
Why Is Hesai Still Growing If Tesla Doesn't Use LiDAR?
October 4, 2026

The cover image is an official Waymo sixth-generation vehicle photo, published on August 19, 2024.
Tesla is pushing its no-LiDAR approach further into real operations. Reuters reported on September 3 that Cybercab has begun offering rides in parts of Austin. Pure vision has taken a step forward, but this progress cannot answer another question: is LiDAR losing customers?
Hesai's unaudited second-quarter financial report disclosed on August 18 presents a contrast. LiDAR shipments reached 628,300 units, up 78.4% year-over-year; group revenue grew 21.9%, while gross margin fell from 42.5% to 40.1%. More units sold, yet revenue and profit margins did not rise in tandem.
So the real question this article wants to pursue is: as pure vision continues to improve, what costs can LiDAR still save for customers, and how much value can remain on the supplier's income statement? As of October 4, the following uses Hesai's latest disclosed second-quarter operating data.
How Much Can You Save by Removing One Sensor?
The appeal of Tesla's approach is concrete. Eliminating LiDAR means buying one less piece of hardware and reducing the corresponding installation, calibration, and maintenance work; cameras can also directly read traffic lights, lane markings, and road signs. If the software can meet requirements in the target scenarios, those savings certainly have value.
But cameras and LiDAR provide different information. Cameras receive ambient light, and algorithms use that to identify objects and estimate spatial relationships; LiDAR actively emits light to measure distance, recording the positions of surrounding objects as a 3D point cloud—essentially a map composed of many distance measurements. What it adds is a different kind of measurement basis.
This also explains why "stronger algorithms" does not automatically lead to "distance-measuring hardware is useless." Seeing clearly what an object is and reliably knowing how far away it is from the vehicle are related but not entirely identical problems. Conversely, adding an extra radar does not guarantee autonomous driving safety either—sensor occlusion, severe weather, and software judgment still need to be addressed.
Tesla's official website still clearly states that FSD (Supervised) for vehicle owners requires active driver supervision and will not make the vehicle an autonomous car. It cannot be conflated with the separately operated Robotaxi service as the same proof of capability. When comparing approaches, one must first align on whether there is a human fallback, the road scope, and operating conditions.
A Taxi's Math Covers the Whole Day
Waymo offers another option. The sixth-generation solution announced in 2024 includes 13 cameras, 4 LiDARs, and 6 radars. On February 12, 2026, the company announced it began fully driverless operations with this generation system, continuing to use camera, LiDAR, and radar fusion.
Fusion means letting different sensors complement each other. The same February technical note also emphasized that when camera visibility is limited, LiDAR and radar provide redundancy; the new-generation LiDAR leverages industry cost reductions and in-house chip and optical design to lower costs. These are Waymo's technical statements and cannot be directly treated as independent safety evaluations.
For a robotaxi operator, hardware procurement is only part of the cost. Whether vehicles can reliably accept rides during target hours, whether they frequently stop and wait for handling, and how much remote support is needed—all of these change the cost per kilometer.
From this, a business inference can be made. As long as the operational benefits brought by additional distance-measuring information exceed procurement and maintenance costs, customers have reason to keep LiDAR. If pure vision achieves equivalent reliability under the same conditions at a lower total cost, that procurement could also be eliminated. The judgment requires the full set of operational data; demo videos cannot provide the answer.

Screenshot from Waymo's official website, February 12, 2026. The company announced that its sixth-generation system began fully driverless operations and stated it continues to use multi-sensor fusion.
Radar Prices Drop: Customers Are Happy, Manufacturers May Not Be
Hesai's shipments provide evidence that demand still exists. In the second quarter, ADAS LiDAR shipments reached 485,900 units, up 60.1% year-over-year; robotics LiDAR shipments reached 142,400 units, up 193.4% year-over-year. This only proves the company's delivery growth; it cannot be used to infer industry-wide market share, nor can all robotics units be counted as humanoid robots.
The financial report also states that declining average selling prices offset part of the delivery growth, and the gross margin decline mainly came from a higher revenue share of lower-margin products. Price cuts make it easier for customers to adopt, but also require suppliers to keep pushing manufacturing costs down.
There is a common miscalculation here. Dividing group revenue by LiDAR shipments does not yield a comparable radar unit price. The group already has new businesses such as robot actuators, and different LiDAR products have different prices. One must distinguish between price cuts on the same product, product mix changes, and contributions from new businesses—the result of division cannot be written as a price decline percentage.
This makes the commercial pressure of pure vision clearer. Even if automakers ultimately keep LiDAR, they can still demand that suppliers prove the benefit brought by every yuan of procurement. Only manufacturers that can achieve mass production delivery and continuously reduce costs have a chance to turn demand into profit; technical specifications alone may not protect selling prices.

Partial original page from Hesai's second-quarter financial report, August 18, 2026. Shipments grew 78.4%, group revenue grew 21.9%, and gross margin fell from 42.5% to 40.1%. Amounts are in millions of RMB.
There Is Also a New Business Investment on the Income Statement
Hesai's second-quarter group operating profit was approximately 2.2 million yuan, down 90.4% year-over-year. If one stops reading here and concludes that LiDAR is no longer profitable, they would miss the segment table.
The same financial report shows that the LiDAR segment revenue was approximately 816 million yuan, with operating profit of approximately 66.24 million yuan; strategic growth businesses had an operating loss of approximately 64.04 million yuan. Only after offsetting the two does it come close to the group's operating profit.
Therefore, the decline in group profit involves both product pricing and mix as well as new business investment, and cannot be entirely attributed to LiDAR. The existing segment profit shows that this business still has operating income, but it is insufficient to prove future pricing power, let alone represent that every supplier can make money.
Physix Frontier