From Li Yifan's interview: How did Hesai overtake RoboSense in the automotive lidar market?
Original · Creator Alliance

From Li Yifan's interview: How did Hesai overtake RoboSense in the automotive lidar market?

U
user_16675997784
Creator AllianceNov 12, 2026

When Li Auto was developing the L9, it asked Hesai to take the lidar from 64 lines to 128 lines while keeping the size, cost, and quoted price the same. The two sides settled on the specs within a few months; design and mass-production delivery took longer. Hesai would proactively tell Li Auto about problems that might come up in production, and the two sides would figure out solutions together. The AT128 later installed on the L9 was built this way.

Li Yifan recalls the Li Auto L9 project. Image from the Zhang Xiaojun interview.

Li Auto fixed performance, installation space, and price all at once, and Hesai had to build the lidar under those constraints. During prototyping, engineers could tune each unit one by one. In mass production, the automaker has its own production schedule, and the lidar has to be delivered on time. The L9 gave Hesai its first passenger-car order and also taught it how to work with an automaker to take a part all the way to mass production.

By 2024, Li Auto still accounted for more than half of Hesai's domestic installations, and Hesai ranked third on the list. RoboSense ranked first, supplying mass-production vehicles for brands including Zeekr, AITO, and XPeng at the same time. How many cars Li Auto delivered had a big impact on Hesai's installations that year. RoboSense had projects with several automakers and could deliver to several mass-production models at the same time. For Hesai to catch up with RoboSense, it also had to sell lidar into mainstream models beyond Li Auto.

In 2025, RoboSense installed about 60,000 more units domestically than the year before, while Hesai installed about 750,000 more. Both companies' installations were growing, and the overtake happened in a market that was still expanding. My judgment is that Hesai caught the wave of automakers expanding standard fitment. Lidar that had previously only been installed on high trims began to be purchased by automakers for more versions. Hesai was the first to make a lidar affordable for base versions too.

Automakers start buying lidar for base versions

Once, Li Yifan was riding in a company-arranged car and found that the driver was unwilling to use driver assistance. The driver felt more confident driving himself and saw enabling it as adding a layer of occupational risk. Li Yifan used this to talk about airbags and seatbelts. He argued that owners can absolutely drive themselves, and automakers still have reason to equip the car with safety hardware.

I care more about what this means for procurement. Navigate-on-autopilot can be reserved for high trims, but once lidar becomes standard across the lineup, automakers also have to order it for entry-level versions. Entry-level versions are already cheap, leaving limited budget for newly added parts. The premise for automakers to expand standard fitment is that there is a lidar on the market they can afford and that fits in the car.

This also explains why Hesai put effort into redesigning first. Early lidars required many discrete components, and Hesai later integrated some functions into chips. When making the ATX, it further simplified the optical structure. The lidar became smaller, making it easier for automakers to find a place to install it, and saving labor in final assembly and calibration at the factory. Li Yifan puts the starting point of cost reduction in design. The more complex the structure, the more work there is to do during assembly. Only by changing the design first can more processes later be handed over to machines.

Hesai ATX. Image source: Hesai Technology.

When making lidar for the next automaker, Hesai can still reuse existing chips and designs. Supplying several models at the same time spreads R&D investment across more lidar units.

Li Auto changed its configuration, Xiaomi made it standard across the lineup

In May 2025, Li Auto updated the L series, and AD Pro also got lidar. Li Auto continued selling driver assistance in two tiers, and both tiers used Hesai's ATL. Automakers can differentiate high and low trims with software and compute, while procuring the lidar uniformly. For Hesai, the ATL was built on the basis of the ATX. Hesai used its existing design to expand supply from high trims to base versions.

Li Auto added lidar to the base versions of its existing lineup. Xiaomi, meanwhile, equipped all versions with Hesai lidar from the launch of the YU7. Of the nearly 130,000 YU7 units Xiaomi delivered in the second half of the year, every one had Hesai lidar. In 2025, Li Auto delivered fewer cars than the year before, yet Hesai still installed more lidar domestically. One new Xiaomi model alone gave it a big order from outside Li Auto.

Xiaomi YU7. Image source: Xiaomi Auto.

According to Li Yifan's recollection, the Li Auto project lost money in its early days. Later the product had gross margin, but even including chip R&D investment, the project might still not have made money. Hesai spent money first on chips and production lines. Only by applying these results to later models did the money spent on the first order have a chance to be recovered.

By 2025, Hesai's shipments exceeded the expectations given in the interview, and the company achieved full-year profitability for the first time. In the annual report, lower-priced automotive lidar sold more, and across all lidar, the average selling price fell. Lidar revenue continued to grow, and the company's operating profit turned positive. Sales volume grew faster than unit prices fell, so Hesai could achieve profitability while shipping large volumes of low-priced lidar.

RoboSense's new products only entered mass production in Q4

Once an automaker settles on a lidar model, the parts supplier still has to go through trial installation and validation, then prepare for mass production. Hesai began delivering the ATX at the start of the year, and its Hangzhou plant Phase II went into operation in August. RoboSense was also changing generations that year; according to its annual report, what it mainly delivered in the first three quarters was still older products, and new products EMX and EM4 only entered mass production in Q4. By the time these two new RoboSense products began mass production, several new cars that had chosen Hesai were already being delivered.

When the Zeekr 001 was facelifted in autumn, its main lidar switched from RoboSense to Hesai. In the same year, RoboSense's new lidar was also installed on the Zeekr 9X. The same brand can choose different lidar for different models. When a facelift comes, the original orders are also redistributed. Starting with this facelift, Hesai supplies the 001.

RoboSense also had changes in existing orders. In its interim report, it said two major customers reduced purchases: one switched to products from a partner component manufacturer, and the other canceled lidar configurations on some models. In 2024, mass-production cars from several automakers pushed RoboSense to first place. When automakers change configurations or switch suppliers, RoboSense has to fill the gap with other models. During this period, RoboSense was also preparing for mass production of its new products.

… replies

G
Ctrl + Enter to reply
Work from Creator Alliance, reviewed and approved.