Hesai: Tesla's 'Abandoned Child'? Unstoppable 'Magnificent Comeback' of LiDAR

In the first part of Hesai's article "'4x' Hesai: Why Has the Lidar Abandoned by Tesla Become 'Bright' Again?", Dolphin Research detailed the reasons for Hesai's market capitalization explosion. In the high-growth lidar sector, Hesai continues to reduce costs through chip integration and increased integration, leading to a continuous decline in lidar costs, thereby enhancing penetration rates and expanding application scenarios.

Following the Q3 2024 report, the market value of $Hesai(HSAI.US) began to rise sharply due to earnings consistently exceeding market expectations. The nonlinear cost reduction in lidar technology has driven significant increases in revenue and shipment volumes. Currently, the new product ATX is gradually replacing AT128 as the main shipment driver for passenger car radars.

In this study, Dolphin Research will continue to address the following two questions:

1. How to view the market space for lidar?

2. In the highly concentrated CR3 lidar track, can the industry pattern of the strong remaining strong be maintained?

3. Is there still room for investment in Hesai?

I. How to view the market space for lidar?

Lidar, like the chip industry, has strong scenario expansion capabilities. Currently, lidar shipment scenarios are mainly concentrated in L2 passenger cars and L4 Robotaxi. Although the current shipment in the robot scenario is mainly lawn mowers, there is still potential for large-scale application unlocking in scenario expansion.

Passenger Car ADAS Track: Lidar is on the verge of an explosion, with penetration rates about to cross the "0-1" to "1-N" inflection point stage:

From the perspective of China's vehicle-mounted lidar penetration rate, the penetration rate of lidar in passenger cars was only 7% in 2024, and only 14% in new energy passenger cars. However, Dolphin Research believes that with rapid technological iteration, the penetration rate of lidar will usher in an inflection point investment opportunity stage from "0-1" to "1-N" in the next three to five years starting in 2025:

① From the supply side, cost reduction through technological innovation and high product maturity drives the popularity of downstream applications:

From the perspective of Hesai's lidar prices, lidar prices have been rapidly declining. From 2020 to 2024, the average price (blended ASP) of lidar dropped from 82,000 yuan to about 3,900 yuan in 2024, with an average price drop of -51% over five years. In 2025, the ATX, which is concentrated in passenger car shipments, is priced at less than $200, equivalent to the "thousand-yuan phone" era of smartphones, making it more suitable for large-scale expansion in high price-sensitive passenger car scenarios.

The rapid decline in lidar costs is mainly driven by the following factors:

a. Architectural innovation: Cost reduction from mechanical to semi-solid state

According to the classification of lidar on the scanning end, lidar is divided into mechanical, hybrid solid-state, and solid-state. The first installation of lidar on the vehicle end (L4 Robotaxi) was the solid-state Pandar series, with a cost of nearly 100,000 yuan per unit in 2020. The high cost limited lidar applications to commercial vehicle Robotaxi pilots. In the passenger car application scenario, which is naturally highly price-sensitive, a significant expansion requires affordable lidar, and prices must drop significantly.

Hesai first released the ATX128 series in August 2021, which entered mass production in 2022. The price dropped sharply from 100,000 yuan per unit for the Pandar series to only 5,000 yuan per unit at the time of mass production in 2022. The core reason for the price drop was the cost reduction brought by switching the scanning architecture from mechanical to semi-solid state.

Although mechanical lidar has the advantage of achieving 360-degree panoramic scanning without blind spots, its complex structure results in a long production process, requiring cumbersome manual assembly and calibration, leading to high manufacturing costs, difficulty in mass production, and susceptibility to damage.

In the ATX series, Hesai replaced mechanical rotation with a one-dimensional rotating mirror in the scanning structure, removed the expensive large motor and heat dissipation module of the mechanical rotation structure, and replaced multiple lens groups with aspheric lenses, reducing the labor cost of dynamic balance calibration, ultimately achieving significant cost reduction. The ATX is 78% lighter than the Pandar128 (360g vs. 1630g) and has only 22.6% of the volume (0.3L vs. 1.33L).

b. Chip integration and integration bring cost reduction:

Hesai's ATX integrates the number of lasers and detectors from 256 discrete devices into 2 ASIC chips through chip integration, achieving self-developed laser emission chip VCSEL, laser receiving chip SPAD, and control chip ASIC, significantly reducing material costs;

Emission end: Hesai replaced EEL with VCSEL lasers on the emission end, using multi-junction technology to improve power density;

Receiving end: SPAD array integration (integrating the original 128 channels into one chip);

Control end: Self-developed SoC replaces FPGA;

At the same time, the chip design also enables lidar to achieve "Moore's Law": chip performance doubles every 18-24 months, and prices halve. Therefore, Hesai's lidar can achieve smaller size, lighter weight, longer detection distance, and higher precision advantages.

Currently, the price of the ATX series is less than $200 (about 1,400 yuan), a 72% drop from the approximately 5,000 yuan of the AT128 when it was first mass-produced in 2022. Chip integration and integration have further reduced lidar prices, making the ATX more suitable for large-scale expansion in high price-sensitive passenger car scenarios, becoming the main driver of Hesai's shipments in 2025.

2. From the demand side:

① Passenger car intelligent driving is still at the tipping point of an explosion:

In 2024, only 7% of new cars in China were equipped with L2+ level ADAS functions, capable of achieving highway and urban NOA. However, since 2025, mainstream car companies such as BYD, Geely, and Changan have launched "intelligent driving equality," bringing highway and urban NOA to models priced between 100,000 and 200,000 yuan. Although the immature technology of intelligent driving and the accident of Xiaomi SU7 have poured cold water on the industry, it has instead raised consumer awareness of lidar as a redundant safety line. After the Xiaomi car accident, the Xiaomi SUV YU7, launched in July, has started to come standard with lidar across the board.

Dolphin Research expects that as electrification innovation nears its end, competition among car companies around intelligence will only become more intense. Referring to the trend of the mobile phone industry from "feature phones" to "smartphones," the transition from electric vehicles to smart cars is a 5-10 year industrial opportunity. Dolphin Research expects that the popularization of L2-level ADAS will still be driven by mainstream car companies such as BYD, Geely, Great Wall, and Changan in 2026.

Leapmotor, with its main model price range of 100,000-200,000 yuan, plans to achieve urban NOA by the end of the year. XPeng will also lower urban NOA to the 100,000-150,000 yuan model Mona M03. Dolphin Research expects that BYD may further lower urban NOA to models below 150,000 yuan in 2026 (currently, BYD models equipped with urban NOA are priced at 200,000 yuan).

② Under the trend of declining lidar prices, lidar assumes the dual role of "intelligent driving eye" and "safety redundancy":

In the downstream vehicle intelligent driving, there has always been a debate between the "pure vision route" led by Tesla and followed by XPeng, and the "multi-sensor fusion route" led by Huawei and Li Auto. The technology stack of intelligent driving has not yet converged, leading to market controversy over the function of lidar as the "intelligent driving eye."

However, Dolphin Research believes that under the price parity brought by cost reduction, even if the "intelligent driving eye" function is controversial, the Xiaomi car accident has verified that the opportunity for penetration rate improvement brought by the redundant safety function is very high for a car costing around ten thousand yuan with a lidar costing around a thousand yuan:

Pure vision, which uses a camera-based perception solution, is limited by optical physical laws and is easily affected by strong light, darkness, heavy rain, snow, and fog, unable to break through optical derivation limits. The "probability guessing" nature of the vision solution determines a higher false detection rate.

Although lidar performance also declines in adverse weather such as rain, snow, and fog, its reliability in extreme weather is significantly higher than that of pure vision solutions. Since lidar data is directly measured 3D point cloud data, not the "probability guessing" nature of vision solutions, the false detection rate is lower, the response speed is faster, and the prediction accuracy is higher, resulting in a higher AEB trigger success rate, making it more suitable for serving as a "safety airbag" function in "corner cases":

1. Lidar reliability in extreme weather and corner case scenarios is significantly higher than pure vision solutions:

a. Higher reliability in rain and fog scenarios:

In heavy rain/fog, lidar detection distance attenuation is only 30%-40% (vision attenuation exceeds 70%), and point cloud noise can be filtered through algorithms to maintain usability. In heavy rain, pure vision recognition distance is less than 40 meters, unable to meet AEB braking requirements.

According to the 2024 FSD accident investigation report by the US NHTSA, the AEB failure probability in daily weather is 0.8 times per million miles, but in heavy rain, the AEB failure probability is 4.2 times per million miles. Pure vision, due to optical scattering and feature point deficiencies, has a 5-fold increase in AEB failure probability in heavy rain compared to daily weather.

b. Lidar has stronger adaptability to extreme lighting:

In strong backlight/tunnel entrance and exit scenarios, lidar, due to its active light source emission, is not affected by ambient light interference (active emission of 905nm infrared light source, isolated from ambient light spectrum), generating stable point clouds with target recognition delay <50ms. Pure vision solutions need to dynamically adjust exposure parameters, and strong light causes image overexposure, leading to motion blur and feature loss, easily resulting in target recognition failure, with high target recognition delay (vision solution delay 200-300ms).

c. Lidar has advantages in corner case scenarios:

In close-range cut-in scenarios, pre-judging the intention to cut in:

Lidar accurately calculates the relative speed and acceleration of the cut-in vehicle through real-time point cloud tracking (10Hz refresh), while pure vision relies on image feature point matching, with frame-to-frame matching errors leading to increased delay.

Facing irregular obstacles: When identifying small targets such as cones and spilled objects, the lidar solution has a higher recognition distance, lower false detection rate, and lower missed detection rate (pure vision false detection rate >15%).

In complex urban road conditions (pedestrians + bicycles + electric vehicles mixed crossing): The accuracy of trajectory prediction and the detection rate of occluded targets are higher than pure vision solutions. The core principle is that lidar data is directly measured 3D point cloud data, while pure vision solutions convert 2D images into 3D solutions through BEV, which is prone to failure in occlusion scenarios. The "probability guessing" nature of vision solutions determines a higher false detection rate.

2. Lidar serves as a "safety airbag" function in extreme weather and corner cases, especially as ATX enters the "thousand-yuan phone" era, this logic becomes very smooth:

a. The implementation of L2 strong standards forces OEMs to standardize lidar:

In February 2025, the "Technical Requirements and Test Methods for Light Vehicle Automatic Emergency Braking Systems" was drafted and entered the public consultation stage on April 30. This standard will replace the current national standard GB/T 39901-2021, upgrading to China's first mandatory national standard for L2-level ADAS, marking the entry of intelligent driving safety regulation into a hard compliance era.

Compared to the current national standard, the new national standard requires comprehensive upgrades in AEBS coverage, activation speed, and false detection & missed detection:

1) Coverage expansion: The new national standard expands coverage from M1 passenger cars to N1 light commercial vehicles (annual sales of over 3.5 million units), achieving integrated safety regulation for commercial and passenger vehicles;

2) AEB activation speed range improvement: The AEBS system activation speed range for M1/N1 vehicles is changed from above 15km/h to 10-80km/h and 10-60km/h, respectively, while clearly defining the activation speed range for effective recognition of pedestrians and two-wheelers as 20km/h-60km/h;

3) Addition of extreme scenarios: The false response test adds corner case scenarios such as right-turn following, stationary bicycles, and pedestrians moving in the same direction, while upgrading test props: the thickness of the iron plate in the lane is increased from 10mm to 25mm.

4) Improvement in experimental missed detection rate: Tightened from ≤40% to ≤10-20%.

Lidar can significantly increase the AEB speed limit by 50% (daytime AEB speed limit without lidar increases from 85km/h to 140km/h, nighttime AEB limit without lidar increases from 80km/h to 120km/h), while better passing tests in extreme scenarios and significantly reducing missed detection rates.

Due to the superior ranging capability of lidar compared to pure vision solutions, it can provide the system with more time to make decisions and respond, significantly expanding the speed range covered by the AEB system while improving safety.

Although the L2 strong standard itself does not mandate the use of lidar, due to the increased testing threshold of the strong standard, OEMs with "lowland" intelligent driving technology find it difficult to pass the strong standard test and need to standardize lidar solutions to make up for the shortcomings of intelligent driving.

Currently, the penetration rate of lidar in the 100,000-200,000 yuan range is only 15.4%, and the mid-to-low-end market is expected to become the main incremental market for lidar, especially under the implementation of the L2 strong standard policy.

Data source: Guojin Securities

b. L3 autonomous driving, due to the transfer of accident responsibility to car companies, provides more motivation for car companies to equip lidar as "safety redundancy"

Policy: The introduction of regulations such as the "Beijing Autonomous Vehicle Regulations," "Wuhan Intelligent Connected Vehicle Development Promotion Regulations," and "Shenzhen Special Economic Zone Intelligent Connected Vehicle Management Regulations" has accelerated the commercialization of L3 autonomous driving by specifying key aspects such as the road process and accident liability definition for L3 autonomous vehicles.

The transfer of L3 responsibility forces the continued upgrade of safety redundancy, and lidar may be a more cost-effective solution:

Unlike L2, the accident liability subject of L3 will shift from the user to the OEM. OEMs are more focused on improving the safety of L3-level autonomous driving to ensure that the cost of liability does not exceed the potential benefits of deploying L3-level systems, requiring higher safety redundancy.

In addition to serving as the "intelligent driving eye," lidar plays another important role as "safety redundancy" when intelligent driving technology has not yet matured, especially as a better solution than pure vision solutions for corner cases (strong and weak light, rain, fog, etc.), significantly reducing accident rates and compensation costs, making it a more cost-effective solution for car companies.

c. Lidar entering the thousand-yuan phone era can serve as a "safety airbag," and it is becoming a trend to standardize multiple lidars for L3 and above:

Lidar entering the thousand-yuan phone era can serve as a "safety airbag," transitioning from "technical optional" to "safety necessity":

With ATX achieving significant cost reduction through chip integration and increased integration, and the scanning module switching from traditional mechanical architecture to semi-solid state, it has entered the thousand-yuan phone era. Even for models priced at 100,000 yuan, the cost of lidar accounts for only about 1% of the BoM cost, making it fully capable of serving as a "safety airbag" component.

With the continued cost reduction driven by the chip integration Moore effect, it is expected that lidar prices can continue to decline, promoting installation on the most popular models priced at 100,000 yuan, becoming the main driver of lidar shipment growth in 2026.

It is becoming a trend to standardize multiple lidars for L3 and above:

At the same time, it has become an industry trend to standardize multiple lidars for L3 and above. From the current high-end models of Huawei Aito M9 and ZunJie S800, the number of lidars has increased from 1 in the 2024 model to 4 in the 2025 model (including 1 192-line main radar + 2 side blind spot radars + 1 rear blind spot radar, achieving 360-degree coverage without blind spots).

Hesai has also released a multi-radar perception solution for L3 and L4 levels, with L3 consisting of 1 ETX main radar and 2 solid-state lidar FTX as blind spot radars, and L4 consisting of 4 AT1440 main radars and 4 FTX as blind spot radars. It is expected that with the accelerated implementation of L3 and above autonomous driving plans by leading car companies, Hesai's multi-lidar solution will begin to scale in 2026-2027, with higher single-vehicle value.

II. How to view the competitive barriers of lidar?

At the industry Beta level, both supply and demand sides are highly certain trends. Can the industry competition pattern ensure that the leading position remains strong, fully enjoying the industry's growth dividends?

From the perspective of domestic competition, there are only four players in the lidar market: Hesai, Suteng, Huawei, and Tudatong. Currently, the market concentration of lidar is extremely high. Since Huawei's main lidar customers are in the intelligent selection car mode and HI mode (integrated intelligent packaging shipments), there is basically no external supply, and Tudatong's lidar is mainly supplied to Nio (the 1550nm wavelength solution is costly). Therefore, in the domestic ADAS lidar third-party market track, the actual competition is focused on the competition between the leading Hesai and the second Suteng.

a. Lidar technology iterates quickly, and the technology stack has not yet shown a convergence state;

From the current lidar technology stack, the lidar technology solution has not yet shown a convergence trend. Hesai's current lidar receiving module has iterated from SiPM to SPAD, and the emission module has iterated from EEL to VCSEL module, but the semi-solid state solution of the scanning module may still be a transitional stage for lidar.

Solid-state lidar has the advantages of the simplest structure, highest integration, smallest size, and lower cost, but in the short term, it is limited by low technical maturity, low emission power, and detection distance, so it is generally used as a blind spot radar. However, the breakthrough direction of lidar in the medium term may still iterate towards solid-state lidar, and the technology stack has not yet shown a convergence state.

At the same time, the chip integration of lidar also has the advantage of "Moore's Law," with fast technology iteration:

The point cloud density and ranging capability of lidar continue to evolve under the "Moore's Law" of chip integration. Hesai's current chip integration architecture has iterated to version V4.0, with fast chip integration technology iteration speed. V4.0 adopts 3D stacking technology + 256-core intelligent engine, and ATX/AT512/AT1440 are all products under version V4.0, achieving another significant improvement in point cloud density and a significant decline in cost.

Among them, the point cloud density of the high-end radar AT512 is 8 times higher than that of AT128, and the maximum detection distance has increased from 260 meters for AT128 to 400 meters. The price of the main shipment ATX is already less than $200 (about 1,400 yuan), a 72% drop from the approximately 5,000 yuan of the AT128 when it was first mass-produced in 2022, and the current price is only half of that of the AT128.

Dolphin Research believes that the fast technology iteration speed of lidar, the efficiency of Moore's Law, strong scale effects, and the ecological binding with OEMs constitute the moat of the lidar track. The dual-factor resonance makes it difficult for new players to catch up, and leading players can achieve the "strong remain strong" effect.

b. Lidar has strong scale effects:

Due to the "volume increase and price drop" technology cost reduction penetration route of lidar, it constitutes a strong scale barrier. Hesai has already achieved significant lidar shipments, with shipments of 1.2-1.5 million units in 2025, establishing a scale effect barrier, bringing significant cost reduction, effectively preventing new players from entering the market.

In 2024, production costs already accounted for 31% of lidar costs, and Hesai still has a strong scale effect cost reduction advantage, making it difficult for new entrants to break through the mass production cost threshold.

c. OEM ecological binding + long car-grade certification:

Car-grade certification requires 50 tests + 30,000 hours, and Hesai has reached a high degree of ecological binding with OEMs.

For car companies, the fast technology iteration of lidar + low value + does not affect the coupling degree of car companies' self-developed algorithms makes the efficiency of car companies' self-research willingness not high:

a. Lidar has low value, only ordinary components:

Lidar has relatively low value. Currently, the price of ATX has dropped to less than $200 (less than 1,400 yuan). The "Moore's Law" of chip integration can support the continuous reduction of lidar costs. Currently, the value of lidar is only the value of ordinary automotive components, accounting for less than 1% of the single-vehicle BOM cost, while intelligent driving chips still account for 5%-10% of the car company's BOM cost. The ROI of car companies' self-research on lidar is not high.

b. Lidar, unlike chips, does not affect the coupling degree with car companies' self-developed algorithms:

As the "sensory" system of intelligent driving, lidar is different from the "brain" of intelligent driving chips. The core competitiveness of car companies is the algorithm definition right (the collaboration of intelligent driving software and hardware chips), not the physical performance of the hardware. Car companies' self-developed customized chips can adapt to the car company's exclusive algorithms, avoiding binding by chip manufacturers.

The interface of lidar has been fully standardized and does not affect the coupling degree with car companies' self-developed algorithms, nor does it have algorithm binding risks.

III. How to view the market space for lidar?

① Passenger car track

a. Domestic market: "Technical parity" and "responsibility transfer" force lidar to become standard:

New energy vehicle sales: Assuming that by 2030, China's annual passenger car sales will reach 30 million units, with a new energy vehicle penetration rate of 90% (high-level intelligent driving is naturally coupled with electric vehicles, and the maturity of intelligent driving also promotes a second leap in new energy penetration), new energy vehicle sales will total 27 million units.

Intelligent driving penetration rate: Assuming that by 2030, the penetration rate of L3 intelligent driving will reach 65% (intelligent driving functions have reached conditional urban NOA), the penetration rate of L4/L5 high-level intelligent driving will reach 30%, and the penetration rate of L2.5 and below will only be 5% (highway NOA), the penetration rate of intelligent driving in new energy vehicles will reach 100%, indicating that the stage of iteration from feature phones to smartphones in the mobile phone stage has been experienced.

Number of lidars per vehicle:

For L3+ high-level intelligent driving, due to the transfer of accident responsibility from users to car companies, the demand for safety redundancy will require higher standards, leading to a rigid increase in the cost of lidar as safety redundancy. Currently, the industry has seen a trend of multiple lidars being equipped for quasi-L3 high-level intelligent driving led by Huawei.

With the continuous decline in lidar prices, entering the "thousand-yuan phone era" also allows lidar to become a "basic safety component" similar to airbags.

Therefore, Dolphin Research assumes that L4/L5 high-level intelligent driving is equipped with 8 lidars per vehicle, L2.9/L3 vehicles need to be equipped with 3 lidars, and L2.5 vehicles are equipped with 1 lidar as a "basic safety component."

Lidar penetration rate: Due to the transfer of responsibility for intelligent driving above L3 to car companies, the more sufficient the need for safety redundancy, the more the number of lidars equipped per vehicle shows a continuous upward trend. Finally, it is assumed that the penetration rate of L4/L5 lidars in L4/L5 high-level intelligent driving is 90%, L2.9/L3 reaches 62%, and L2.5 penetration rate is 40%. Finally, the average number of lidars equipped per vehicle is calculated to be 3.4.

Value of lidar per vehicle: Dolphin Research expects that the price of lidar in the industry will still follow the "volume increase and price reduction" route under the cost reduction of chip integration and integration technology. It is expected that by 2030, the price of a single lidar will drop to 445 yuan (an average annual decline of 18%), but the number of lidars equipped per vehicle shows a continuous upward trend. Finally, the average value of lidar per vehicle in 2030 is 1,500 yuan.

China's lidar market space: Finally, in China, with 27 million new energy vehicles, the penetration rate of intelligent driving in new energy vehicles reaching 100%, the average number of lidars equipped per vehicle is 3.4, and the value of a single lidar is 445 yuan, assuming an average value of 1,500 yuan per vehicle, the scale of China's lidar market will reach 40 billion yuan by 2030.

b. Overseas market: Lidar can become a key track for the globalization breakthrough of China's supply chain:

Lidar itself, as a sensor with strong manufacturing barriers (testing mass production capabilities) and fast technology iteration speed, is very suitable for the overseas route.

From the current overseas lidar players Valeo and Innoviz, the performance of overseas lidars is still significantly behind Hesai's new generation products (ranging distance, point cloud density), but the terminal price is very expensive.

With its unique "fast technology iteration speed + mass production barriers significantly reducing manufacturing costs" dual attributes, Hesai's lidar has the advantage of "high performance + low cost" compared to foreign competitors, making lidar a key track for the globalization breakthrough of China's supply chain again.

From Hesai's perspective, Hesai has already secured exclusive cooperation with top European car manufacturers, covering multiple models of fuel vehicles and new energy vehicles. It is a long-term cooperation project spanning 2030, and it is the largest order in the overseas pre-installed mass production lidar field so far. Overseas shipments are also expected to start scaling at the end of 2026 and 2027.

The company's acquisition of a global point from a top European OEM this time means that Chinese lidar has truly entered the core supply chain of the global automotive industry, opening up the market for Chinese lidar manufacturers to sell overseas.

At the same time, EU policies and regulations also promote the installation of lidar: According to the UN-R157 standard, which will be mandatory in the EU in 2027, L3 vehicles must have "fail-operational" redundancy capabilities. The redundancy design and fault handling of sensors and systems must ensure the safe operation of the entire vehicle, requiring an independent redundant sensor system to ensure that a single fault does not lead to an accident, further promoting the installation of lidar in Europe.

Due to the lack of truly competitive overseas markets (low performance and high cost), Dolphin Research assumes that the average value of lidar per vehicle in the European market is three times that of the Chinese market, and other markets are 1.5 times that of the Chinese market. The US market is not considered for the time being due to policy sensitivity factors.

Finally, based on assumptions, Dolphin Research expects that by 2030, the global vehicle-mounted ADAS total market size will reach 71.8 billion yuan, with a compound annual growth rate (CAGR) of 73% from 2025 to 2030. Lidar is still at the penetration rate take-off stage, or the penetration rate is at the "0-1" to "1-10" inflection point stage.

c. Robot track - upward option attribute

Lidar, like the chip industry, has strong scenario expansion capabilities. Although the current shipment in the robot scenario is mainly lawn mowers, lidar itself has functions such as positioning and mapping, obstacle perception, navigation path planning, and environmental understanding, and still has the potential for large-scale application unlocking in various robot scenarios.

Because this scenario is still in the "0-1" startup stage, Dolphin Research only makes a simple assumption: by 2030, the global robot shipment volume will be 4.29 million units. Assuming that the average ASP of lidar in consumer/industrial robots is 2,250 yuan, and the average ASP in humanoid robots is 4,000 yuan, it is expected that the market size of lidar in robot application scenarios will reach 11.9 billion yuan in 2020.

IV. Is there still room for investment in Hesai?

Dolphin Research believes that the fast technology iteration speed of lidar, strong scale effects, and ecological binding with OEMs constitute the moat of the lidar track. Leading players can achieve the "strong remain strong" effect, and the concentration of the lidar track will remain high, with leading players forming a "strong remain strong" trend.

Dolphin Research expects that due to fast technology iteration and leading technology, Hesai's global market share will continue to increase from the current approximately 33% to 44% by 2030, forming a leading position with Suteng in China, and accelerating the replacement of overseas lidar players with weak mass production capabilities, weak performance, and high prices.

In terms of profit margins, Dolphin Research expects that after lidar enters the mature stage in 2030, the profit margin in China can be comparable to mid-to-high-end manufacturing, while in the US and European players, due to difficulties in mass production and low performance of finished products, there are almost no truly competitive players, and the profit margin of overseas exports will be higher. Assuming that Hesai's profit margin in the lidar ADAS track in the Chinese market is 15%, and the comprehensive profit margin in the overseas market is 20%.

In the robot track, due to the need for higher-end lidar shipments compared to the passenger car track, and the robot track is still in the development stage, the cost sensitivity of downstream applications is not as high as the passenger car track. Assuming that the profit margin of the robot scenario is 25%.

Finally, Dolphin Research gives Hesai a valuation of 20 times PE in the ADAS passenger car track in 2030, and 25 times PE in the robot track, with a WACC discount rate of 10.8%. Finally, it is expected that Hesai's market value will be 81.5 billion yuan, with a 172% upside potential compared to the current market value of 30 billion yuan (equivalent to the optimistic valuation of considering all scenarios).

However, due to the relatively low certainty of large-scale mass production shipments of lidar in the robot and new energy vehicle markets outside of China and Europe, the neutral market value after removing these two scenarios is about 45 billion yuan, still with a 50% upside potential.