.
03 Endgame Speculation Under AI Transformation
Looking further, based on current autonomous driving trends, the lidar industry faces a paradox: the ultimate leap in smart driving technology may actually reduce reliance on high-precision physical sensors like lidar.
2025 is the year of concentrated deployment of end-to-end autonomous driving models. When the systems are no longer dependent on rules written by engineers but instead use massive video data to directly train neural networks, achieving seamless "perception-decision-control," the demand for three-dimensional high-precision point cloud data weakens.
If vision AI is powerful enough to infer three-dimensional spatial and depth information from two-dimensional images like humans, the value of lidar will be diluted.
This is the logic behind Tesla and XPeng’s shift to pure vision: High-power chips allow the system to process vast image data, but point cloud data from lidar is hard to integrate into end-to-end models.
Thus, maturity in autonomous driving algorithms may lead to less dependency on lidar.
Even if the industry agrees to keep lidar for 0.01% ultimate safety, its growth logic will change. As products become homogenized, lidar may become a standardized industrial commodity with narrowing margins for manufacturers.
However, the policy dividends of L3-level autonomous driving are bringing new variables for lidar.
In December 2025, the Ministry of Industry and Information Technology issued the first batch of L3-level autonomous vehicle trial permits, with Changan Deep Blue SL03 and BAIC Arcfox Alpha S (L3 version) successfully approved for pilots. This means that with L3-level responsibility shifting from drivers to carmakers, lidar’s status goes from optional to mandatory, with per-vehicle installations likely increasing from the current one to three–six, bringing new opportunities for the lidar industry.
For leading companies like Hesai, the breakthrough may lie beyond cars. In 2025, Hesai delivered about 240,000 lidars in the robotics sector, up 425.8% year-over-year. This growth far outpaces its automotive business. Suteng Innovation’s robotics business also performed well, with gross margins of 45%, much higher than the 17.4% from the automotive business.
It’s reported that Unitree Robotics’ entire line equipped with Hesai JT-series lidars appeared in the 2026 Spring Festival Gala; the series has exceeded 200,000 deliveries. Hesai signed an exclusive supply deal for 10 million JT radars with Dreame, setting a record for the consumer robotics sector.
With the rise of general artificial intelligence, fields like embodied intelligent robots and unmanned logistics vehicles are on the brink of boom, and lidar will have broad space for applications. Morgan Stanley predicts that by 2050, global robotic lidar demand will increase nearly 300-fold compared to 2025.
Unlike cars driving on structured roads, robots operate in complex, unstructured three-dimensional environments, requiring much higher spatial awareness. This non-automotive market is becoming the next blue ocean for lidar capacity.
With its robust 2025 financials, Hesai Technology has proven its victory in the first round of the hardware knockout—building a cost moat with self-developed chips and spreading fixed costs through scale effects, achieving first profitability in the lidar track.
But in the era of rapidly advancing AI foundation models, the ultimate rival for the lidar industry is no longer the competition, but ever-evolving “pure vision AI.”
The hardware cycle, after all, depends on the software.Risk Warning and Disclaimer ClauseThe market carries risks, and investment requires caution. This article does not constitute personal investment advice, nor does it consider the unique investment goals, financial situation, or needs of individual users. Users should determine whether any opinions, viewpoints, or conclusions in this article suit their specific circumstances. Any investment based on this is at their own risk. ```