Hesai Group: Betting the World's Cars See in 3D

I. Cold Open & Episode Roadmap

In October 2014, in a rented office in Shanghai, three engineers in their late twenties incorporated a company called 上海禾赛光电科技 Shanghai Hesai Photonics. Their first product was not a self-driving car sensor. It was a laser device for sniffing out methane leaks in natural gas pipelines.³ The pivot to lidar came two years later, almost as an afterthought — a side project that ate the company.

Twelve years on, that side project has produced one of the strangest résumés in global technology. 禾赛科技 Hesai is the only lidar company on earth that has ever reported a full year of profit under U.S. accounting standards. It shipped more than 1.6 million laser sensors in 2025 — roughly one every twenty seconds, around the clock, for a year.¹ It is the named lidar partner on Nvidia's flagship autonomous-driving reference platform.²⁸ It has been chosen by Mercedes-Benz to supply the eyes for the German company's Level 3 hands-off driving programs in Europe and China.²⁵

It is also, according to the United States Department of Defense, a Chinese military company. Hesai was placed on the Pentagon's Section 1260H list on January 31, 2024, sued to get off it, was removed and re-listed on the same day in October 2024, lost in federal district court in July 2025, appealed, and remains on the list today.² Since June 30, 2026 — five weeks ago — companies on that list have been barred from contracting with the Department of Defense.² Three U.S. senators have formally asked the SEC to kick Hesai off the Nasdaq entirely.¹⁸

So here is a company that raised money twice, on two continents, in the two largest lidar equity raises in history, that out-executed a decade of better-funded Silicon Valley and Detroit-backed rivals on cost, and that spends a meaningful share of its legal budget arguing it is not an arm of the People's Liberation Army.

This is the story of how a laser-ranging startup won the sensor war and inherited a geopolitical one. The spine runs like this: what lidar actually is and why the entire auto industry has been fighting about it for fifteen years; how Hesai survived a decade as a Chinese hardware startup while the American cohort blew itself up on SPAC money; how it rode China's advanced-driver-assistance boom to the first profit anyone in the category has ever booked; why Washington's designation is not a footnote but arguably the dominant variable in the equity; what the Hong Kong dual listing was really for; who else is in this race and who is actually winning; how much of the robotics and "spatial intelligence" story is real optionality versus press-release optionality; and finally, what a skeptical investor should be watching over the next four quarters.

Start with the technology, because almost every argument about this company eventually reduces to a physics-and-cost argument that most people get wrong.

↑ Back to Top

II. What Is Lidar, and Why Does This Industry Exist

Picture a bat in a dark cave. It shrieks, waits, and listens for the echo. The delay tells it how far away the wall is. Do that thousands of times a second, in every direction, and the bat builds a map of a space it cannot see.

Lidar — Light Detection and Ranging — is the same trick with light instead of sound. The sensor fires a pulse of invisible laser light, waits for it to bounce off something and come back, and measures the round trip. Light travels about thirty centimetres per nanosecond, so the timing electronics have to be brutally precise. Fire a million of those pulses per second in a sweeping pattern and you get a "point cloud": a live, three-dimensional dot-sculpture of the world, accurate to a few centimetres, that works in pitch darkness and does not care whether the object in front of the car is a pedestrian, a fallen mattress, or a shipping container that fell off a truck.

That last part is the whole argument. A camera is a fantastically cheap sensor that produces a flat image; software then has to guess depth and guess what it is looking at, based on patterns it has seen before. Lidar does not guess. It measures. A camera-based system has to recognise a white truck across a bright sky in order to avoid it; a lidar simply registers that something solid is 40 metres ahead and closing.

This is why the industry split into two camps. Tesla bet that cameras plus enough neural-network training would eventually match or beat human vision, and stripped lidar out entirely on cost grounds. Almost everyone else — every serious robotaxi operator and, increasingly, every automaker chasing hands-off highway driving — landed on "sensor fusion": cameras for classification and colour, radar for velocity and bad weather, lidar for ground truth on geometry. The fusion camp's argument is not that lidar is smarter. It is that lidar fails differently from cameras, and two sensors that fail differently are far safer than one sensor that fails confidently.

The economics are where the story actually lives. The spinning drum on the roof of Google's first self-driving prototypes was a Velodyne unit that cost roughly as much as a luxury car. At those prices lidar was a research instrument, not an automotive component. The entire fifteen-year history of this industry is the story of driving that number down by roughly three orders of magnitude — from a machine you bolt to a test mule to a part you can specify on a mass-market sedan.

The mechanism was replacing motion with silicon. Early lidars were mechanical: a stack of lasers and detectors physically spinning at 600 rpm inside a housing. Reliable enough for a lab, hopeless for a car that must survive a decade of potholes and Arizona summers. The industry moved to "semi-solid-state" designs — a tiny mirror oscillating inside a sealed unit, far fewer moving parts — and then began pulling the discrete components onto custom chips. Hesai's version of this is a proprietary application-specific integrated circuit that folds the transmit and receive electronics onto silicon, which is why the company now describes itself as having "full-stack proprietary ASIC capabilities."³² In plain terms: the more of a lidar you can print on a chip instead of assembling by hand, the faster the price falls, and the more the business starts to behave like semiconductors rather than like precision optics.

A useful analogy for the transition: early lidar was a hand-built telescope, assembled and calibrated by technicians. Modern lidar is closer to a smartphone camera module — a chip, a lens, a bit of packaging, produced by machines that never get tired and never mis-align an optical path by half a micron. Companies that made the jump from the first world to the second survived. Companies that tried to scale the first world by hiring more technicians did not.

Two businesses in one company

One more distinction matters for everything that follows, because Hesai is really two businesses wearing one name. The first is ADAS lidar: long-range, forward-facing units bolted behind a windshield or into a roofline on passenger cars, designed to see 150 to 250 metres down a highway. These sell in enormous volumes at low prices. The second is robotics lidar: shorter-range, wide-field-of-view units for robotaxis, delivery robots, warehouse vehicles, humanoids, and — genuinely — robotic lawn mowers. These sell in smaller volumes at better margins. In 2025, ADAS accounted for about 1.38 million of Hesai's 1.62 million units; robotics was the other 239,000, growing more than five-fold year over year.¹

Hold that split in your head. Nearly every argument in this story — margins, competition, geopolitics, valuation — is downstream of it.

↑ Back to Top

III. Founding and the First Decade: Surviving as a Chinese Hardware Startup (2014–2022)

There is no monsoon scene here, no thunderclap of inspiration. The founding of Hesai is a story about three engineers who were, by their own trajectory, mildly overqualified for the problem they first chose.

李一帆 Li Yifan — "David" Li to the English-speaking world — took a bachelor's degree in engineering at 清华大学 Tsinghua University, then a master's and a PhD in mechanical engineering at the University of Illinois at Urbana-Champaign, with a research focus on robotics. He went to work at Western Digital in Silicon Valley in August 2013 and lasted fourteen months.² In October 2014, he and two co-founders — 孙恺 Kai Sun, who became chief scientist, and 向少卿 Xiang Shaoqing, who became chief technology officer — incorporated the company in Shanghai.² Li was 28. All three are roughly the same age, and remarkably, all three are still there: as of the 2025 annual report, Li was 40, Sun 40, and Xiang 41, and all three sit on the board as executive directors.²

Founding-team continuity of that length in hardware is rare enough to be worth pausing on. It is not automatically a virtue — long-tenured founders can also mean entrenchment — but it does mean the institutional memory of every technical bet the company has made still sits in the room.

The first bet was wrong, or at least small. Hesai's initial focus was high-performance laser sensors for natural gas and industrial applications — the methane detector business.³ It worked well enough to keep the lights on. But by 2016 the founders had shifted the company's primary business to lidar, and by 2019 they had a product, the Pandar64, that a specific customer set desperately wanted.³

That customer set was robotaxi developers. In 2018 and 2019, every autonomous-driving program on earth needed a good mechanical spinning lidar, and there were perhaps three companies that could make one. Hesai became one of them. The Pandar64 accounted for 64.2% of Hesai's revenue in 2019.³ This was a wonderful place to be commercially and a terrifying place to be strategically: a single product, sold to a handful of research-stage customers, in a market whose total addressable size was measured in tens of thousands of units.

It also attracted the attention of the incumbent. In August 2019, Velodyne — the American company that had invented the category — sued Hesai for patent infringement in U.S. federal court and at the International Trade Commission. Hesai countersued in Frankfurt and in the Shanghai Intellectual Property Court. On June 24, 2020, the two settled with a global cross-licence covering existing and future patents on both sides, under which Hesai agreed to pay Velodyne a one-off fee plus an annual royalty running through 2030.³

At the time this read as a Chinese upstart paying protection money. It aged into one of the most valuable contracts Hesai ever signed, for reasons that would only become clear three years later.

The capital that funded the pivot came from an unusually strategic register. By the time of the Nasdaq listing, Hesai's shareholder base included Baidu, Bosch, and Xiaomi, each holding around 5–6% of the company on an as-converted basis.³ Money from a German Tier-1 supplier and from China's leading autonomous-driving software platform is not just money. It is a signal to procurement departments at automakers that this supplier has been diligenced by people who know what they are looking at.

The factory bet

Then came the decision that made the company. Around 2021–2022, Hesai committed to designing cheap, mass-manufacturable lidar for ordinary passenger cars — not for robotaxi fleets — and to building the factory to produce them before the orders existed. The Maxwell centre in Jiading, Shanghai, a combined R&D and intelligent-manufacturing site, entered trial operation in December 2023; a second facility, Hertz, had begun phased operation earlier.² The logic was that lidar was becoming a semiconductor-like business, and in semiconductor-like businesses the low-cost producer at scale wins, because every doubling of volume pulls unit cost down and lets you underbid the next tender.

That is a capital-intensive, faith-based bet. It is also the exact opposite of what the American lidar cohort was doing. Between 2020 and 2021, Velodyne, Luminar, Innoviz, Ouster, Aeva, AEye, Cepton, and Quanergy all went public via SPAC, most of them on projections of robotaxi-scale volumes at automotive prices. They raised money on decks and then went looking for the cost curve. Hesai found the cost curve and then went looking for money.

By late 2022, Quanergy had filed for bankruptcy. The rest were burning cash and revising forecasts down. And Hesai, quietly, was preparing to list in New York.

↑ Back to Top

IV. The Nasdaq IPO and the U.S. Lidar Shakeout (Feb 2023)

February 2023 was not an obvious month to take a Chinese company public in America. The audit-inspection standoff between Washington and Beijing had only just been defused. Chinese ADRs had been a value trap for two years. And the lidar sector specifically had become a punchline — an entire cohort of de-SPACed companies trading at a fraction of merger prices.

Hesai went anyway. On February 9, 2023, it priced 10.0 million American depositary shares at $19.00 each, the top of a $17–$19 range, raising $190 million gross.⁴ The stock opened at $23.75 the next morning, 25% above the offer, and closed its first session at $21.05, giving the company a market capitalisation of about $2.6 billion.⁵ It was the largest U.S. IPO by a Chinese company in eighteen months, and the first deal in which Goldman Sachs, Morgan Stanley, and Credit Suisse had jointly underwritten a U.S.-listed Chinese issuer since 2021.⁵

The symbolism ran deeper than the size. Four days after Hesai's debut, on February 13, 2023, Ouster and Velodyne completed what both companies called a "merger of equals."⁶ Read the phrase carefully. Two of the most celebrated names in American lidar — one of them the literal inventor of the category — had combined not to accelerate anything but to stop bleeding. Quanergy was already gone.

The contrast is the lesson of this entire era, and it is worth stating precisely rather than triumphantly. Hesai did not have better physics than Velodyne. Its early patents were close enough to Velodyne's that it paid to license them. What Hesai had was sequencing: it built manufacturing scale and drove unit cost down before asking public markets for money, so its listing was a financing event rather than a survival event. The American cohort inverted that order, and the capital markets, eventually, charged them for it.

The legal thicket that came with the ticker

The Nasdaq honeymoon was short. Two months after the IPO, on April 7, 2023, Hesai and certain of its officers, directors and underwriters were named in a putative securities class action in the Eastern District of New York, alleging that the IPO offering documents contained false or misleading statements. The case was later transferred to the Southern District of New York; on February 18, 2026, the court granted in part and denied in part the defendants' motion to dismiss, meaning parts of it survive.² Four days after that class action was filed, on April 11, 2023, Ouster sued Hesai for patent infringement in Delaware and simultaneously asked the ITC to open a Section 337 investigation that could have barred Hesai products from the United States entirely.²

Here the 2020 Velodyne cross-licence came back to life. Hesai's argument was essentially that Ouster, having merged with Velodyne, had inherited Velodyne's obligations — including the covenant not to sue. Hesai pushed the fight into arbitration. The ITC terminated its investigation in October 2023. In late March 2025, the arbitration tribunal issued an interim decision finding that Ouster was indeed bound by the Hesai–Velodyne agreement; in May 2025, the Delaware court dismissed Ouster's patent case with no conditions, no settlement payment, and no injunction; and on September 15, 2025, the tribunal's final decision awarded Hesai roughly $6.4 million in costs and fees, which Ouster paid in the fourth quarter of 2025.²

Consider what was at stake in that ITC case. A Section 337 exclusion order does not award damages; it bars products at the border. Had Ouster prevailed, Hesai's entire U.S. robotics and autonomous-mobility business would have been shut off by customs officers rather than by a court judgment — a far more immediate commercial death than any monetary award. The company avoided that outcome not through superior lawyering in 2023 but through a contract signed in 2020, when it was the smaller party paying to settle.

For investors, the takeaway is not that Hesai is litigation-proof — it plainly is not. It is that a settlement the market once read as a defeat turned out to be the instrument that neutralised the single most dangerous legal threat to Hesai's U.S. market access. Good IP hygiene in hardware is a slow-compounding asset that shows up in exactly one place: the absence of a catastrophe.

Meanwhile the operating business was doing something the share price did not yet reflect.

↑ Back to Top

V. The Inflection: Winning China's ADAS Volume War and Reaching Profitability (2023–2025)

Somewhere around 2023, a feature war broke out in Chinese showrooms, and it changed lidar from a science project into a commodity.

The feature was NOA — "navigate on autopilot," the ability of a car to handle highway on-ramps, lane changes, and eventually city streets with the driver supervising rather than steering. Chinese electric-vehicle brands discovered that NOA sold cars, and that a visible lidar pod on the roofline was proof to a showroom customer that the car had the good version. The sensor became a marketing object as much as an engineering one — the automotive equivalent of a camera bump on a phone.

Hesai was positioned almost perfectly for this. It had the factory, the automotive-grade quality systems, and a chip-integration roadmap that let it keep cutting price without destroying margin. What followed was one of the steepest volume ramps in automotive components in recent memory.

The numbers tell it plainly. Full-year 2024 revenue was RMB 2,077.2 million on 501,889 lidar units, and the company reported non-GAAP net profit of about RMB 14 million — swinging from a non-GAAP loss of RMB 241 million a year earlier.⁷ That RMB 14 million is, in absolute terms, a rounding error: a profit margin under one percent. In context it was a genuine landmark, because no lidar company anywhere had ever produced a full year in the black. Hesai also reported positive operating cash flow of RMB 63 million and a net cash position of about RMB 1.3 billion.⁷

Then 2025 turned the landmark into a business. Revenue reached RMB 3,027.6 million, up 45.8%, on 1,620,406 units — total shipments more than tripled.¹ GAAP net income was RMB 435.9 million, against a GAAP net loss of RMB 102.4 million in 2024; non-GAAP net income was RMB 550.5 million.¹ The fourth quarter alone did RMB 1,000.5 million of revenue and RMB 153.2 million of net income.¹ The company finished the year with a cash reserve of RMB 7,511.0 million and total shareholders' equity of RMB 8,958.8 million.¹

What the numbers actually prove

What does that actually prove? Three things, and one non-thing.

It proves demand is real and broad rather than concentrated. Hesai's top five customers accounted for 67.5% of revenue in 2023, 59.9% in 2024, and 55.8% in 2025 — concentration falling steadily even as revenue grew.² More strikingly, the single U.S.-headquartered global OEM that contributed 28.4% of 2023 revenue fell below 5% in both 2024 and 2025, and the company still grew.² Losing a quarter of your revenue base and doubling anyway is the clearest possible evidence that the customer set widened.

It proves the manufacturing thesis. Hesai ended 2025 with 1,118 employees, of whom 587 were in research and development and only 206 in production and supply chain, supported by 1,879 contract workers.² A company producing 1.6 million precision optical instruments a year with roughly 200 direct production employees is running a genuinely automated line, not a labour-arbitrage one. That distinction matters enormously for whether the cost advantage survives wage inflation or a move offshore.

It proves operating leverage exists. Between 2024 and 2025, revenue grew about 46% while net income swung by more than half a billion renminbi. Fixed-cost absorption in a heavily automated plant is the most straightforward margin mechanism in industrial businesses, and here it worked as advertised.

The non-thing is durability, and this is where the bear starts sharpening. Gross margin was 42.6% in 2024 and 41.8% in 2025.¹ Then in the first quarter of 2026 it fell to 39.1%, from 41.7% a year earlier, on revenue of RMB 680.6 million and 471,723 units.¹¹ Management's explanation, from CFO 范鹏 Peng Fan on the Q4 2025 call, was that the anticipated decline in average selling price "is mainly driven by product mix" rather than by price war, with the high-volume ATX product heading toward roughly $150 per unit in 2026, and future declines expected to narrow.¹³

That explanation is plausible and also unfalsifiable in the short run — "mix, not price" is what every component supplier says on the way down. The honest reading is that Hesai is deliberately trading margin percentage for volume and share, which is the correct strategy in a scale business and simultaneously the thing that destroys equity value if the volume stops arriving. Two quarters of stable-to-rising gross margin would settle the argument in management's favour. Two more quarters of decline would not.

There is also a quality-of-earnings question sitting inside the 2025 profit that deserves its own paragraph, and it will get one in the management section, because it involves a company the founders themselves control.

Externally, the scoreboard reads well. Yole Group ranked Hesai first in long-range ADAS lidar shipments for passenger cars in 2025 with a 43% volume share, its fifth consecutive year at number one; cumulative deliveries passed two million units in November 2025, and monthly output exceeded 200,000 units.⁸ Design wins spanned 40 automotive brands and more than 160 vehicle models.⁸

Those are impressive numbers. They are also, as the next-but-one section shows, a carefully chosen slice of a market where Hesai is not actually the largest player by units.

↑ Back to Top

VI. Geopolitics as a Structural Risk, Not a Footnote: The Pentagon Blacklist and the Short-Seller Fight

On January 31, 2024, without warning and without a hearing, the U.S. Department of Defense added Hesai to the list of "Entities Identified as Chinese Military Companies Operating in the United States" maintained under Section 1260H of the FY2021 National Defense Authorization Act.² The company's stock fell. Its American customers started asking questions its salespeople could not answer.

Hesai's response was unusually combative for a Chinese issuer. It announced within days that it would challenge the listing.¹⁴ On May 13, 2024, it filed suit in the U.S. District Court for the District of Columbia, arguing that the DoD had exceeded its statutory authority, acted arbitrarily and capriciously under the Administrative Procedure Act, and violated the Constitution both in adding the company and in refusing to remove it.²¹⁵¹⁶ After obtaining the Department's decision memorandum — which Hesai said failed to justify the accusation — the company asked for an expedited summary-judgment schedule and moved for judgment on July 3, 2024.²

What happened next is the part that tells you how this system really works. On October 15, 2024, the DoD removed Hesai from the list — and on the same day, added it back, this time on a fresh administrative record built from "the latest information available."² The government had, in effect, mooted the lawsuit's target and rebuilt the designation underneath it.

Hesai kept fighting and lost. On July 11, 2025, Judge Paul Friedman upheld the designation, and the company filed notice of appeal two days later.²¹⁷ The district court's reasoning is worth understanding precisely, because it is both narrow and, for an investor, more alarming than a broad finding would have been. The court did not find that Hesai's products had been used for military purposes. It did not find a direct or indirect connection between Hesai and the Chinese military. It upheld the designation on the basis that lidar has potential military applications, that Hesai's R&D sites sit within certain industrial districts, and that the company cooperates with Chinese government-linked institutions.¹⁷

Sit with that. If the standard for designation is "your technology is dual-use and you are physically located in China," then no amount of corporate behaviour by Hesai can cure it. That is not a compliance problem management can fix with better disclosure or an internal audit. It is a category assignment.

The D.C. Circuit heard oral argument in the appeal on March 19, 2026.² In December 2025 the appellate court had declined to coordinate Hesai's appeal with drone maker DJI's parallel challenge. As of the company's most recent annual report, filed April 24, 2026, Hesai remained on the list and stated it was unable to predict the outcome.² No appellate decision has been reported as of this writing.

Meanwhile the practical consequences started biting on a fixed schedule. Effective June 30, 2026, entities on the 1260H list and their controlled affiliates became prohibited from entering into contracts with the DoD for goods, services, or technology; effective June 30, 2027, the DoD will additionally be barred from buying listed entities' products indirectly through third parties. Listed entities and their subsidiaries are also barred from Department of Homeland Security contracts and funding.²

Hesai does not sell to the Pentagon, so the direct revenue hit is negligible. The company's own filing is candid about where the real damage lands: the continued listing "has led to a false perception that we are associated with the Chinese military, which has had a negative impact on our ongoing contract negotiations and disrupted our existing and potential customer relationships globally, in particular those in the United States."² That is a disclosed, management-acknowledged commercial harm, not a speculative one. When an American fleet operator or a Tier-1 integrator weighs a Hesai design-in against a competitor, the 1260H line item is a procurement risk that has to be escalated, and escalation kills deals quietly.

The pressure has also escalated politically rather than subsided. On September 15, 2025 — the day before Hesai's Hong Kong shares began trading — Senators Rick Scott, Marsha Blackburn, and Bernie Moreno wrote to the Treasury Department and the SEC demanding that Hesai be delisted from the Nasdaq outright and added to Treasury's Chinese Military-Industrial Complex Companies List under Executive Order 13959, citing alleged use of Hesai lidar on PLA combat vehicles and in Xinjiang surveillance.¹⁸ Hesai has denied military involvement throughout. And in July 2026, CNBC reported that U.S. officials and analysts were framing Hesai's expanded hardware partnership with Nvidia as a fresh cybersecurity concern — the argument being that Chinese-made sensors deployed across American roads constitute a data-collection surface, by design or by exploitation.¹⁹

That is three distinct escalation vectors in under three years: a defence designation, a securities-delisting campaign, and a data-security framing. The pattern is the point. This is not a one-time shock that can be modelled as a discrete event and discounted; it is a rolling process in which each new administration, each new NDAA cycle, and each new headline creates another chance for the rules to tighten.

Then there is the short seller. On March 18, 2025, Blue Orca Capital published a report and disclosed a short position; the stock fell that day. Blue Orca's central financial allegation was that Hesai's Q4 2024 profitability was manufactured — that the roughly $20 million net profit existed only because of a one-time $20 million break fee from a terminated contract with its largest customer, which it said had not been properly disclosed. It also questioned whether reported revenue was consistent with that customer's actual purchase volumes, claimed to have photographic evidence of Hesai lidar on Chinese military vehicles, and noted that Mercedes-Benz had not confirmed a supply relationship reported by Reuters on March 11, 2025, after which the stock had surged 33%. Blue Orca further highlighted that on March 13, 2025 — two days after that surge and five days before the report — company insiders filed planned share sales totalling roughly $62.58 million.²¹

Hesai's response, issued on March 19, 2025, was two sentences of substance: the company "strongly disagrees with the allegations in the report and believes that the report is without merit," and reiterated its "unwavering commitment to stringent standards of business ethics and regulatory compliance."²⁰ It did not address the break-fee claim, the revenue-reconciliation claim, or the military-imagery claim individually.

For an investor assessing management credibility, the form of that response matters as much as its content. A blanket denial is legally safe and analytically empty. It leaves the market to resolve the allegations on its own, which is a strange choice for a company whose entire Western-market problem is a trust deficit.

To be fair on the substance: the 2025 annual report does now disclose the underlying transaction. Hesai confirms that in 2024 it received a one-off payment from the U.S.-headquartered OEM following that customer's termination of a March 2023 contract, intended to compensate for R&D investment and actual costs of work-in-progress and raw materials, after the customer suspended the relevant project — and warns that "there is no guarantee that we will always receive compensation for our investments and costs incurred resulting from terminated contracts."² So the economic substance of Blue Orca's most concrete financial claim — that a one-off termination payment flattered the first-ever profit — appears in the company's own filing. Whether the original disclosure was adequate is what the surviving portions of the securities class action will test.

The clean summary for investors is uncomfortable but simple: Hesai's commercial execution question has largely been answered. Its permission-to-operate-in-the-West question has not, and the evidence to date runs against the company. Which is precisely why, in September 2025, it went looking for a second home.

↑ Back to Top

VII. The Hong Kong Dual-Primary Listing: Insurance Policy or Vote of No Confidence? (Sept 2025)

On the morning of September 16, 2025, Hesai's Class B shares began trading on the Main Board of the Hong Kong Stock Exchange under the code 2525. It became the first lidar company anywhere to hold a dual-primary listing in both the United States and Hong Kong.²²

The mechanics were substantial. The company sold 19,550,000 Class B shares at HK$212.80 each, raising gross proceeds of about HK$4.2 billion — roughly $533 million before the over-allotment option, and about $614 million after it was exercised.²²²³ It was the largest equity raise in the global lidar sector's history, comfortably exceeding the company's own Nasdaq debut.

Now: why do it?

The defensive reading is the obvious one, and management has never had to state it out loud because the risk factors state it for them. A dual-primary listing is not a secondary or depositary arrangement; each listing stands on its own regulatory feet. If a U.S. delisting order ever arrived — via the SEC route the senators demanded, via Treasury sanctions, or via a future NDAA provision — Hesai's shares would not become untradeable. They would simply trade in Hong Kong. Given that the company was, at that moment, under active political attack in Washington and had already lost its district-court case, building a fully independent trading venue was the single highest-value piece of corporate insurance available. The senators' letter landing the day before pricing was, if nothing else, exquisite timing.

The offensive reading is also real, and investors should not dismiss it as spin. Hong Kong opens access to Mainland capital via Stock Connect and to Asian long-only institutions who understand Chinese industrial supply chains better than a U.S. small-cap fund does. A shareholder base that prices Hesai as a Chinese advanced-manufacturing champion rather than as a China-risk ADR is a structurally different, and probably more patient, base.

Both readings can be true at once, and the sequencing suggests they are: the raise was large enough to be opportunistic, and timed early enough to be defensive.

There is a third reading, less flattering and worth stating. Raising $614 million when the company had just turned profitable and was generating operating cash suggests either that management sees a capital-intensity ramp ahead that internal cash cannot fund — the Thailand build and a capacity doubling both qualify — or that it wanted the balance sheet armoured before the political environment got worse. Neither is a criticism. But a company that raises heavily immediately after achieving profitability is telling investors that the profitability is not yet sufficient to self-fund the strategy. Ending 2025 with more than RMB 7.5 billion of cash against RMB 3.0 billion of annual revenue is a war chest, not a working-capital buffer.¹

Governance plumbing and the buyback that hasn't happened

The follow-through has been about governance plumbing. At an extraordinary general meeting and separate class meetings held in Suzhou on March 3, 2026, shareholders approved a package of resolutions: re-designating 50 million authorised, unissued shares as Class B ordinary shares; adopting a third amended and restated memorandum and articles of association; granting directors a general mandate to allot and issue additional Class B shares; and granting a general mandate to repurchase the company's own shares and ADSs.²⁴ A second extraordinary general meeting was scheduled for August 28, 2026, with the board recommending shareholders vote in favour of the proposed resolutions.³²

The repurchase mandate is the one to watch, and it is worth being precise about what it is. A general mandate is permission, not a programme. It authorises the board to buy back stock; it commits nothing. As of the most recent monthly securities returns filed with the Hong Kong exchange, the company had not reported share repurchases under it.³¹ For a business that ended 2025 with more than RMB 7.5 billion of cash and had just turned GAAP-profitable, the gap between authorisation and execution is the cleanest available test of whether management thinks its own shares are cheap.

The other real-time signal the dual listing created is the spread between the Hong Kong shares and the Nasdaq ADSs. Because both represent the same economic claim, any persistent premium in one venue is the market pricing venue-specific risk — most obviously, delisting tail risk in New York. That spread is a live, daily referendum on the geopolitical thesis, and it costs nothing to watch.

Which brings us to the question the market has never fully settled: is Hesai actually winning?

↑ Back to Top

VIII. Competitive Landscape: Who Else Is in This Race, and Who's Actually Winning

Here is a fact that does not appear in Hesai's investor materials.

In China's passenger-vehicle lidar market — the largest in the world by a wide margin — Hesai was not the number-one supplier by installations in 2025. 华为 Huawei was. Across January to December 2025, Huawei installed 1,406,294 lidar units for a 41.5% share; Hesai installed 1,143,162 for 33.8%; 速腾聚创 RoboSense installed 575,661 for 17.0%; and 图达通 Seyond installed 259,335 for 7.7%. The top three together took roughly 92% of a market of about 3.4 million units.¹⁰

Both statements are true simultaneously: Yole ranks Hesai first globally in long-range ADAS lidar for passenger cars with 43% share,⁸ and Chinese installation data ranks Huawei first in all passenger-vehicle lidar installations in China.¹⁰ The difference is segment definition and geography. Investors should hold both, and treat any presentation that shows only the first as marketing.

Now, the war game.

Huawei is the most misunderstood competitor because it is not really a lidar company. It is a full-stack advanced-driving platform vendor that bundles its own sensors with its own compute and its own software, sold into an alliance of partner brands. Its product cadence has been ferocious — a 96-line unit in 2020, 192-line in 2023, and by March 2026 an 896-line dual-optical-path "image-grade" lidar debuting on flagship models.¹⁰ The threat Huawei poses to Hesai is not price. It is disintermediation: when an automaker buys Huawei's whole driving stack, the lidar line item disappears into a system contract that Hesai was never invited to bid on. The counterweight is that many automakers — particularly those with global ambitions or their own software teams — have strong strategic reasons not to outsource their brain to Huawei. That reluctance is Hesai's addressable market.

RoboSense is the direct rival, and the head-to-head is now clarifying in a way that flatters Hesai on economics. RoboSense's 2025 revenue was about RMB 1.94 billion on roughly 912,000 lidar units, with annual gross margin of 26.5%, and it achieved its first-ever quarterly net profit — about RMB 104 million — in the fourth quarter of 2025.²⁶ Compare the margin lines: 41.8% at Hesai versus 26.5% at RoboSense on comparable-generation technology.¹²⁶ That fifteen-point gap is the single best available proxy for Hesai's cost and mix advantage, and it is the sort of evidence that distinguishes a genuine scale economy from a slogan.

But RoboSense has out-flanked Hesai where the growth rate is highest. It shipped roughly 303,000 robotics lidar units in 2025 and claims the global number-one position in that segment, with robotics revenue of RMB 347 million — about 49% of product sales.²⁶ Hesai shipped 239,273 robotics units in the same year.¹ So the two companies are converging on each other's territory from opposite directions: Hesai descending from high-value automotive into robotics, RoboSense ascending from robotics volume into automotive value. Neither has a defensible fortress in the other's home segment.

Livox, DJI's lidar affiliate, matters for one reason: 大疆 DJI's cost engineering is arguably the best in Chinese consumer hardware, and Livox inherits it. In the consumer-adjacent robotics categories where Hesai is now pushing hardest — mowers, service robots — that is precisely the wrong competitor to meet.

Seyond, formerly Innovusion, holds a real if smaller position.¹⁰ It is also where Hesai's current CFO came from, which is an unusual piece of talent flow between direct competitors and is discussed below.

The Western cohort is largely a cautionary tale by now. Beyond the Ouster–Velodyne combination and Quanergy's bankruptcy, the SPAC-era survivors — Luminar, Innoviz, Aeva, Cepton — have generally depended on a small number of Western OEM programmes with long timelines. Valeo remains the one legacy Tier-1 auto-parts giant still genuinely relevant. The structural fact underneath all of it is stark: Chinese lidar companies collectively account for roughly 95% of the global market.⁹ This category localised faster and more completely than almost any other automotive component in modern history.

Myth versus reality

Three consensus statements about this company deserve correction.

Myth: Hesai is the world's largest lidar company. Reality: it is the largest by long-range passenger-car ADAS shipments as defined by Yole,⁸ and second by total lidar installations in China, behind a competitor that mostly supplies itself.¹⁰ Both facts are real; only one appears in the pitch.

Myth: profitability proves the business model. Reality: profitability proves the manufacturing model. The 2024 non-GAAP profit was supported by a contract-termination payment, and roughly RMB 185 million of 2025 income came from transactions with a founder-associated entity.² The underlying operating trend is genuinely improving; the headline overstates how far it has come.

Myth: the DoD listing is a headline risk that will fade with the political cycle. Reality: it survived judicial review on structural grounds, the procurement prohibitions began on a legislated schedule, and the company's own filing states it has disrupted customer relationships.²¹⁷ It is an operating cost with a legal calendar attached.

Run Porter's five forces across that landscape and the picture is unromantic.

Supplier power is low and falling, because Hesai has pulled its critical components onto in-house ASICs, which converts an external bargaining relationship into an internal design problem.³² Buyer power is high and rising: automakers multi-source by policy, re-tender aggressively, and are themselves in a price war. Rivalry is intense, price-led, and concentrated among three well-capitalised Chinese players — the worst possible structure for sustained pricing power. Threat of substitutes is genuine but has weakened: camera-only autonomy remains the cheapest architecture and has not gone away, but the convergence of global OEMs and every serious robotaxi operator on sensor fusion has moved the argument from "will lidar be used" to "how much lidar per car." Barriers to entry are now about capital and manufacturing know-how rather than optics: Hesai held 632 granted patents and 728 pending applications in China plus 157 granted and 554 pending elsewhere as of end-2025, but the harder barrier is that a new entrant would need to build a Maxwell-class automated line and then sell at prices that only make sense at a million units a year.²

Through Hamilton Helmer's 7 Powers, Hesai holds one power clearly and one partially. Scale economies are real and measurable — that fifteen-point gross-margin gap over its closest peer is the evidence. An emerging process power exists in vertical chip integration, where the accumulated knowledge of designing lidar ASICs and running a highly automated line is genuinely hard to copy quickly. Cornered resource is arguable via the patent estate and the Velodyne cross-licence. What Hesai conspicuously lacks are the powers that make investors sleep well: there are no network effects, no meaningful switching costs once a vehicle programme ends, no counter-positioning that rivals cannot imitate, and no brand power with end consumers who could not name their car's lidar supplier.

The blunt conclusion: Hesai's advantage is a cost-and-execution lead in a commoditising component market. That is a real advantage. It is not a moat in the way a payment network or an operating system is a moat, and the industry's own history — in which Hesai unseated Velodyne, the company that invented the field — is proof that leadership here is rented, not owned.

Which is exactly why management has been talking about something other than lidar.

↑ Back to Top

IX. The Next Act: Robotics, Robotaxis, and "Spatial Intelligence" as the Real Optionality

There is a device that mows suburban lawns while its owner is at work. It has no rails, no boundary wire, and no idea what a lawn is. It knows only that there is a solid object 40 centimetres ahead that was not there yesterday, and that it should go around.

That device is now one of the more interesting lines in Hesai's order book, and the way it got there says something about how technology cost curves actually end.

Hesai launched its JT-series lidar — a compact, hemispherical unit designed to be hidden inside a chassis — at CES in January 2025, aimed at robotic lawn mowers. An initial partnership with 追觅科技 Dreame included an order for 300,000 units. On March 26, 2026, the two companies signed an exclusive supply agreement under which Hesai will supply up to 10 million JT-series units for robotic mowers across the Dreame and MOVA brands — described by Hesai as the largest order in lidar history, on top of more than 300,000 units already delivered.²⁷

Ten million units is an eye-catching number that deserves immediate deflation. It is a framework covering multiple years, not a purchase order; consumer robotics units sell at a fraction of automotive ASPs; and Dreame's own volumes have to materialise for any of it to convert. What the deal genuinely demonstrates is that lidar unit costs have fallen far enough that a sensor once reserved for research vehicles now clears the bill-of-materials hurdle for a garden appliance. That is the cost curve reaching its logical destination, and it opens categories nobody was modelling three years ago.

The robotaxi and autonomous-fleet business is more mature and more strategically valuable. Hesai's customer list across autonomous mobility and robotics spans 百度 Baidu's Apollo Go, 文远知行 WeRide, 小马智行 Pony.ai, DiDi, 宇树科技 Unitree, Neolix, Zelos, and 美团 Meituan.¹ The investment logic is that Hesai does not need any particular robotaxi operator to win; it needs the category to deploy. That is a materially better risk position than a supplier levered to one programme.

Humanoid and quadruped robots sit at the far speculative end of the same thesis — Hesai counts 宇树科技 Unitree among its robotics customers¹ and claims leadership across humanoid, quadruped, robotaxi, robovan, and lawn-mower sub-segments, with recent wins including a 200,000-unit order for Zelos robovans and a design win with two-wheeler maker NIU Technologies.¹² These are worth a sentence rather than a section: the units are small, the customers are early-stage, and the category's economics are unproven.

Two upgrades to credibility

Two 2026 developments upgraded the story's credibility in different ways.

The first was Nvidia. On January 7, 2026, Hesai announced it had been selected as a lidar partner for Nvidia's DRIVE Hyperion 10 — the reference compute-and-sensor architecture Nvidia sells to automakers as a production-ready path to Level 4 capability.²⁸ Being designed into a reference platform is a distribution mechanism rather than a single sale: every automaker that adopts Hyperion inherits Hesai as a validated option. On the Q4 2025 call, CEO Li described this as enabling "a scalable turnkey model" for international expansion.¹³ The irony is unmissable — the same partnership that gives Hesai a global distribution channel is precisely what triggered the U.S. cyber-risk narrative six months later.¹⁹

The second was Mercedes-Benz. Disclosed alongside Q1 2026 results on May 19, 2026, Hesai announced it had become strategic lidar partner and confirmed supplier for Mercedes-Benz models enabling Level 3 autonomy, covering programmes in Europe and China, with production supported by the new Galileo manufacturing centre in Thailand.¹¹²⁵ This is the single most important design win in the company's history for a reason that has nothing to do with volume: it is direct evidence against the "Hesai is a China-only supplier that Western OEMs cannot touch" thesis. A German luxury brand putting a Chinese sensor into a Level 3 system — where the manufacturer, not the driver, carries liability when the system is engaged — is a substantive vote of confidence.

Two caveats sit right beside it. Volumes and timing were not disclosed; CEO Li characterised the new contracts on the Q1 call as "a continuation of the existing collaboration with additional volumes and car models."¹² And the Thailand plant that is supposed to serve Western customers from outside China is not yet running — the company said the Bangkok Galileo facility was expected to commence production in early 2027.²⁸ The geopolitical hedge is real, but it is a construction project first and a hedge second.

The most speculative leg is what management calls Strategic Growth Initiatives. In April 2026 Hesai unveiled Picasso, described as the world's first 6D full-colour ultra-sensitive lidar SPAD-SoC — a chip that captures colour and spatial data together, with launch expected in the second half of 2026 — and Kosmo, a spatial-intelligence device pairing lidar with AI algorithms and targeting recurring revenue across robotics simulation, immersive media, and 4D entertainment.¹¹ In plain terms: instead of selling a sensor that reports distances, sell a device that produces a continuously updated, photorealistic 3D model of a physical space, and charge for the data pipeline rather than the hardware.

If that works, it changes the margin profile of the company. CFO Fan told analysts that Kosmo "should carry a structurally higher margin profile because it integrates hardware, software and data workflows," and guided to roughly RMB 100 million of SGI revenue in 2026, beginning in the second quarter, scaling toward RMB 500 million in 2027.¹²

Investors should size that honestly. RMB 100 million against 2025 revenue of RMB 3.0 billion is about three percent — a rounding error dressed in a strategy. The guidance is useful precisely because it is specific and near-term: it is falsifiable within two quarters. If SGI revenue shows up on schedule and is disclosed separately, the "beyond hardware" narrative earns a hearing. If it slips or gets folded into an aggregate line, that is information too.

The market's own verdict on the Q1 package was notably cool. Despite announcing the Mercedes win and a return to quarterly profitability in the same release, coverage noted the stock down 15.8% over the period, with commentary pointing to margin pressure from aggressive pricing and capital-intensive capacity expansion as the dominant concern.³⁰ When a company delivers its best strategic news in years and the shares fall, the market is telling you which variable it is actually underwriting. It is not design wins. It is gross margin.

↑ Back to Top

X. Current Management: Incentives, Credibility, and Capital Allocation

David Li has been chief executive since inception. He became chairman of the board only in September 2025, coinciding with the Hong Kong listing.² The executive board today comprises the three founders plus 杨彩莲 Cailian Yang, vice president of operations, with three independent non-executive directors — 张毅 Yi Zhang, 任佳 Jia Ren, and 王晖 Hui Wang.³¹

Control is concentrated by design. Hesai runs a dual-class structure in which Class A shares carry ten votes and Class B one. As of March 31, 2026, the founders together beneficially owned all outstanding Class A shares plus a small Class B holding — approximately 17.4% of total share capital but 67.6% of voting power.² Minority shareholders own most of the economics and almost none of the votes. That is standard for founder-led Chinese technology listings and it is not, by itself, a scandal. It does mean the governance check on this management team is disclosure and market pressure, not the ballot.

Which makes the behaviour worth examining closely.

The CFO seat. On May 13, 2024 — the same day the company filed its lawsuit against the Department of Defense — Hesai announced that Global CFO and board director Louis T. Hsieh had resigned "for personal and family reasons, effective today," with finance director Norman Fan taking over on an interim basis.²⁹ Hsieh was not an unknown quantity; he was one of the more experienced U.S.-market CFOs in Chinese ADR land. An abrupt, same-day departure of a credentialed CFO on the day of a major litigation filing is the kind of coincidence that a sceptical investor is entitled to notice, while acknowledging that the stated reason may be exactly the true one.

The permanent replacement is where it gets more interesting. Peng "Andrew" Fan has served as CFO since November 2024. Immediately before joining, he was chief financial officer of Seyond Holdings — formerly Innovusion — from May 2021 to September 2024.² Seyond is a direct competitor and, on the Chinese installation data, the fourth-largest lidar supplier in the market.¹⁰ Hiring your rival's finance chief is either a shrewd acquisition of competitive intelligence or a signal about how small the pool of qualified lidar CFOs really is. It is not a red flag. It is a detail worth knowing.

The related-party question. This is the item that a short seller would lead with, and it barely surfaced in the sell-side conversation.

Hesai's 2025 accounts disclose a company called Sharpa — an early-stage AI robotics business described as "an associate of our founders," which Li has served as a director of since January 2026.² The transaction sequence is unusual. In April 2025, Hesai indirectly subscribed for roughly 10% of Sharpa for $13.9 million, about RMB 100.0 million. In August 2025 — four months later — Hesai sold that entire stake to two independent third-party investors for $38.4 million, about RMB 275.6 million, booking a net gain after taxes and expenses of roughly $20.7 million, or RMB 148.0 million, recorded in other income. In May 2025, Hesai also agreed to transfer certain internally generated know-how to Sharpa for RMB 37.0 million, valued by an independent appraiser; the transfer completed in August 2025 and the RMB 37.0 million was booked as a gain in other operating income. In December 2025, Hesai transferred property to Sharpa for RMB 4.0 million at net book value.²

Add the two gains and roughly RMB 185 million of 2025 pre-tax income came from transactions with a company the founders are associated with. Against GAAP net income of RMB 435.9 million,¹ that is a large enough share that it materially changes how one reads the profitability milestone. The gains were disclosed, appraised where required, and the stake was sold to independent third parties at a price those parties chose to pay — this is not an allegation of impropriety. But it is a quality-of-earnings observation with real weight: the operating business is less profitable than the headline suggests, and an investor treating FY2025 net income as a clean run-rate is overstating the base. It also raises the structural question that always accompanies founder-adjacent ventures — when the founders build an AI robotics company on the side, who decides which opportunities belong to the listed entity and which to the affiliate?

Capital allocation. The record here is genuinely disciplined and deserves to be said plainly. Hesai has made essentially no acquisitions. The capital has gone into manufacturing: Maxwell, Hertz, and now Galileo in Thailand, leased in May 2025 and under renovation through the 2025 annual report.² Announced capacity was to double from roughly 2 million units in 2025 to 4 million in 2026.²⁸ This is a company that says it is a manufacturer and then spends like one. Compared with the far more common pattern — promise discipline, then buy something expensive and unrelated — that consistency is a credit.

The open item is shareholder returns. The buyback mandate exists; execution does not yet.²⁴³¹ With a large net cash position and a share price that has punished the company for margin optics, the coming quarters will reveal whether the mandate was a genuine capital-allocation tool or a box-ticking exercise at a Hong Kong listing.

Guidance discipline. Management has been specific in a way that makes it testable, which is the right behaviour. It guided to 3–3.5 million lidar units for full-year 2026 on the Q4 2025 call, reaffirmed that range on the Q1 2026 call, and guided Q2 2026 revenue to RMB 850–900 million, implying 20–27% year-over-year growth, on roughly 650,000 units.¹²¹³ Note what that arithmetic implies: hitting even the bottom of the annual range requires a dramatic second-half acceleration. The company will report second-quarter results on August 18, 2026, following a board meeting the same day.³¹ That print, two weeks from now, is the first hard checkpoint on whether the 2026 guidance was ambitious or promotional.

On tone: Li's answers to competitive questions have been consistently confident and consistently unspecific — "We welcome competition. We remain confident in our long-term structural advantages and our ability to continue launching category-defining products," he told analysts on the Q1 call.¹² Geopolitics went unraised by management on that call.¹² Investors should read that as management choosing to route the political story entirely through legal filings and the risk-factor section rather than the earnings narrative. That is a defensible communications strategy. It is also an incomplete one, given that the risk factors themselves acknowledge disrupted customer relationships.²

↑ Back to Top

XI. Bull Case vs. Bear Case

Strip away the narrative and the disagreement reduces to a single question: is Hesai a scale manufacturer whose cost lead compounds, or a commodity component supplier in a price war with a political target on its back?

The bull case rests on evidence rather than aspiration, which is unusual for a company at this stage. Hesai is first in long-range ADAS lidar globally and has been for five consecutive years.⁸ It is the only company in the category to have produced a full year of GAAP profit, while its nearest domestic rival reached its first profitable quarter only at the end of 2025 and runs at fifteen points lower gross margin.¹²⁶ Customer concentration has fallen every year even as revenue compounded, and the business absorbed the loss of a customer that was once more than a quarter of revenue without breaking stride.² The Mercedes Level 3 selection is direct evidence that the "China-only" ceiling is not absolute.²⁵ The Nvidia reference-platform position converts individual selling into channel distribution.²⁸ The Dreame framework demonstrates that the cost curve has opened consumer categories.²⁷ And the dual listing removes the most extreme financing and tradeability tail risk.²²

The bear case is equally concrete. Gross margin has now declined in consecutive periods, and management's "mix, not price" explanation cannot yet be verified.¹¹¹¹³ The DoD designation has been upheld by a federal court on reasoning — dual-use technology, Chinese location, cooperation with domestic institutions — that no corporate action can remedy, with procurement prohibitions that began taking effect on June 30, 2026 and a further tranche in 2027.²¹⁷ There is an active senatorial campaign for outright Nasdaq delisting.¹⁸ A securities class action has partially survived a motion to dismiss.² A short-seller report was met with a two-sentence denial rather than a point-by-point rebuttal.²⁰ Roughly RMB 185 million of 2025 income came from transactions with a founder-associated entity.² Huawei, not Hesai, led Chinese passenger-vehicle lidar installations in 2025 and is escalating specifications aggressively.¹⁰ And the entire demand base depends on Chinese EV manufacturers continuing to spend on driver-assistance differentiation during a domestic price war — a spending pattern with no guarantee of persistence.

The activist stress test. What would a hostile investor actually attack? Four things. First, disclosure quality: the gap between a two-sentence short-seller response and the detailed related-party and contract-termination disclosures that later appeared in the annual report invites the question of why the latter was not offered when it mattered. Second, related-party architecture: a founder-associated AI robotics venture that receives Hesai know-how, buys Hesai property, and generates a large one-year gain for Hesai deserves a clearer articulation of the corporate-opportunity boundary. Third, insider selling: planned sales of roughly $62.58 million filed days after a 33% news-driven rally and days before a short report is the kind of sequence that invites scrutiny regardless of pre-arranged plan status.²¹ Fourth, the buyback gap: authorisation without execution, in a net-cash company whose stock fell on good news, is exactly where an activist would apply pressure.

The honest synthesis. Hesai has a credible, evidence-backed answer to "why do we win the sensor": lowest cost at scale, deepest patent position, broadest design-win base, first to profit, and now a Western premium-OEM validation. The evidence supports that claim more strongly than it supports most such claims in emerging hardware.

Hesai does not have a credible answer to "why do we win the geopolitics," because it may not be a question a company can answer. The designation survived judicial review on grounds that are effectively structural. The political pressure has broadened rather than narrowed. And a business whose entire international expansion thesis runs through Western OEMs and American reference platforms is exposed to a variable it does not control and cannot hedge beyond building factories in Thailand and listing shares in Hong Kong — both of which it has now done.

An investor's view of this stock is, to an unusual degree, a view on that single unresolved question.

↑ Back to Top

XII. What to Watch: KPIs, Risk Radar, and the Next 12–24 Months

If an investor tracked only three things about Hesai, these would be the three.

Gross margin trajectory, quarter by quarter. This is the whole argument compressed into one line. The sequence — 42.6% for 2024, 41.8% for 2025, 39.1% in the first quarter of 2026 — is the bear case's strongest evidence and the bull case's biggest vulnerability.¹¹¹ Management's claim is that mix, not price competition, is driving it and that declines will narrow as cost structure improves.¹³ Stabilisation, or a return above 40% while volumes grow, would validate the scale-economy thesis. Continued erosion would suggest that in a three-player Chinese price war, being the low-cost producer buys share rather than profit.

Unit shipments and implied ASP by segment. Shipments are the cleanest read on demand, and the ADAS/robotics split — disclosed each quarter — lets an investor infer where pricing is holding.¹¹¹ Robotics carries better economics and is growing faster; the faster it grows as a share of the mix, the more support there is for margins. Against the guided 3–3.5 million units for 2026, the quarterly cadence is directly checkable.¹²

SGI revenue disclosure. Management has committed to roughly RMB 100 million in 2026, beginning in the second quarter, scaling toward RMB 500 million in 2027.¹² Whether that appears as a separately disclosed line, and whether it hits, is the test of the "beyond hardware into software and data" claim. Narrative-to-revenue conversion within four quarters is a rare and useful check.

The risk radar, restricted to mechanisms that actually apply here.

Regulatory and political escalation is the dominant one. The transmission is not a revenue line; it is the willingness of Western customers and platform partners to commit to a supplier carrying a federal designation, plus the ever-present possibility of export controls, a Treasury listing, or a delisting order. The D.C. Circuit appeal is the near-term binary.²¹⁸

Margin compression from China's EV price war. Automakers under margin pressure push component prices down. Hesai's defence is cost position; its exposure is that its two largest rivals are also well-capitalised and equally willing to trade margin for share.¹⁰²⁶

Execution risk in Thailand. Galileo is the physical embodiment of the geopolitical hedge and the production base for Mercedes programmes, and it was expected to start production in early 2027.²⁸ Any slippage delays exactly the revenue that most needs to arrive to prove the non-China thesis.

Customer-programme concentration. Although top-five concentration has fallen, one Chinese new-energy manufacturer still represented 14.7% of 2025 revenue, down from 33.7% in 2024 — a reminder that individual vehicle programmes can move the P&L.²

Litigation overhang. The partially surviving securities class action carries an undetermined outcome and cost, which the company states it cannot estimate.²

What would change the case fastest. On the bull side: a D.C. Circuit ruling vacating or narrowing the designation; gross margin stabilising above 40%; buybacks actually executed under the March 2026 mandate; SGI revenue reported and material. On the bear side: further U.S. restrictions extending beyond defence procurement into commercial sales; a second consecutive quarter of margin decline paired with a shipment miss against the 3–3.5 million guidance; or Thailand slipping past early 2027.

The next data point arrives on August 18, 2026.³¹

↑ Back to Top

XIII. Playbook: Durable Business and Investing Lessons

Manufacturing beats physics, eventually. Hesai did not out-invent the American lidar cohort. It licensed Velodyne's patents rather than defeating them.³ What it did was build the automated capacity to make the thing cheaply before the demand arrived, and then meet demand at a price nobody else could match. In hardware categories that are transitioning from research instrument to mass component, the durable winner is usually the company that solved manufacturing, not the one that published the best spec sheet. That is a screenable pattern.

"First to profitability" is a real signal — and a claim to audit. Being first to profit in a capital-intensive young category genuinely tells you something about relative cost position. But the first year of profitability is also the year most vulnerable to flattering one-offs. In Hesai's case, a contract-termination payment supported the 2024 non-GAAP result² and a related-party stake disposal plus a know-how transfer contributed roughly RMB 185 million to 2025 income.² The discipline is simple: before treating a maiden profit as a run-rate, read the other income line and the related-party note.

Geopolitical risk in Chinese technology names is a standing cost, not a tail event. The most useful thing about the Hesai case study is how the risk actually manifested — not as a single ban, but as a designation, then a re-designation on a new record, then a lawsuit, then an adverse ruling on structural grounds, then a delisting campaign, then a cyber-risk framing, then procurement prohibitions on a fixed calendar. It expressed itself in valuation multiple, in financing structure (hence a $614 million Hong Kong raise), in factory location (hence Thailand), and in the company's own admission that customer relationships were disrupted.²²²²⁸ It should be modelled as an ongoing operating variable with a cost, not as a probability-weighted one-time shock.

Diversify the demand base, but size each leg by economics, not by narrative appeal. Hesai's spread across passenger-car ADAS, robotaxis, delivery robots, humanoids, and lawn mowers is genuinely more resilient than a bet on any single application reaching scale. The trap is letting the most exciting leg dominate the mental model. Robotics was about 15% of 2025 units,¹ and Strategic Growth Initiatives are guided at roughly three percent of revenue in 2026.¹² Optionality is worth paying something for. It is not worth paying for as though it had already arrived.

A licence you pay for can be worth more than a lawsuit you win. In 2020, paying an incumbent a settlement fee plus royalties through 2030 looked like a tax on a challenger.³ Five years later that agreement was the mechanism that terminated an existential trade-barrier case and produced a $6.4 million cost award in Hesai's favour.² In IP-dense hardware, the cheapest form of insurance is often a licence signed early, when the fee is small because the licensee is small.

Watch what an authorisation becomes. A repurchase mandate, a capacity announcement, and a framework supply agreement are all permissions rather than performance. The interval between authorisation and execution is one of the most information-rich things a public company produces, and it is free to observe.

↑ Back to Top

XIV. Outro

There is a particular kind of company that arrives at the top of its industry precisely when the ground beneath the industry starts moving.

Hesai today is the cost and technology leader in a component the entire global automotive industry has converged on. It sells to forty automotive brands across more than 160 models,⁸ to nearly every serious robotaxi operator on earth,¹ to a German luxury marque betting its Level 3 liability on the sensor,²⁵ and to a Chinese appliance maker putting laser rangefinders into lawn mowers.²⁷ It has done what none of its better-funded Western predecessors managed: make money doing it.¹

And it does all of this while formally designated by the United States government as a Chinese military company, with a federal court having upheld that designation on grounds the company cannot cure through any change in its own conduct.²¹⁷

That is the tension to hold. Not "is the technology good" — the design wins settle that. Not "can they manufacture" — 1.6 million units and a 41.8% gross margin settle that.¹ The question is whether a company can hold a global leadership position in a safety-critical sensor that Western regulators are increasingly inclined to treat as an instrument of state, and whether the margin structure survives a three-way Chinese price war long enough for the robotics and spatial-intelligence bets to become material.

Two of those three variables are measurable every ninety days. The third is decided by a court, a Congress, and an executive branch that have so far shown no interest in Hesai's engineering.

Hesai Group

Table of Contents

Hesai Group: Betting the World's Cars See in 3D

I. Cold Open & Episode Roadmap

II. What Is Lidar, and Why Does This Industry Exist

III. Founding and the First Decade: Surviving as a Chinese Hardware Startup (2014–2022)

IV. The Nasdaq IPO and the U.S. Lidar Shakeout (Feb 2023)

V. The Inflection: Winning China's ADAS Volume War and Reaching Profitability (2023–2025)

VI. Geopolitics as a Structural Risk, Not a Footnote: The Pentagon Blacklist and the Short-Seller Fight

VII. The Hong Kong Dual-Primary Listing: Insurance Policy or Vote of No Confidence? (Sept 2025)

VIII. Competitive Landscape: Who Else Is in This Race, and Who's Actually Winning

IX. The Next Act: Robotics, Robotaxis, and "Spatial Intelligence" as the Real Optionality

X. Current Management: Incentives, Credibility, and Capital Allocation

XI. Bull Case vs. Bear Case

XII. What to Watch: KPIs, Risk Radar, and the Next 12–24 Months

XIII. Playbook: Durable Business and Investing Lessons

XIV. Outro

References