Li Yifan’s 11 Years Building Hesai—and Where Opportunity Comes From
Summary
Hesai’s core financial story is not that its stock doubled, but that it drove LiDAR from RMB200,000–300,000 per unit in 2017 to roughly $200, while targeting profitability in the “this year” referred to in the interview. Li Yifan said the company has raised nearly $1B cumulatively, delivered more than 600,000 units last year and expects 1.2M–1.5M this year; the stock can diverge from value in the short term, “but over the long term they will converge.” What matters is the product, customer value and ability to deliver consistently.
The 99.5% cost reduction came mainly not from scale, but from ASICs, internally developed core components, product redesign and automated manufacturing working together to rewrite the cost structure. From its first-generation product in 2017, Hesai understood that a discrete-component product costing hundreds of thousands of yuan could never reach the $1,000 range through volume alone, so it transformed itself from an “assembly plant” into a chip company and now mass-produces fourth-generation chips. Li Yifan’s hardware doctrine is simple: “Most of the cost reduction in the hardware we see today doesn’t come from volume. It comes from design.”
LiDAR is shifting from an expensive Robotaxi “feature” to a “safety component” in mass-market vehicles, changing market capacity, price elasticity and configuration logic. Hesai’s early Robotaxi products carried gross margins of 70%–80%; later, its first mass-production program upgraded from 64 lines to 128 at Li Auto’s request without changing size, cost or quoted price. It began shipping with the L9 in July 2022 and exceeded 10,000 units of monthly production in September. Consumers can reject an assisted-driving feature, Li Yifan argues, but few would voluntarily remove airbags, seat belts or brakes: “A safety component has nothing to do with features. It has value for everyone.”
Li Yifan sees the supposed “technology-route debate” as a capital-markets narrative; the real competition begins after choosing technologies such as 905 nm and a scanning architecture, and is about executing better on performance, quality, cost and manufacturing. He compares it with four-cylinder versus five-cylinder engines: an irrational unconventional design can be packaged as an attempt to overtake on the inside, but the industry ultimately converges, while companies using the same approach can still differ enormously in execution. Pure vision and LiDAR are not a simple either-or; LiDAR is an always-on “A+” that provides independent evidence, like a hand touching an obstacle when vision fails to detect it.
Hesai’s most important early organizational assets were commercial feedback, equal partnership among its 3 founders and conservative cash management. The founders held nearly equal stakes, despite universal investor opposition; they divided responsibilities across business, frontier technology, engineering and manufacturing, and reduced friction through a shared office and synchronized context. Even when the company pursued businesses such as gas detection and PM2.5 monitoring that later proved unattractive, it insisted on real revenue because “a company without revenue is an original sin.” Its first major customer, Drive.ai, prepaid $2M and treated Hesai as a “secret weapon” it did not want competitors to know about.
The next phase of globalization cannot stop at proving that Chinese products are better and cheaper; it must answer what shared benefits Hesai can create for local customers, governments and citizens. Hesai spent years on POCs before entering a top European automaker’s global mass-production system, and built local teams attuned to America’s directness and Germany’s indirectness. Li Yifan calls the domestic market a “gym for building muscle,” while overseas markets are where the company makes money from those capabilities. He now sees the flaw in “my product is better, so you have to buy it”: as a company grows, it must contribute jobs, capabilities and local value rather than simply “come here to extract money.”
Li Yifan sees hard tech as a marathon without network effects, and ultimately explains the industry’s opportunity through the nation and its people. Companies cannot burn through their ammunition using the internet playbook that treats every battle as decisive; they must “look 10 years out, think 3 years out and execute 1 year at a time,” converting each win into capabilities in chips, manufacturing, quality and global service. He ultimately extends the mission from exporting engineering capability to exporting brand and culture: “Where do the industry’s opportunities come from? They come from the nation and its people,” with the end state being for the world to know not only Chinese manufacturing, but “how cool Chinese people are.”
Deep dive
1. The Stock Is a Voting Machine in the Short Term; Products and Customer Value Are the Scale in the Long Term
The quick-fire company profile: Hesai Technology was founded in 2014, went through Series A, B and C rounds and an IPO, and has raised nearly $1B cumulatively. Li Yifan was born in 1986, identifies as ENTJ, and expects the company to turn profitable in the “this year” referred to in the interview.
Faced with the stock’s roller coaster, Li Yifan tells the team not to watch the price minute by minute. The company cannot genuinely multiply its value several times over in the short term, nor lose its long-term value because of a period of low prices: “Short-term stock-price movements can diverge from a company’s long-term value, but over the long term they will converge.”
His management message is the same at highs and lows: gains reflect macro conditions, sentiment and trading dynamics, while declines do not mean the product has lost its value. The variables a listed company can actually control are whether it builds good products and technology and creates more value for customers.
2. LiDAR Is an “Eye” That Emits Light and Measures Distance Directly
Li Yifan calls LiDAR the “eyes” of cars and robots. It identifies pedestrians, vehicles and road obstacles, but unlike a camera, which passively receives light, it emits light and receives the reflection, giving it a different source of information in darkness and complex ambient light.
The more important difference is distance. Cameras generally have to infer depth; LiDAR measures it directly using time of flight—“a bit like a ruler that accurately measures distance.” For a fast-moving car or robot, the distance to an obstacle can be an input that “determines life or death.”
Around 2008–2016, the industry was still dominated mainly by US companies, whose products were expensive and not necessarily high quality. From 2017–2018, Chinese players entered with new architectures, manufacturing methods and sustained R&D, changing performance, quality and cost simultaneously.
3. A 99.5% Cost Reduction in 6–7 Years Turned a Car-Sized Price into a Component
When Hesai first shipped in 2017, a LiDAR sold for RMB200,000–300,000 in both China and overseas—“about the price of a car.” Six or 7 years later, the product was down to roughly $200, a cumulative cost reduction of 99.5% by Li Yifan’s calculation.
He acknowledges that the initial price was excessively high and irrational, and that autonomous driving’s expansion into assisted driving created scale. But he still stresses that most industrial products could never achieve this magnitude of change through volume alone: “If cars could fall 99.5% through innovation, how happy would everyone be?”
Heavy R&D spending also pushed the industry faster than its natural pace. Hesai spends roughly RMB1B a year on R&D, with peers investing heavily as well. Li Yifan says the sector may have been “overheated” for a period, but the result was that products that might otherwise have stopped at $1,000 kept falling, creating a cost dividend for consumers and automakers.
4. Hardware Cost Reduction Starts with Design; Scale Only Amplifies Good Design
Early products were manually “assembled” from discrete components. They were large, expensive and poorly suited to automated production. Hesai realized that automation did not mean putting a hand-built product directly on a production line; the components, structure and processes had to be redesigned first.
Li Yifan’s “chipification” is not mysterious. It is essentially an ASIC, or Application-Specific Integrated Circuit: reintegrating a system for a specific application so it can be replicated like printing, while shrinking its size, improving performance and lowering cost.
When the first-generation LiDAR was completed in 2017, the team asked: how do we turn a product costing hundreds of thousands of yuan into one costing $1,000? The answer was not to wait for sales, because there was not enough “water” in the cost base. It was to bring dispersed core modules that had been purchased externally into internally developed chips, generation by generation.
Hesai now mass-produces fourth-generation chips, and most core components have been chipified. Li Yifan sums it up: “Most of the cost reduction in the hardware we see today doesn’t come from volume. It comes from design.” Automation and scale come later to improve quality.
5. The Engineer’s Instinct Began with Coding at 5 and Physics Competitions
Li Yifan’s father studied automated control of industrial enterprises, whom he jokingly calls “China’s first-generation AI engineer”; his mother was a medical-school professor. The family already had a computer in 1991, and his father believed the next generation would struggle to survive without programming skills, so Li Yifan says he was “writing code when he was still running around naked.”
He developed a long-standing interest in mathematics, physics, programming and opto-mechatronics, and gained admission to Tsinghua’s precision-instrumentation program through physics and informatics Olympiads. In high school, he acknowledged that some classmates had better grades, but he was still class monitor and a first-class scholarship recipient; at university, he became student-union president.
6. His PhD Studied Control; Only at Graduation Did He See That Perception Was the Real Bottleneck
During his robotics PhD at UIUC, Li Yifan studied how robots could walk without falling, grasp objects and determine how far they had fallen. “Artificial intelligence” was not yet a common term; this work was called advanced control.
He later changed how he judged the value of the research. When a robotic arm fails to grasp a cup, the root cause is often not insufficiently precise control, but failure to recognize the cup, measure its position correctly or identify the outer edge of a double-walled glass. “It took me many years to realize that I had studied a problem that was not the most important one in the industry.”
This reflection was not a rejection of control research. It was an acknowledgment that he had continued “carving flowers” at the frontier while being constrained by robots’ limited understanding of the 3D world. Even today, many failures in automotive and robotic systems begin with “not perceiving clearly enough.”
7. A 1% Optimization Can Be Extremely Valuable Without Exciting an Entrepreneur
After graduation, Li Yifan became a principal engineer at Western Digital, optimizing hard-drive head control. The precision was comparable to having a plane flying at full speed pick up every golf ball along a runway. A 1% improvement in track-seeking time, multiplied across millions of hard drives and enormous numbers of operations, still had substantial commercial value.
He uses YouTube’s recommendation algorithm as a similar example: increasing the time global users spend on the platform by 1% can generate enormous advertising revenue. Li Yifan recognizes the value of Silicon Valley experts’ “carving flowers,” but found that his own excitement lay not in optimizing a huge installed base by 1%, but in building a new physical product.
8. Western Workplace Experience Became His Ability to Communicate with Overseas Customers
His year-plus at Western Digital did not produce a new academic theory, but it taught Li Yifan the operating rules of a technology company: how bosses manage engineers, how colleagues communicate, what behavior is acceptable and which boundaries cannot be crossed.
The longer-term payoff was learning how Western companies make decisions. Hesai later served large European and US automakers, dealing with organizational cultures rather than individual buyers. Without firsthand experience, aggressively selling a product on its merits alone could easily miss the point.
He observed that Western customers often want suppliers to explain their weaknesses first and dislike having their time wasted by circumlocution. Eastern communication is more attentive to relationships and conflict avoidance. Both cultures have to be understood; the same sales language cannot be mechanically copied.
9. He Abandoned the Scientist’s Path Because He Wanted Technology to Be Used in the Real World
During his PhD, Li Yifan felt control had entered a phase of correction and refinement, making foundational breakthroughs less likely. He also acknowledged that not every PhD becomes a scientist; his own interests leaned toward putting technology in front of more users.
He applied to Boston Dynamics. The company said much of its work was related to defense and that it did not often hire non-US citizens, especially Chinese nationals. There was no Unitree and no second comparable company at the time, so the rejection disappointed him. In retrospect, general-purpose robotics in 2014 still lacked a clear commercial path, and joining would probably have meant remaining a researcher.
Zhang Xiaojun asked why he did not pursue what Unitree is doing today. Li Yifan’s answer was timing: he preferred an industry in which a finished product could immediately create productive and economic value, rather than one dependent on long-term financing before commercialization.
10. Commercialization Is Not Money; It Is the Hardest Market Feedback to Fake
Li Yifan defines commercialization as “the feedback you can trust most.” Financing, valuation and fame can all be indirect signals; customers will not keep paying real money for a bad product. Industries without revenue are more prone to internal arguments and strategic fragmentation because feedback is distorted.
He does not deny the value of long-cycle projects. If the dream, entrepreneur and investors all have enough patience, it can make sense to invest in directions that are not commercial today but could be extremely valuable later. Hesai’s preference, however, is to have clear commercialization paths in the short, medium and long term.
He also believes that the many splits among AI companies over the past decade were partly related to long-term commercialization failing to materialize. Once a company is far from its customers, it becomes difficult to objectively determine whether the boss, investors or technical staff are right.
11. Three Founders Held Nearly Equal Stakes, Putting Complementary Capabilities Above Single-Person Control
Li Yifan, his senior schoolmate Xiang Yaoxing and Silicon Valley acquaintance Sun Kai had been friends since around 2009. They shared an interest in technology, products and entrepreneurship, and all believed they would eventually return to China to build something rather than stay in the US for a more comfortable life.
Their long-term division of labor was clear: Li Yifan handled business, customers, operations and capital; chief scientist Sun Kai handled frontier strategy and “0 to 1”; Xiang Yaoxing, an exceptionally strong engineer who had worked at Apple, handled “1 to 10,” manufacturing and quality systems.
The equity split was nearly one-third each. Sun Kai initially held an extra percentage point because the first laser-detection technology came from his laboratory—what Li Yifan calls a “rounding error.” Almost every investor opposed equal ownership, but the team accepted that some would decline to invest as a result.
Li Yifan does not believe a single founder holding absolute control is a universal answer. Every company needs to stick to its own organizational assumptions; Hesai’s results at least show that this model can work.
12. Disagreements Come from Values or Context, Not from the Absence of a Final Decision-Maker
Zhang Xiaojun repeatedly asked who made the final call when the founders disagreed. Li Yifan believes this often reduces the question to the wrong level: a values conflict does not disappear because a major shareholder forces a vote, and when information is inconsistent, even a boss can make the wrong call by not understanding the technology.
The 3 founders share similar views on fundamental right and wrong. One will not want to serve customers while another wants to exploit them. When their views differ, they first assume their contexts differ and resolve the issue by sharing information rather than escalating it into a power struggle.
They shared an office and could hear one another talking and typing. The arrangement was inspired by the open executive offices at a large European automaker: informal encounters, overheard conversations and casual chats outside formal reporting were all part of the top-level information-synchronization system.
13. Returning to China and Starting a Company Were One Decision in 2014; $2M Was the Common Starting Line
The 3 never seriously considered starting a company in the US. They saw America as more comfortable, but not well suited to building a complex hardware-manufacturing company. Returning to China and founding a company were not treated as separate decisions.
The team agreed to resign and return together once they raised $2M. Li Yifan initially tried to raise $1M for 20% and found no takers. He then approached 2 institutions separately, asking each to invest $1M for 10%, while telling each that the company would be formed if the other participated.
The 2 early institutions were Yuanzhan Capital and Shenzhen-based Dami Capital. Li Yifan denies this was deceptive: once both agreed, he immediately disclosed the full structure. Financing can be designed, but the facts a company presents externally must be accurate.
14. The First Gas-Inspection Order Was Worth RMB20M—and Exposed a Market Mismatch
The team’s first direction after returning was laser-based gas-leak detection. Certification in the traditional energy industry was complicated and sales cycles were long. Customers then asked for the sensor to be mounted on a drone to enable fully automated pipeline inspection—a normal company might simply say it could not be done.
After speaking with DJI, Hesai assembled the laser, drone, iPad and software into a complete solution and sold it to Xiyang Gas. Within months of founding, it signed an order worth roughly RMB20M, delivered in batches, with apparently high gross margins.
The project trained the team to deliver full-stack solutions but did not prove the direction was right. After making frequent trips to Langfang, Li Yifan concluded that the commercial value created by the technology was limited while sales and certification were too burdensome. The product ultimately looked more like an isolated exhibit in a “museum.”
15. PM2.5 Revenue Kept Growing, but Exposed the “Hammer Looking for a Nail”
As air quality deteriorated, the team developed compact PM2.5 monitoring devices, worked with environmental authorities and monitoring stations, mounted the devices on buses to collect city data, and built modeling and app systems.
These businesses generated several million yuan in revenue, and the company continued to grow every year from its founding. But a team of just over 10 people was exhausted by delivery, while the market ceiling was too low. Li Yifan says that several million yuan in revenue “made everyone that tired,” which was itself a bad signal.
Zhang Xiaojun summarized this as a common problem for hard-tech founders: “holding a hammer and looking for nails.” They understand the boundaries of their technology but not the market’s pain points; a little customer interest can lead the team to commit heavy assets and mistake technological appeal for commercial value.
16. Slow Negative Feedback Is How Technical Entrepreneurship Turns into “Intracranial Self-Amusement”
Li Yifan believes Eastern culture typically avoids saying “bad idea” directly. Customers, friends and investors are more willing to encourage young teams. That friendliness reduces emotional harm, but also makes it harder for entrepreneurs to confirm quickly that a project has no real demand.
An early investor told them, “We are investing in you. Whatever you want to do, we support.” Li Yifan remains grateful for that trust. But if entrepreneurs have to wait for investors to tell them what to do, that is also unhealthy; responsibility for direction cannot ultimately be outsourced.
The real danger is a team absorbed in technology without measuring commercial value or knowing how to obtain market feedback, creating “intracranial self-amusement.” The gas and air-monitoring projects soon revealed their limits, forcing Hesai to search for a new direction.
17. The 2016 Blackboard Exercise Turned Entrepreneurship from Technical Interest into Market Selection
In early 2016, Li Yifan filled a blackboard several meters long with industries that lasers might enter: industrial cutting, food-ingredient and calorie analysis, various sensors, and even shining light on plates to measure how many calories someone had consumed.
The screening criteria approached a complete entrepreneurial framework for the first time: the market had to be large enough, differentiation clear enough, opportunities abundant enough and relevant to the team’s capabilities. They began formally discussing demand, market size, product uniqueness and unfair advantage.
LiDAR stood out not because the sensor market itself was large, but because it enabled robots and cars. The ceiling for cutting equipment depends on how many people need things cut; the ceiling for 3D perception depends on the development of the entire robotics industry.
18. A Berlin Automaker Executive’s “Eyes Lighting Up” Turned a Difficult Problem into a Definite Direction
In April 2016, Li Yifan was invited to Berlin to speak on “Why Chinese Innovation Is Powerful.” He joked that the company was not yet successful, but used the trip to visit Tsinghua alumni and automotive professionals in Munich, Stuttgart and elsewhere.
Whenever people knew LiDAR, they would light up at the news that Hesai wanted to build it and repeatedly say, “Please do it. Hurry up.” Mass-market vehicles did not yet use LiDAR widely; only advanced R&D teams could afford equipment costing tens of thousands of dollars. But industry professionals were convinced that demand would eventually emerge.
Li Yifan returned from Germany and went straight to the company to ask about progress. The answer—“we built it”—meant a beam of laser on a lab bench, one MATLAB number read roughly every 20 seconds, and the apparatus rotated on a small wheeled refrigerator before eventually drawing a square room.
19. Four Months Took the Product from a Refrigerator Turntable to a Formal Point Cloud; Speed Was the Startup’s First Advantage
The primitive setup proved the underlying feasibility. By August 2016, the point cloud Hesai released already looked close to modern LiDAR. Li Yifan believes the prototype principle was not necessarily difficult; engineering it was, and the team led by Xiang Yaoxing “was simply fast.”
He sees speed as a basic qualification for an early-stage startup: “If you say your advantage isn’t speed, I find that pretty strange.” At one point there may have been roughly 50 LiDAR companies in China, but most did not remain on the same track long enough to reach the present.
20. Pandar40 Was Deliberately Not a Drop-In Replacement; Switching Costs Tested Product Value
Velodyne already had 32-line and 64-line products, so the easiest strategy would have been to build a seamless Chinese substitute. Hesai instead chose 40 lines based on actual perception needs and named it Pandar40, combining the Chinese panda with LiDAR.
Li Yifan acknowledges that this raised customers’ switching costs, which a business textbook might call “crazy.” But the team did not want adoption based on being a “cheaper version,” nor did it want customers to think it was copying Velodyne. It wanted to test whether customers would rebuild their data around a more rational product.
This “literary man’s arrogance” was not a rejection of cost, but a product-prioritization choice: pursue absolute performance and structural rationality first, then let the market prove it. Hesai’s first product consequently became an industry-recognized phenomenon.
21. Anti-Interference and Waveform Parsing Pushed Sensor Innovation Further into Chips
In 2017, a US customer envisioned hundreds of autonomous vehicles operating simultaneously, with their lasers potentially interfering with one another. Hesai began adding unique codes to the outgoing light so the receiver could distinguish its own echoes.
Another capability later deployed across the product line was a waveform-analysis engine. Previously the system could only say whether something had been detected; the new system attempted to distinguish rain, fog, dust and material characteristics. When a vehicle splashed water, the LiDAR could determine that the reflection was not another car.
Li Yifan describes the evolution as “compressing the engine and transmission into chips.” Chipification both internalized the supply chain and created a vehicle for performance innovation; miniaturization, higher performance and lower cost accounted for roughly 90% of the product’s progress.
22. Even Low-Quality Revenue Can Unite a Team When Customers Pay Real Money
Li Yifan’s early principle was: “A company without revenue is an original sin.” Gas detection and PM2.5 later proved to be poor directions, but real customer payments made the team organize around the people it served rather than argue over untested judgments from the boss or investors.
He confirms only that revenue was positive at the time and does not claim cash flow was positive. Before finding LiDAR, the company had raised little and the team was small. The real pressure was not imminent starvation, but the fact that more than a year into entrepreneurship they still did not know what they would do in the long term.
Hesai remained financially conservative. It began the next financing round after spending roughly half of the previous round; in hindsight, Li Yifan says it “gave away a lot of shares.” In exchange, it secured supply continuity: automakers need to rely on suppliers for 5–10 years and cannot accept a plan that says supply ends if next year’s financing fails.
23. Drive.ai Prepaid $2M Because It Did Not Want to Share Its “Secret Weapon”
After receiving samples in August 2017, Li Yifan and Xiang Yaoxing took the product from company to company among US autonomous-driving firms. Within months, Drive.ai tested it and proposed an order worth roughly $2M. Li Yifan doubted the seriousness of the offer and demanded payment upfront; the customer then wired the full amount.
He asked to use the switch from Velodyne to Hesai as a customer case study, but the request was rejected under an NDA. When he pressed for the reason, the customer gave the real answer: “You are the only one that works, and I don’t want anyone else to know.”
Drive.ai believed others distrusted Chinese brands, while it had taken the risk of discovering that Hesai worked. That created a competitive advantage, leaving no reason to recommend Hesai to peers. Li Yifan accepted the blunt answer and treated it as the strongest early feedback on product quality.
24. Revenue Grew Roughly 10x from 2017 to 2019; Overseas Customers Recognized Chinese Hardware First
Li Yifan recalls revenue of roughly RMB30M in 2017, more than RMB100M in 2018 and more than RMB300M in 2019. B2B growth required solving capacity, product structure, supply-chain and organizational issues simultaneously; it was not simply a matter of selling more of the same consumer product.
Around 2017–2018, Hesai raised roughly RMB100M in Series A and then closed a Series B involving Baidu and Lightspeed China. Li Yifan speculates that Baidu had already invested in Waymo and understood its products, but judged that Hesai would eventually surpass Waymo.
Early major customers were mainly in the US, and Li Yifan traveled there almost monthly. Domestic autonomous-driving teams existed, but switched later because Chinese customers instinctively believed that expensive, complex equipment should come from established overseas brands.
25. The 1M-Unit Vision for 2019 Went on the Wall When Shipments Were Still in the Thousands
After attending Hupan University, Li Yifan accepted that a company needs a mission, vision and values. Hesai had not lacked values, but its goal was still simply “build a good LiDAR,” not a statement everyone could repeat together.
In June 2019, roughly 30–40 key decision-makers locked themselves in a conference room for 2 days. The task was not for the boss to dictate a slogan, but to summarize how people who best represented Hesai were already behaving and write it down for those who came later.
The team proposed giving 1% of the world’s cars 3D-perception capability by 2025, translating into a business target of roughly 800,000–1M LiDAR units. At the time, annual shipments were only 2,000–3,000 units; the company later disclosed more than 600,000 deliveries last year and a forecast of 1.2M–1.5M this year.
The 4 values remain unchanged: pursue excellence, embrace challenges, let actions speak louder than words and act like an owner. Each month, one of the 3 founders spends an entire day training new employees on the traits that help someone “thrive” at Hesai.
26. Robotaxi Customers Paid for the Best Product; Hesai Was Never a Value-for-Money Substitute
Early Robotaxi customers were still developing systems and collecting data. They did not ask for cheap LiDAR, only the best performance and reliability; like a professional photographer choosing equipment, they would not accept a loss of critical capability to save money.
Although Hesai was designed and manufactured in China, its average selling price was at one point higher than Velodyne’s, and gross margins on related products stayed around 70%–80%. Li Yifan directly rejects the idea that customers defected “for better value”: the industry data showed clearly better quality and higher ASP.
This premium starting point gave a hard-tech company a rare window: customers would pay for quality, allowing high margins to fund continued R&D. Entering from the low end could lock product definition around a single cost target.
27. “Technology Route” Is a Narrative; Victory Comes from Thousands of Specific Choices
In more than a decade, Li Yifan has rarely used the term “technology route,” because it implies that a company cannot turn back after making a choice. He prefers “technology choice”: others can make the same choice, and the question is who executes it better.
His engine analogy: “People making four-cylinder engines don’t talk about technology routes; people making five-cylinder engines do.” A non-mainstream design can be packaged as an inside pass, but returning to four cylinders after failing does not mean the competition is over.
Hesai decided early that 905 nm was the long-term preference for maturity, cost, performance and heat dissipation. Some peers once used 1550 nm to achieve longer range, but the industry gradually converged back toward 905 nm.
Scanning methods are similar. Hesai chose multiple lasers paired with a simple mechanical rotating mirror, while peers tried galvanometers and other approaches. Even with the same route, thousands of technical parameters remain comparable; 2 four-cylinder engines can differ enormously in quality.
28. LiDAR Is Not Vision’s Plan B; It Is an Always-On “A+”
Li Yifan’s unconditional view is that a sensing method based on different physical principles increases safety. LiDAR is not a Plan B that activates only after vision fails; it is an “A+” operating in parallel, proactively saying, “I scanned something,” when vision might miss it.
Zhang Xiaojun suggested that inconsistent signals could confuse a machine. Li Yifan rejected the premise directly: LiDAR sends out light and “touches” objects. In a dark cave, a person may not see an obstacle but feel it with a hand, and would normally trust the hand.
He acknowledges that Tesla’s formal position may be that as vision improves, LiDAR’s marginal contribution will shrink. But how small “shrink” means should be answered by accident statistics and actual system data, not belief alone.
Market direction also cannot be judged from a few models without LiDAR, because historically almost every car lacked it. Li Yifan says to look at incremental installations; the industry data he sees still shows annual growth above 50%.
29. From Feature to Safety Component, Lower Prices Changed Consumer Decision-Making
Li Yifan once viewed LiDAR as a “feature”: it enabled more autonomous-driving functions, but a good driver could handle the car without it. He now believes it is becoming a “safety component,” able to participate in braking at night and around irregular obstacles even when a high-level function is not activated.
A company driver refused to use assisted driving because, as a professional driver, he did not think the system was safer than he was; using it could increase his occupational risk. Li Yifan advances the argument with a question: when buying a car, would the same person voluntarily remove airbags, seat belts or better brakes to save money?
The shift depends on the 99.5% cost reduction. A RMB300,000 LiDAR could never be standard equipment; at roughly $200, automakers can include a capability that is extremely unlikely to be needed but could save a life.
30. “People Drive with Their Eyes Only” Is Biomimicry, Not First Principles
Li Yifan finds that “first principles” is often used as an authority phrase when people cannot explain something. When he asks colleagues to define it, many simply say they heard Elon Musk use it and do not understand the meaning.
He interprets first principles as deriving from the most atomic mechanism rather than relying on statistical analogy. “Humans use only their eyes, so machines only need cameras” is actually biomimicry—and arguably points in the opposite direction from first principles.
Biomimicry can also fail. Cars do not imitate tigers running, and planes do not simply copy birds flapping their wings. Li Yifan notes that Leonardo da Vinci proposed flight using birds’ method around 1500, while methods using fluid-dynamic pressure differences to generate lift emerged around 1700. Yet he believes humanity still spent a long time trying to build things that looked like birds. Whether pure vision works should likewise be judged by results.
31. Bosch Bet $170M; Hesai Shifted from Expensive Equipment to Automotive Industrialization
In August 2019, Bosch Group led a roughly $170M investment in Hesai. Li Yifan says Bosch had reviewed more than 100 LiDAR companies globally and had spent nearly 2 years in discussions with Hesai’s business division before calling it “the only hope in the whole village.”
The investor was Bosch’s operating business, not Bosch Ventures. The 2 sides were both partners and competitors: Bosch continued developing its own LiDAR while sharing automotive supply-chain, quality and mass-production experience on a reciprocal basis, only abandoning the direction years later.
The investment pushed Hesai to industrialize its small-volume, high-price Robotaxi products more seriously. The greatest challenge was not building a sample, but convincing automakers that a Chinese company experienced with tens-of-thousands-of-yuan equipment could make it cheap, automotive-grade and stable at scale.
32. Li Auto’s First Mass-Production Design Win Upgraded 64 Lines to 128 Without Changing 3 Things
From late 2020 to early 2021, Hesai won a design award from Li Auto and became its first mass-production passenger-car customer. Li Yifan says Li Auto’s rare judgment at the time was not to blindly trust overseas suppliers, but to believe Chinese automakers and suppliers could eventually surpass overseas peers.
After reviewing the sample, Li Auto said it was good overall, but 64 lines were insufficient. The product had to be upgraded to 128 lines without changing its size, cost or quoted price. For Hesai at the time, this was nearly “impossible.”
Product definition was completed within months; the real difficulty was full design, validation and mass production. The project was finalized in early 2021, shipped with the Li Auto L9 in July 2022 and reached monthly production above 10,000 units in September.
Li Yifan calls Li Auto’s approach genuinely first-principles: instead of imposing legacy automaker standards on a supplier, the 2 sides jointly analyzed the hardest parts of the new product, exposed risks early and decided what each would need to do.
33. In High-Difficulty Projects, Customer-Supplier Blame Games Only Produce 2 Losers
Hesai’s engineering head later described working with Li Auto as “2 departments of the same company.” Both sides were aligned around solving the same problem rather than guarding contractual boundaries and waiting for the other to fail.
When a technical issue arises, a traditional customer can easily say, “It’s all your problem; I have no problem.” But if the supplier cannot solve it, the vehicle still cannot be delivered on time. Li Auto was willing to ask, “What do I need to do?” and, when necessary, take on risks that did not formally belong to the customer.
Li Yifan’s conclusion is that complex new technologies cannot produce a customer winner and supplier loser: projects that emphasize traditional customer-supplier identities often end with both sides losing. Getting a large organization to abandon that power reflex is itself difficult.
34. Xiaomi, Meituan and Hillhouse Were Looking at More Than a Potential Supplier
In 2021, as Hesai continued raising money for mass production, Xiaomi invested more than $100M. Li Yifan valued not only Xiaomi’s car ambitions, but its ability to combine product definition, supply chain, manufacturing, quality control, industrial design and capital into a system that was “safe to buy with your eyes closed.”
Lei Jun understood Hesai’s new changes and its logic for building its own capacity after a relatively brief discussion. Meituan was already using Hesai products in autonomous delivery vehicles, while Wang Xing’s long-term thinking about automotive and robotics strategy provided another basis for cooperation.
Zhang Lei of Hillhouse also concluded after speaking with the team that Hesai had thought through its strategy and already had a product, and became bullish. Li Yifan does not view these conversations merely as sources of capital; financing is never a product milestone worthy of an internal celebration.
35. China Is the “Gym”; Overseas Markets Monetize the Muscle
Early mass-production programs were low-volume and R&D-intensive. If the full investment in chips, factories and quality systems were included, the business could still lose money. Li Yifan does not blame Li Auto for low margins; he acknowledges that Chinese automakers face intense competition and pressure on their own net margins.
He wants the supply chain to focus on whether capabilities have been built: whether quality, manufacturing, chips and R&D have withstood the test of the Chinese market. “In China, you go to the gym to build muscle; overseas, you make money from it.”
Overseas peers face neither China’s high-pressure automotive competition nor the same pace of opportunity. If Hesai builds its products and delivery system in China before entering overseas markets, competing becomes much easier.
36. Starting with Premium Customers Prevented Cost from Becoming the Only Product Definition
After Li Auto, Hesai’s customers expanded to Lotus, HiPhi, Jiyue, Changan and Great Wall; China joint-venture programs included Ford, Audi and GM systems. Among companies that had not yet adopted its products, Li Yifan mentioned Tesla directly; other reasons should be answered by the customers.
Whether in Robotaxi, domestic mass-production vehicles or overseas markets, Hesai tries to start with higher-end brands. Premium customers ask about performance and quality first, then demand sustained cost reduction; lower-end models are more likely to ask only about cost, forcing suppliers to sacrifice other metrics.
Li Yifan attributes this starting point to both choice and luck. Its first domestic mass-production customer, Li Auto, was relatively expensive at the time, allowing Hesai to prove the best product first and then move down the scale and cost curves rather than climb upward from a low-cost label.
37. A European Global Mass-Production Win Came from Years of POCs, Not One Brilliant Sales Pitch
Hesai has disclosed a global mass-production partnership with a top European automaker but, under confidentiality obligations, does not name it. This is different from a project for an overseas brand’s Chinese joint venture: the product enters the automaker’s global system and follows headquarters’ quality and procurement processes.
Trust began with Robotaxi projects around 2018–2019 and developed through multiple proofs of concept. Each small project required Hesai to deliver the relevant result, complete validation and submit the findings in full before earning the next project.
The supplier pool narrowed from more than 10 companies to 2 or 3 and then to the final award. Li Yifan emphasizes that European customers will not trust a Chinese company after one conversation; they need to see delivery, transparency and quality systems repeatedly before putting a long-term vehicle program on it.
38. Velodyne’s ITC Case Once Directly Threatened Hesai’s US Business
In 2019, Velodyne sued Hesai and RoboSense in the US for patent infringement. Constrained by confidentiality agreements, Li Yifan does not disclose the settlement terms, but says the team’s first priority was to hire the best lawyers.
From August 15, 2019 to June 15, 2020, Hesai went through the full US International Trade Commission and patent-court process. The parties submitted evidence and sued one another over nearly a year, consuming enormous management bandwidth.
The danger of the ITC is its speed: a loss could lead to a US sales ban. In 2019, many of Hesai’s customers were still in the US, so this was not an ordinary legal expense but a matter of survival. Li Yifan received the complaint while ill, held an internal meeting the next day and flew to the US.
39. Shanghai’s 2022 Lockdown Left No Failure Option for the L9 Program
Hesai was the sole supplier of the LiDAR on Li Auto’s L9, and the system was standard equipment, leaving no option to absorb failure by reducing configuration or switching to a second supplier. The project was already in the most stressful phase before delivery when the pandemic and semiconductor shortages added further pressure.
The factory moved to closed-loop production, with employees living onsite. Li Yifan did not move into the factory for symbolic reasons; he judged that his role was to solve issues such as new capacity, and traveled to the US to coordinate as the pandemic neared its end.
His management standard is not that the boss must appear in the most visible place, but that “everyone must be where they most need to be.” The project shipped on schedule in July and reached monthly production above 10,000 units in September.
40. The IPO Was a Financing Milestone; Products and Customers Were Worth Celebrating
Before Hesai’s US listing, most employees did not know the news was confirmed until the day before the bell. The bell rang at 9 a.m. New York time; core staff in China were summoned several hours earlier and told, while the office prepared only signs for group photos.
The 3 founders returned to China without a celebration dinner, and the company held no large IPO event. The only commemorative item was a power bank that Li Yifan helped design. He says listing was “far less valuable than delivering a product, developing something new or winning a major customer.”
During IPO preparation, Li Yifan estimated the probability of success at only 20%, leaving the sales team surprised that the company was investing so much. He replied that almost none of Hesai’s truly meaningful achievements began with a 90% probability: “Most particularly meaningful things are 20% things.”
The point is not blind risk-taking, but that entrepreneurs have a reason to believe the real probability is higher than market consensus and prepare for the worst case. Not building an overseas substitute, winning overseas customers first and listing in the US all initially looked like 20% choices.
41. Globalization Must Move from “Our Product Is Better, So You Should Buy It” to Local Mutual Development
Li Yifan reflects that Hesai’s past overseas logic was close to “extracting money”: its product was better than local suppliers’ and sometimes cheaper, so customers should choose it. That logic works for a small company, but inevitably encounters political and social resistance at scale.
Mature global companies explain how cooperation will create local jobs, capabilities and industrial benefits, allowing companies, governments and citizens to share the gains. Chinese companies that emphasize only competitiveness but do not answer “what are you doing for me?” will struggle to build lasting partnerships.
Hesai has encountered patent, international-trade and Department of Defense issues. Li Yifan stresses that the company has not participated in military projects, but acknowledges that he previously had not thought enough about how to coexist with the world. Globalization requires not only greater aggression, but greater intelligence.
42. America Wants Directness, Germany Wants Indirection; Global Business Requires Local Syntax
Li Yifan calls the US a “new-money” culture: customers value efficiency, and you can say directly that you are number one; they will then verify whether you really are. US customers are also willing to state product weaknesses, commercial motives and the real reason they do not want publicity.
The European “old-money” culture represented by Germany is more wary of self-promotion. If the first page of a deck is full of “number one in the world,” the customer may first assume you are a fraud. Hesai usually starts with patents, technical details and problems it has solved, allowing customers to discover its market share “unintentionally.”
Hesai’s European team therefore relies on local employees from Germany and elsewhere for communication. Germany, France, the UK and Spain also have complicated differences among themselves. Chinese customers cannot be treated as one market either; communicating with Li Auto, BYD and SAIC requires different rhythms and organizational interfaces.
43. Hard-Tech Founders Must Respect Traditional Functions and Protect Time for Strategy
Early on, because Robotaxi products could “sell themselves straight off the table,” Hesai assumed the world did not need sales. Once it entered mass-production vehicles, it learned to respect traditional functions: the volume of organization-to-organization communication is enormous, and even the finest wine can be buried in a lane that is too deep.
The same lesson applied to finance, legal, HR, quality, manufacturing and sales. Li Yifan once believed smart people could learn everything from zero, but later acknowledged that experienced professional managers solve problems far more effectively than founders trying to make up for a lack of industry experience.
His second reflection is that survival consumed so much time that there was not enough left to think about strategy several years out. A first-time founder rarely sees the whole route before setting out, but “we may not survive” cannot become a permanent excuse to abandon moats, capability accumulation and the long-term path.
His third suggestion is to reduce entrepreneurial isolation. The 3 founders could speak openly about anything, but they shared the perspective of one company. External exchanges at Hupan University and the World Economic Forum allowed people from different fields to change his thinking in a single short conversation; whether they were called friends mattered less.
44. Primary Markets Seek Non-Consensus; Secondary Markets Require Understanding How Consensus Forms
After more than 100 IPO roadshow meetings, Li Yifan finally “realized” that secondary-market communication is not only about explaining why the company is good, but why other investors will also think it is good, because stock prices require broader consensus.
Entrepreneurs naturally like to describe what they did right that others have not yet understood. That works for emphasizing fundamentals and non-consensus thinking in early-stage investing; in public markets, talking only about contrarian decisions can scare away people unwilling to make an isolated bet.
He calls the adjustment “making peace with the world”: understanding how primary- and secondary-market investors operate and communicating to their respective concerns does not mean pandering or inventing stories. Then comes “making peace with yourself”—volatility is part of the entrepreneur’s profession and need not produce extreme elation or despair.
45. Hard Tech Without Network Effects Is a Marathon; Winning the First Battle Guarantees Nothing
Li Yifan believes the previous generation of internet companies was “winner takes all.” Network effects made “every battle is decisive” a rational strategy: once ahead, data and users continued reinforcing the leader, making the early pressure immense and the later stages relatively easy.
Cars, robots and most physical-world AI products mainly have economies of scale, not irreversible network effects; the leader can still lose its lead. Even for GPT- and OpenAI-type products, he does not believe services with low switching costs have formed classic network effects.
Hard tech therefore cannot fire all its ammunition in the first battle. It must follow Zeng Ming’s principle: “look 10 years out, think 3 years out and execute 1 year at a time.” The more important question is not whether the next battle can be won, but what capabilities will be accumulated after winning and how the window will become a durable moat.
46. Manufacturing Is Part of R&D; Automotive Capabilities Can Be Reused Downmarket in Robotics
Hesai developed chipification and its own factories early and insisted that “manufacturing is part of R&D,” rather than developing first and handing drawings to a contract manufacturer. Li Yifan believes design and manufacturing must form a closed loop; a company without real manufacturing capabilities will struggle to have the strongest R&D.
The next growth opportunity after cars is not a single scenario, but a collection of robots: ports, factories, lawn mowers, humanoid robots and robot dogs. Giving these customers automotive-grade products is a reuse of platform capability compared with Hesai’s earlier consumer or industrial products.
At the time of the interview, nearly 40% of Hesai’s business came from non-automotive applications and carried good margins. Each vertical was too small on its own, but together they rode on the passenger-car chip and manufacturing platform. Automotive business continued along the “safety component” path to raise adoption rates.
Li Yifan imagines LiDAR truly disappearing only if humanity fully enters the metaverse, with every brain floating in a vat and no interaction with the physical world. As long as sensing value exceeds cost, phones and robot vacuums already show that applications do not need to involve human life to make sense.
47. The AI-Education Opportunity Is Turning One-on-One Tutors from a Scarce Resource into a Universal Product
Asked about current venture opportunities, Li Yifan chose AI education. He agrees that the highest-quality education is one-on-one tutoring because education must adapt to the individual; the historical constraint is the limited supply of excellent teachers, while AI could theoretically multiply marginal supply.
His example is studying the history of the US Civil War. Books and public videos are not tailored to an individual’s knowledge structure; the ideal product would be like an expert on the subject sitting in front of you and giving a continuous 3-hour explanation based on your questions. AI makes “a customized teacher for everyone” feasible.
He believes the larger unresolved question remains US-China relations: how much the 2 countries should continue embracing each other versus moving toward separation is still unclear. That directly affects the capital, customers and globalization path of Chinese hard tech.
48. Industry Opportunities Come from the Nation and Its People; Engineering Output Must Be Followed by Brand and Cultural Output
Li Yifan believes founders are narrow-minded if they attribute all success to themselves and their teams. Team effort matters, but without opportunities created by industries such as new-energy vehicles, there would be no stage on which to perform. Those opportunities did not appear from nowhere; they were built through the nation’s investment over a 10-year horizon.
A trip to India reinforced the point. Li Yifan believes China and India may have started from similar positions in the 1990s, but the gap is now enormous; Indian companies and officials openly discuss “China Plus One” and catching up in manufacturing. Compared with cities that have always been wealthy, such as New York and London, this contrast made China’s development more tangible to him.
He sees 3 tasks for this generation of Chinese entrepreneurs: first export world-class engineering capability, then make global users remember Chinese brands, and ultimately export national culture and soft power. America’s strength comes not only from Apple, but also from the NBA and Hollywood, which created an image of “a better world worth learning from.”
When overseas customers visit China for the first time, Li Yifan takes them to the Bund and puts them on a high-speed train before talking about LiDAR. If they do not believe China can improve on imported technology, they are unlikely to accept an expensive Chinese brand. His final formulation is: “Where do the industry’s opportunities come from? They come from the nation and its people,” and the goal is for the world to know “how cool Chinese people are.”