Lidar Prices Fell 99.5%, and It Wasn't Volume — It Was Design
Hardware cost reduction doesn't come from scale, it comes from design capability: shrinking a lidar that cost hundreds of thousands of yuan into a chip is what lets it go from a feature to a safety component, and finally to standard equipment.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
The 99.5% price cut came from design, not volume
When Hesai first shipped in 2017, a lidar cost 200,000 to 300,000 RMB — as much as a car. Six or seven years later it was down to 200 USD, a reduction of roughly 99.5%. Li Yifan's judgment: most of the hardware cost reduction you see today does not come from volume, it comes from design — you get better at integrating things, and only after chipification can you achieve real cost reduction. He draws an analogy to Tesla: Tesla's biggest impact on the auto industry was not better automated production, but proposing a more advanced automotive design philosophy, cutting cost at the design level. So from 2017 Hesai realized it couldn't just be an assembly shop — it had to become a chip company too. Today it has mass-produced its fourth-generation chip.
— Li YifanThree founders split the equity evenly, and every investor objected
Hesai's three founders hold roughly equal stakes; Li Yifan is CEO but does not hold the largest share. He says every investor objected, on the grounds that "three people shouldn't split equity evenly." His response: listen to everyone's advice, but don't follow everyone's advice — "listening and following are two different things." He argues further that once you need someone to make the call, it's essentially "a very low-level decision-making mechanism": the vast majority of disagreements come from different backgrounds and different values, and equity can't solve those — even if the largest shareholder holds more than double the second-largest, it still can't. The more common source of disagreement is inconsistent information, which has nothing to do with voting rights and can only be solved by designing a system and communicating deliberately. The three of them share one office.
— Li YifanIn year one they signed a 20 million RMB natural gas order
In their first year back in China they were using lasers to detect natural gas leaks, and a customer asked whether the laser could be mounted on a drone for fully automated inspection. Hesai actually turned the laser into a fully automated natural gas leak detection solution, delivered with an iPad and a software system; the customer was one of China's largest natural gas companies, and the order was signed for 20 million RMB. Li Yifan says signing a 20 million order in the first few months of a startup, at very high gross margin, felt great — but they soon realized this DNA wasn't suited to such a traditional business, and that it was only a direction the customer had suggested; nobody ever really turned it into a mass product. "It just stayed in that museum." The next year they did air quality monitoring, a few million in revenue, a team of about thirty, and by early 2016 they could already see the end of the road.
— Li YifanThey deliberately made a 64-line unit into 40 lines to raise switching costs
The first product was called Pandora 40. At the time Velodyne was at 32 and 64 lines, and the most sensible move would have been to build a drop-in replacement — but Hesai concluded after analysis that 40 lines was more reasonable, and deep down they wanted to build something different from the other guy: "I was afraid the reason you'd use mine was that you thought I was just a cheaper substitute." Li Yifan admits this carried a certain "literati arrogance": if a customer wanted to use it, they'd have to abandon all their existing data, and switching costs would be higher rather than lower — business school people would think you'd gone mad. His reasoning was that switching costs actually weren't high, and that this was the only way to filter for people who genuinely believed in the product. As for the name, the most Chinese elements are the five-star red flag and the panda, so it was called Panda — add an R and you get Pandar.
— Li YifanThe customer wouldn't let us publicize it, because we were the only usable option
Silicon Valley self-driving company Drive.AI placed a 2 million USD order and demanded to pay the full amount upfront in one transfer. Li Yifan asked whether they could do a case study, and the other side refused, citing an NDA. Two weeks later, face to face in Silicon Valley, they gave the real reason: you're the only usable one, and I don't want anyone else to know — because nobody trusts Chinese brands, I took the risk and found out you were usable, so why would I recommend you to my competitors? Li Yifan says that in that era he thought Velodyne wasn't good enough and wanted to find something better, found Hesai, and so Hesai was his secret weapon. Around the same period there were also customers like Aurora and Nuro; 2017 revenue was about 30 million RMB.
— Li YifanThere are only technology choices, not technology routes
Li Yifan says that in over a decade Hesai has never used the phrase "technology route." He draws a distinction: a technology route means that once a company goes in one direction it can never come back, going further and further, one right and one wrong; a technology choice means what you're doing is something others can do too, and it's just a question of who does it better. He gives an example: so-called solid-state and semi-solid-state are essentially just changes in the mechanical scanning method — they sound cool and make later players feel there's an opportunity to overtake on the curve. He offers an analogy: people making four-cylinder engines don't talk about technology routes, people making five-cylinder engines do, because that's an unreasonable design — once you've worked on it for a while and find the five-cylinder is a struggle, you go back to four cylinders and stop talking about it too. He acknowledges that the two choices — 90 nanometers, and multiple lasers plus a simple mechanical scanning structure — were decided under Sun Kai's guidance.
— Li YifanNot using lidar is mostly a fear that Musk is right
On the Tesla approach, Li Yifan lays out a few facts: lidar is a sensing method based on a different principle, and it will definitely make cars safer — it's not plan B, it's plan A plus: while A is executing, A plus is always working. He uses a blind person as an analogy: if you can both touch and see, and in a dark cave your eyes see nothing but your hand touches something, who do you believe? He thinks many of the people who are genuinely worried are worried that Musk is right, rather than looking at the data and concluding it isn't needed. He also takes apart the term "first principles": what Musk is actually talking about is biomimicry — the logic being that humans drive using only their eyes, so machines driving shouldn't need anything beyond that principle — whereas first principles in essence means analyzing the most atomic nature of things without relying on statistical patterns. The two are opposites. Airplanes are in fact the least biomimetic thing; Da Vinci's attempt to fly the way a bird does was wrong.
— Li YifanChinese companies going global need to replace the grab-the-money logic
Li Yifan says his biggest recent realization is that in new industries like cars and the automotive supply chain, Chinese people are far ahead of most Western competitors — but "when you are clearly far ahead, how do you find a good way to get along with the world" is something they haven't thought through enough. He reflects that historically international business was a "grab-the-money logic": my thing is better than your local one, more reliable, sometimes cheaper, so please use mine. That logic only works when you're very small; once you reach scale, in a mature industry what matters is — if you do business with me, what can I do for the local area, so that companies, government and the public all benefit. He mentions that Hesai has historically faced not only the International Trade Court and patent litigation, but also issues with the Department of Defense.
— Li YifanHard tech has no network effects — it's a marathon
Li Yifan compares two generations of entrepreneurs: the internet is an industry centered on winner-take-all, where every battle is a decisive battle and you have to spend all your ammunition, because once the gap with the number two widens it never narrows again; this generation of hard tech has no network effects, and no one has emerged who, like Baidu, Didi or Uber, becomes first and then is very hard to dislodge — it's closer to economies of scale. He thinks network effects were "a fairy tale of one era," and that entrepreneurs of that era had an extremely exaggerated winner-take-all opportunity. So the bigger problem for hard-tech founders is realizing this is a marathon — you can't use up all your energy at the start. He quotes Zeng Ming's saying, "look at ten years, think about three years, do one year," and says that in the early days he didn't understand what investors meant when they asked "how do you build a moat."
— Li YifanIn their own words · checked verbatim
Because you're the only usable one, and I don't want anyone else to know — this is my competitive advantage.
因为你们是 唯一能用的 而这件事 我不想让别人知道 这是我的竞争优势
Li Yifan1:26:58
In over a decade at Hesai, I have never said those four words. Why? Because I don't think there's any such thing as a technology route.
和赛十几年 我从来没有说过这四个字 为什么 因为我觉得 没什么技术路线
Li Yifan1:41:59
That day I randomly asked a colleague, I said, tell me, what is first principles? Everyone's reaction was, we've actually never heard of it, we've only heard Musk say it.
我那天随机问一下同事 我说 你们跟我说一下 啥是第一性原理 所有人的反应都是 我们其实没听过 只听马斯克说过
Li Yifan1:53:03
I said if back then we were seeking statistics, we shouldn't have started a T1, because the historical success rate of T1s is far too low.
我说如果当年我们去寻求统计 我们就不应该创一家T1 因为T1历史上成功率失败的太低了
Li Yifan2:35:31
When I thought of America, I didn't think only of Apple — I thought of MBA, I thought of fun things.
我想到美国的时候 我没觉得仅仅是苹果 我想到的是MBA 我想到的是好玩物
Li Yifan2:56:50
Think carefully about where the industry's opportunities come from — they're actually the nation's opportunities, the people's opportunities.
你仔细去想行业的机会来自于哪儿 其实是国家的民族的机会
Li Yifan3:00:52
Figures
| Lidar unit price reduction | From 200,000-300,000 RMB per unit to 200 USD, roughly a 99.5% cost reduction | 5:02 |
| Hesai cumulative funding | Close to 1 billion USD | 2:01 |
| Hesai founding date | 2014 | 2:01 |
| Funding threshold for returning to China to start up | 2 million USD | 44:30 |
| Drive.AI order amount | 2 million USD, paid in full in one transfer | 1:23:58 |
| Series A funding | 110 million RMB, at the end of 2016 before 2017 shipments | 1:27:58 |
Glossary
- ASIC
- Application Specific IC — an integrated chip system redesigned for a specific application, capable of being mass-replicated.
- ITC
- The US International Trade Commission, which hears patent and trade disputes; it is fast, and losing means a ban on sales in the US.
- POC
- Proof of Concept — a customer first gives you a small project to verify your delivery capability before discussing a large order.
- Robotaxi
- L4/L5 autonomous driving ride-hailing service; Hesai's main source of early customers.
- Tier 1
- A supplier that sells directly to automakers; Li Yifan uses it to describe Hesai's positioning.
How to listen
Hard-tech founders, automotive supply chain professionals, investors and engineers watching Chinese manufacturing go global.
12:05 to 32:13, the family upbringing and Silicon Valley work history — low information density, skippable.