He Saite Technology, Li Yifan: Lidar should be a commodity, not a loss leader for each unit sold | 36Kr exclusive interview

Title: The Battle Between Pure Vision and LiDAR Continues

After the emergence of pure visual solutions in companies like Huawei's intelligent driving, the LiDAR companies, which have been pushed to the opposite side, have had to endure repeated interrogations.

"Lidar is like a 'cheating weapon.' It can improve the assisted driving capability by 2 years, and the performance in many scenarios is improved by more than 10 times. Why not use it?" Li Yifan, CEO of Hesai Technology, responded at a recent product launch.

In fact, the war between pure vision and LiDAR has never stopped. Led by Tesla, domestic intelligent driving companies have adopted visual perception solutions such as BEV (bird's-eye view) + OCC (occupancy network), which can also generate three-dimensional space and perceive common obstacles on the road.

LiDAR is seen as another redundant perception solution. Li Yifan has also emphasized multiple times that LiDAR is like a safety airbag, solving the 1% safety challenges that pure visual solutions find difficult to handle.

However, against the backdrop of the fierce price war in the automotive industry, the LiDAR, which costs thousands of yuan per unit, has made mainstream car companies hesitant.

In response, Hesai has also launched a cost-effective LiDAR solution, continuing to compete with pure vision and prove its own value.

On April 19, Hesai released a wide-angle long-range LiDAR, ATX, which is an important entry into the 150,000-200,000 yuan market. Hesai stated that ATX has already obtained multiple mass production orders and will be mass-produced in the first quarter of next year.

Compared to the previous generation product AT128, the ATX has reduced its volume by 60% and weight by 50%, and can be integrated into positions such as the roof, behind the windshield, and inside the headlights.

In terms of performance, the ATX has improved in detection distance, resolution, and horizontal field of view. According to Hesai, the ATX can detect up to 300 meters, support up to 256 lines, and achieve the best angular resolution of 0.08° x 0.1°, enabling faster recognition of vehicles and pedestrians during high-speed driving. At the same time, the horizontal field of view reaches 140°, capable of handling situations such as vehicles cutting in, pedestrians, or non-motorized vehicles crossing the road.

In addition to higher hardware integration, the Hesai ATX also incorporates a self-developed intelligent point cloud analysis engine, which can identify rain, fog, vehicle exhaust, dust, etc., reducing the interference of rain and fog on the LiDAR.

With the release of AXT, Hesai has also reorganized its product strategy. The previously launched high-performance LiDAR AT512 is aimed at L3-level autonomous driving applications, while AXT is targeted at L2+ intelligent driving, with more competitive pricing.

Li Yifan has repeatedly emphasized that the price of AXT will not be publicly disclosed. However, he also stated, "In the automotive field, the basic principle is to win the lowest bid. If everyone has a similar number of projects, the product prices are likely to be similar." Previously, SureStar Juchuang also launched a mid-to-long-range LiDAR priced at less than $200, similarly targeting the cost-effective market.

This is also a counterattack by LiDAR companies under the questioning of their route and price pressure.

However, igniting a price war may place an even heavier burden on LiDAR companies, whose already low profit margins may be further affected. According to Hesai's statistics, by 2023, ADAS LiDAR companies generally have negative gross margins, and it is only with the expensive Robotaxi LiDAR products that Hesai's overall gross margin reaches 35%.

At the same time, companies such as Geely E-Kua Tong and DJI Automotive have not truly given up on LiDAR products, and the degree of internal competition is still escalating.

However, Li Yifan believes that the mass production capacity of LiDAR companies, selling tens of thousands of units per month, will also form a barrier, including in terms of capital, cost, and time period, which will keep some new players out, and the industry will ultimately become more and more consolidated.

The dispute over intelligent driving routes is not yet settled, and Hesai still needs to continuously prove the value of LiDAR. Once the necessity of the product's installation is clear, making money will only be a matter of time.

The following is a conversation between 36Kr Automotive and other media and Li Yifan, CEO of Hesai Technology, as edited:

Question: Is OCC (occupancy network) in pure vision in competition with LiDAR?

Li Yifan: LiDAR solves 1% of the problems. Everyone is talking about how OCC can achieve 99.9%, but when it comes to safety, LiDAR still clearly solves some scenarios better, and in some cases, only LiDAR can solve them.

For L2++ systems, everyone still has a choice; but for L3-level autonomous driving, there must be safety redundancy, and there must be two different principle intelligent driving systems.

Question: In what stage do you think LiDAR is currently at? Is it in the elimination round, or the quarterfinals or semifinals? The industry had already been in a state of consolidation, and now there are players such as DJI, Huawei, E-Kua Tong, and the industry is becoming more diversified again?

Li Yifan: It's hard to judge. In China, if LiDAR hasn't reached a monthly production of tens of thousands, there are definitely many things that are not known, and we have also come step by step. Today, our understanding and awe of the industry are definitely different from before we achieved large-scale production. Companies that have completed the closed-loop production of LiDAR, globally, are no more than a handful.

The industry will indeed become more consolidated. There are several reasons: unfortunately, during the rapid decline in the selling price of LiDAR, the cost has not yet been reduced enough. ADAS LiDAR companies generally have negative gross margins, which is very strange, but that's how it is. So for an industry where the vast majority are operating at a loss, and you have to invest a lot of money, do you still want to do it?

There are also barriers such as capital, time period, and cost, which have formed some barriers. Especially, LiDAR should be a commodity, and it shouldn't be the case that every unit sold results in a loss. Many players in the industry are facing this problem. In this situation, will it become more and more open? This is unreasonable.

I believe the industry will be divided into three stages: samples, products, and commodities. Samples are the coolest concept, with new technological routes popping up every now and then, but mainly for VC. Products are what are delivered to customers, such as Robotaxi. The transition from products to commodities involves large-scale delivery. We are doing well today, but the "battle" is far from over.

Question: What is the appropriate level for the gross margin of ADAS lidar?

Li Yifan: If there is only one or two companies in the final competitive landscape, occupying 70%-80% of the market, the gross margin will be higher. But if the market becomes more open with 10 companies, the possibility is very low. However, I think the lidar industry has barriers and will not become like the 3% or 5% gross margin for components in many vehicles.

Question: What is your view on the 150,000-200,000 RMB car market? Many car companies in this range have not yet used lidar. Will car companies bite the bullet and install it?

Li Yifan: Our target market is above 150,000 RMB, roughly 6 million vehicles. If we can't hold the line at 150,000 RMB, then the market is too small. The mission of each company is the same.

For lidar, if car companies are reluctant to use it because they don't think it can help make money and sell cars, then they shouldn't use it. More people want to use it, and after it is installed, it becomes the biggest selling point for car companies. This is the opportunity I see. There is still a big gap between the experience of ADAS today and what can be achieved in the future. In this process, the value of lidar is constantly increasing.

Question: Hesai is entering the intelligent manufacturing stage. How do you balance the yield issue during the rapid iteration of products?

Li Yifan: If the company iterates too quickly, it will be a challenge for various aspects such as R&D, manufacturing, automated production lines, quality control, and yield. The way to alleviate this is, first, to make the platform as mature as possible. If the platform itself is mature, the problems encountered by the products will be relatively fewer. Second, we invest in building our own production lines. Only when R&D and manufacturing are together can all the problems be solved at an earlier stage.

Question: Will the arrival time of FMCW (emerging lidar route) technology be delayed?

Li Yifan: We are also developing FMCW technology, although there is no product line for the public. When it comes to making money and rapidly iterating products, I will definitely choose to iterate products quickly. To talk about commodities, it is a prerequisite to maintain the most competitive products.

FMCW products depend on suppliers, and whether the underlying technology is mature determines whether the product is mature. When the underlying technology is not mature, it's a bit like making electric cars when battery technology was not mature back then. It's just empty talk without meaning.