Hesai Group
A Chinese maker of the laser scanners that let cars and robots see in three dimensions, shipping more of them than any listed rival.
The scores
Growth3/4 ↓
Units rose 78.4% in Q2 and revenue only 21.9%, so the engine adds volume faster than it adds money; and 2024 grew just 10.7%.
Business quality3/4 →
A cost position its rival's own accounts confirm, and no pricing power - the price it charges fell a third in a year.
Financial position4/4 →
Net cash worth about a third of the market value, trivial borrowings, no maturity wall, and no equity raised this year to fund anything.
Execution record3/4 →
Revenue landed inside its own guided range and unit guidance was reiterated; against that, three consecutive large misses against a consensus it does not guide.
Per share capture3/4 →
Share count up about 7.7% in 2025 against revenue up 45.7%, no buyback, and a dual-class structure that makes issuance costless to control.
Potential2/4 →
Design wins carry no volume, term or price, and the growth story everyone is excited about is capped by a related-party agreement.
Risk profile2/4 ↑
A national-security designation that a court has sent back rather than cancelled, plus five customers at 55.8% and a related-party growth engine.
Valuation3/4 →
About 3.9 times revenue, a third of the market value in cash, needing 13.5% growth - but operations earn nothing.
Evidence quality3/4 →
Everything load-bearing is filed. The largest item in 2025's profit, first read as unexplained, is a one-off gain on the Sharpa stake.
If it works 1.2× · if it fails you keep about 100%.
The Verdict
We rate Hesai neutral. It is a Chinese maker of the laser scanners that let cars and robots see in three dimensions, and it won the price war, shipping more than any listed rival at a better margin, so the business is solid. But winning has not paid yet: without government grants it has never made a yearly operating profit. After counting its large cash pile, the price is demanding but not absurd, so it is fair. As a bet it is defensive: its young robot business might add a little if it works, and if it fades shareholders still own the scanner leader and its cash.
The long caseIf humanoid robots reach factory volumes, each needs dozens of joint motors, and Hesai can sell its actuators to robot makers beyond the one related customer it serves now.
The evidence under this - 8 claims
- The two mixes are not identical: RoboSense's 2025 robotics share of product revenue was about 49%, and robotics sensors carry higher prices than driver-assistance ones, so a like-for-like comparison would need a split neither company publishes.
- The gap is also narrowing from Hesai's side rather than widening: its own margin fell 2.4 points year on year while RoboSense's rose.
- The trend is the point that runs the other way: the loss excluding grants narrowed from RMB598.1M to RMB12.7M in two years, which is a RMB585M improvement and is not an accounting effect.
- Q4 2025 did earn RMB41.8M excluding the line, on revenue of RMB1,000.5M, so at the scale the second half now reaches the business does produce an operating profit without help.
Interest on a genuinely owned RMB7.05B of cash is real income that accrues to shareholders, not an accounting trick, and a company with net cash worth a third of its market value should be expected to earn it. What it is not is evidence about the lidar business.
- The share count is the one at 31 March and predates the July subdivision, whose ADS ratio change leaves the traded unit unchanged; if any shares have been issued since, the market value is understated.
- The four quarterly revenue figures are convenience translations struck at four different exchange rates, so the $480.1M is not a figure any single document states.
Three of the four inputs are mine. At a 25 times exit the required growth falls to 8.5% and at 15 times it rises to 20.2%; at a 10% required return every figure drops further. What does not move is the comparison that matters: the same arithmetic applied to Ouster on the same day needs 38.4% compound growth, because Ouster starts from about a third of the revenue at more than twice the enterprise value.
The raise is attributed to demand from Sharpa, a company Hesai's chief executive co-founded, and the related-party cap being raised alongside it is RMB300M - the whole of the new guidance range. A guidance increase whose stated driver is a counterparty the chief executive co-founded is not the same evidence as one driven by arm's-length orders.
- The governance around it is real and documented, not nominal: cost-plus pricing benchmarked by a Big Four transfer-pricing study, an independent board committee, an independent financial adviser recommending approval, and a vote from which the interested parties are excluded. That is a stronger check than most related-party revenue gets.
- The cap is a ceiling rather than an order, and RMB42M was actually transacted in the first three months - so the RMB300M figure describes headroom management asked for, not revenue it has booked.
- 2026-09-27editedC23's meeting was held: the poll result (S17) replaces 'subject to approval', and the filed interim report (S18, note 15) adds the half-year revenue from Sharpa and the receivable due from it at 30 June 2026
The company
A lidar fires laser pulses and times their return to draw a three-dimensional picture of what is around it. Hesai builds them in its own factories in China and Thailand, in two shapes: a slim bar for Chinese carmakers' driver-assistance systems, 485,904 in the June 2026 quarter, and a spinning unit for robotaxis, delivery vans and lawn-mowing robots, 142,371.
In a car it replaces nothing; it is added beside cameras and radar. What it competes with is the carmaker deciding not to fit one.
The evidence under this - 3 claims
The penetration figures are management's with no cited source, and a doubling of penetration inside one year is the single most convenient assumption available to a lidar seller. The regulatory date is checkable in principle and was not verified against the standard itself here.
How it makes money
It sells sensors one at a time, and the price falls fast: about $1,100 on average in 2023, $530 in 2024 and $260 in 2025. Hesai sells ever more of them, and still keeps about 40 cents of each sales dollar after the cost of making the sensor.
The evidence under this - 2 claims
Where the money comes from, stream by stream
| Stream | Share of sales (FY2025) | How it earns it |
|---|---|---|
| Lidar products | 98.2% | Laser sensors sold to Chinese carmakers for driver assistance and to robot makers for robotaxis, lawn mowers and delivery vehicles. |
| Engineering services | 1.2% | Paid design, development and validation work for carmakers adapting a sensor; shrinking as one-off projects end. |
| Other products and services | 0.6% | Other hardware and services, including robotic actuation modules, which reached about 2.9% of revenue in the first half of 2026. |
Who runs it
Yifan Li co-founded Hesai and has run it since it started in 2014; his two co-founders lead its science and technology, and the finance chief, Peng Fan, joined in 2024 from a rival lidar maker. The founders own under a fifth of the shares but about two thirds of the votes, through shares that carry ten votes each. The filings name no target their pay rides on: their shares simply vest with time. All three also sit on the board of Sharpa, a robot maker that buys Hesai's parts.
The evidence under this - 5 claims
- The governance around it is real and documented, not nominal: cost-plus pricing benchmarked by a Big Four transfer-pricing study, an independent board committee, an independent financial adviser recommending approval, and a vote from which the interested parties are excluded. That is a stronger check than most related-party revenue gets.
- The cap is a ceiling rather than an order, and RMB42M was actually transacted in the first three months - so the RMB300M figure describes headroom management asked for, not revenue it has booked.
- 2026-09-27editedC23's meeting was held: the poll result (S17) replaces 'subject to approval', and the filed interim report (S18, note 15) adds the half-year revenue from Sharpa and the receivable due from it at 30 June 2026
Why it wins, and who could beat it
Why customers come back
Hesai owns its factories, designs its own receiver chip and ships more than anyone, so its cost per sensor falls faster than its price. That lets it lead prices down and still make money on each unit.
The proof comes from a rival's accounts. RoboSense, the nearest listed competitor, earns a 26.5% gross margin against Hesai's 40.1%, on about half the units.
The evidence under this - 3 claims
- The two mixes are not identical: RoboSense's 2025 robotics share of product revenue was about 49%, and robotics sensors carry higher prices than driver-assistance ones, so a like-for-like comparison would need a split neither company publishes.
- The gap is also narrowing from Hesai's side rather than widening: its own margin fell 2.4 points year on year while RoboSense's rose.
The threat
RoboSense says it sold more robot sensors than anyone in 2025. Some of Hesai's own customers have said they will build their own. And the biggest rival is still no laser at all: about four in five Chinese electric cars were sold without one in 2025, on management's own figure.
The evidence under this - 3 claims
The penetration figures are management's with no cited source, and a doubling of penetration inside one year is the single most convenient assumption available to a lidar seller. The regulatory date is checkable in principle and was not verified against the standard itself here.
The rivals: RRX, INVZ, OUST, MVIS, AEVA, LIDR, and 4 more that are private, in-house or listed abroad.
Who else does this, rival by rival
| Ticker | Rival | Where it competes | How far along | Priced against it |
|---|---|---|---|---|
| RoboSense TechnologyPriced against it: Listed abroad | Lidar | On sale | Listed abroad | |
| Seyond, formerly InnovusionPriced against it: Listed abroad | Lidar | On sale | Listed abroad | |
| In-house | Hesai's own customers, building in housePriced against it: In-house | Lidar | Announced | In-house |
| Not listed | Camera-and-radar driver assistance, fitted instead of lidarPriced against it: Not listed | Lidar | On sale | Not listed |
| RRX | Regal RexnordPriced against it: One segment only | Strategic Growth Initiatives | On sale | One segment only |
| INVZ | Innoviz TechnologiesPriced against it: Yes | - | On sale | Yes |
| OUST | OusterPriced against it: Yes | - | On sale | Yes |
| MVIS | MicroVisionPriced against it: Yes | - | On sale | Yes |
| AEVA | Aeva TechnologiesPriced against it: Yes | - | On sale | Yes |
| LIDR | AEyePriced against it: Yes | - | On sale | Yes |
How big it could get
How big this could get
The two largest listed makers shipped about 2.53 million sensors in 2025, and Hesai made 64% of them. At about $260 each that is only about $660M of revenue a year between them, less than half Hesai's market value.
So growth has to come from more sensors per car. China's mandatory self-driving safety standards take effect in July 2027, and new models already carry three to six.
The evidence under this - 5 claims
Adding two companies' units is not a market measurement: it excludes every private and non-listed supplier, and it double-counts nothing but also cannot say what share the two hold. No trade body publishes a unit total that could check it - the rankings that exist are cited by the suppliers themselves.
The penetration figures are management's with no cited source, and a doubling of penetration inside one year is the single most convenient assumption available to a lidar seller. The regulatory date is checkable in principle and was not verified against the standard itself here.
Who has signed
Great Wall, Li Auto, GAC Toyota and Volkswagen's Chinese ventures have chosen Hesai for new models, but none of these wins carries a disclosed volume. The one written agreement is with Sharpa, a robot maker co-founded by Hesai's chief executive. It bought RMB44.9M in the first half, and in August outside shareholders approved a cap on 2026 sales of RMB300M.
The evidence under this - 2 claims
Not one carries a disclosed volume, term or contract value, and a design win is an award to be a supplier rather than an order. Hesai's own risk factors describe carmakers' ability to re-source or cancel vehicle programmes, and one prior customer terminated a contract, for which Hesai received compensation.
- The governance around it is real and documented, not nominal: cost-plus pricing benchmarked by a Big Four transfer-pricing study, an independent board committee, an independent financial adviser recommending approval, and a vote from which the interested parties are excluded. That is a stronger check than most related-party revenue gets.
- The cap is a ceiling rather than an order, and RMB42M was actually transacted in the first three months - so the RMB300M figure describes headroom management asked for, not revenue it has booked.
- 2026-09-27editedC23's meeting was held: the poll result (S17) replaces 'subject to approval', and the filed interim report (S18, note 15) adds the half-year revenue from Sharpa and the receivable due from it at 30 June 2026
The results
The numbers
Sales grew 46% in the year to Dec 2025, faster than the 11% in the year before.
Operating margin is 5.6%, up from −31% in the year to Dec 2022.
Running costs take 36¢ of each sales dollar, down from 90¢ in the year to Dec 2021.
Shareholders' money earned 6.8% a year, up from −133% in the year to Dec 2023.
It used CNY 192M of cash in the latest year, so the CNY 4.76B it holds lasts about 24.7 years at that pace.
It holds CNY 4.76B of cash against CNY 727M of debt, CNY 4.03B more cash than it owes.
The share count rose 39% in 4.0 years, and pay in shares costs 3.8% of sales.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-04. Fiscal years end in December. In CNY.
| Dec 2021 | Dec 2022 | Dec 2023 | Dec 2024 | Dec 2025 | |
|---|---|---|---|---|---|
| Size and growth | |||||
| Sales | CNY 721M | CNY 1.20B | CNY 1.88B | CNY 2.08B | CNY 3.03B |
| Sales growth | +73% | +67% | +56% | +11% | +46% |
| Profit | |||||
| Gross margin | 53% | 39% | 35% | 43% | 42% |
| Operating margin | −37% | −31% | −30% | −9.9% | 5.6% |
| Net margin | −34% | −25% | −25% | −4.9% | 14% |
| What growth costs, per sales dollar | |||||
| Research and development | 51% | 46% | 42% | 41% | 26% |
| Sales and marketing | 9.6% | 8.7% | 7.9% | 9.3% | 6.3% |
| General and admin | 33% | 17% | 17% | 15% | 9.5% |
| Pay in shares | 7.5% | 8.7% | 13% | 5.6% | 3.8% |
| Capital spending | 31% | 19% | 22% | 12% | 10% |
| Cash | |||||
| Cash from operations | −CNY 228M | −CNY 696M | CNY 57.3M | CNY 63.5M | CNY 117M |
| Free cash flow | −CNY 448M | −CNY 927M | −CNY 349M | −CNY 196M | −CNY 192M |
| Balance sheet | |||||
| Cash and short-term investments | CNY 2.79B | CNY 1.86B | CNY 3.14B | CNY 3.20B | CNY 4.76B |
| Debt | - | CNY 18.5M | CNY 398M | CNY 615M | CNY 727M |
| Net cash | - | CNY 1.84B | CNY 2.74B | CNY 2.59B | CNY 4.03B |
| Returns and shares | |||||
| Return on capital | - | - | - | - | 2.2% |
| Shares (diluted) | 105M | 116M | 125M | 129M | 146M |
What growth costs
Growth has made Hesai cheaper to run: research, selling and office costs barely moved while sales grew, so each sale carries far less overhead than three years ago. The catch is that each sensor now earns less, and new robot projects add costs, so this year the savings have mostly been given back.
The evidence under this - 3 claims
- 2023's share-based pay included one-off option expense on the US listing, so part of the fall in that line is the base year rather than leverage (S01, Item 5).
- Q2 2026 research spending rose 16.0% on new robotics work (C66).
Excluding other operating income Hesai still lost money at the operating line in 2025 (C32), so the leverage has not yet produced an operating profit of its own.
The last report (18 Aug)
Quarter ended June 2026, filed figures.
Units grew 78.4% in the June 2026 quarter, but revenue only 21.9%, to RMB860.8M, because each sensor sold for about 32% less. Gross margin slipped to 40.1% from 42.5%.
Profit told the same story. Operating income fell 90.4% to RMB2.2M, while net income rose 60.0%, all of it from interest and investment income. For the September quarter the company expects revenue up 38% to 45%, with new lines near half of it.
What it did to the thesisIt is the clearest evidence for both sides of the view. Hesai keeps winning volume, but prices fell faster than volume rose, operating profit nearly vanished and all the net profit came from interest. That is why the view is Neutral: cheap, for a real reason.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $0.01 against $0.07 expected.
The evidence under this - 7 claims
- The mix explains part of it by construction: driver-assistance sensors are the cheaper product and they grew 60.1% while robotics grew 193.4%, so a falling average price is partly the arithmetic of a shifting mix rather than price cuts on any one product.
- Hesai's own guidance says robotics and other non-automotive revenue should approach half of total in Q3, which would push the average the other way if the mix explanation is right.
Interest on a genuinely owned RMB7.05B of cash is real income that accrues to shareholders, not an accounting trick, and a company with net cash worth a third of its market value should be expected to earn it. What it is not is evidence about the lidar business.
- The implied sequential step is about 30% on revenue, larger than any the company has guided recently, and it rests on a seasonality argument rather than on a disclosed order book - there is no backlog figure anywhere in the filings.
- The half-of-revenue claim for non-driver-assistance is the one that carries the mix argument for gross margin, so the same guide is doing double duty: if it lands, the margin question resolves favourably; if it slips, both readings slip together.
What management did with the money
It has kept its cash, about $1.04B in June against small borrowings, but the pile is shrinking: RMB7.51B to RMB7.05B in six months. Average shares rose from 124.8M in 2023 to 139.1M in 2025. The founders hold 67.6% of the votes on 17.4% of the shares.
The evidence under this - 4 claims
The price, and where we disagree
In short. Fair. Cheap for a winner, but the cheapness has a reason: without government grants, Hesai has never made a yearly profit from running its business.
The shares are 41.2% below their one-year high, down 21.3% over a year in which the business grew. Two fears hang over it: an American news report in July calling the sensors a cyber risk, and the Pentagon's listing of Hesai as a Chinese military company.
In August an appeals court sent that listing back for a fairer process, without clearing Hesai. The next test is the September-quarter results in November.
The evidence under this - 4 claims
How we read the multiple
Hesai holds about $945M of cash, a third of its market value. Take that off and buyers pay about $1.89B for the business, about 4 times a year's sales; a data vendor, not taking the cash off, says about 5.1.
To return 12% a year for five years, Hesai must:
- Reach about $1.11B of sales in 2031.
- Keep 15 cents of operating profit on each dollar of sales.
- That is 13.5% growth a year: demanding without grants, but not absurd.
Against its own history. Far cheaper than its own normal on sales. That normal was set while it lost money, so the discount reads as fear rather than decline.
The evidence under this - 3 claims
- The share count is the one at 31 March and predates the July subdivision, whose ADS ratio change leaves the traded unit unchanged; if any shares have been issued since, the market value is understated.
- The four quarterly revenue figures are convenience translations struck at four different exchange rates, so the $480.1M is not a figure any single document states.
Three of the four inputs are mine. At a 25 times exit the required growth falls to 8.5% and at 15 times it rises to 20.2%; at a 10% required return every figure drops further. What does not move is the comparison that matters: the same arithmetic applied to Ouster on the same day needs 38.4% compound growth, because Ouster starts from about a third of the revenue at more than twice the enterprise value.
- The trend is the point that runs the other way: the loss excluding grants narrowed from RMB598.1M to RMB12.7M in two years, which is a RMB585M improvement and is not an accounting effect.
- Q4 2025 did earn RMB41.8M excluding the line, on revenue of RMB1,000.5M, so at the scale the second half now reaches the business does produce an operating profit without help.
Priced against its rivals
The competitors its own annual report names, each confirmed against the rival's filing. Measured on company value for each $1 of sales, cheapest first:
| Ticker | Company | company value for each $1 of sales |
|---|---|---|
| INVZ | Innoviz Technologies | 2.0× |
| HSAI | Hesai Group | 5.1× |
| OUST | Ouster | 12.2× |
| MVIS | MicroVision | 20.2× |
| AEVA | Aeva Technologies | 54.2× |
| LIDR | AEye | 102.9× |
The middle of the other 5 is 20.2. Hesai Group is 2nd cheapest of 6.
Rivals' figures come from a data provider, so read the spread of the group rather than any single pair.
Far below the middle of the listed lidar makers, though it is the only one with an operating profit, so the discount is Hesai's own and not for weakness.
Private or foreign rivals that cannot be priced: RoboSense Technology Co., Ltd., Seyond (formerly Innovusion), Koito Manufacturing Co., Ltd., Luminar Technologies, Inc., Valeo SA and other Tier 1 automotive suppliers.
The bet: what it pays if it works
This bet is on one young business: robot joints, factory services and 3D data, sold mostly to Sharpa, a robot maker the chief executive co-founded. The lidar business and the cash are counted on their own, and they already cover nearly all of today's price. So even management's best figure for 2027 adds little: an expensive bet, but a small one.
If the young business fades, shareholders still own the lidar leader and its cash pile.
| Strategic Growth Initiatives revenue in two to three years | Customers the size of Sharpa's 2026 cap needed | At the stated multiple |
|---|---|---|
| $100.00M (the big win) | about 2 at 100% kept | 1.2× |
| $300.00M (market forecast) | about 7 at 100% kept | 1.4× |
IDC expects over 750,000 humanoids a year by 2030; at 30 hand and arm joints each and a 5% share, Hesai ships 600,000 modules in 2029, about US$300M of revenue. On that forecast, $300.00M pays 1.4× from today's value: an Expensive bet on the market case. Strategic Growth Initiatives would be worth about $600.00M, and the price pays nothing for it.
To pay 2× (a Fair bet) it would need about $1.02B of Strategic Growth Initiatives revenue (about 23 customers the size of Sharpa's 2026 cap); to pay 5× (a Cheap bet), about $4.49B of Strategic Growth Initiatives revenue (about 102 customers the size of Sharpa's 2026 cap), at the same multiple.
Each row is Strategic Growth Initiatives alone, plus the rest of the company (Lidar) counted at about $2.60B.
The steps it has to take
- Done Independent shareholders approve the RMB300M Sharpa cap for 2026
- In progress 2026 Strategic Growth Initiatives revenue lands inside RMB200M to RMB300M due 2027-03
- Not started Related-party revenue stays under a quarter of Strategic Growth Initiatives revenue in 2027 due 2028-03
- Not started About US$100M of Strategic Growth Initiatives revenue in 2027, at breakeven due 2028-03
The detail behind the label
The evidence under this - 5 claims
The raise is attributed to demand from Sharpa, a company Hesai's chief executive co-founded, and the related-party cap being raised alongside it is RMB300M - the whole of the new guidance range. A guidance increase whose stated driver is a counterparty the chief executive co-founded is not the same evidence as one driven by arm's-length orders.
- The governance around it is real and documented, not nominal: cost-plus pricing benchmarked by a Big Four transfer-pricing study, an independent board committee, an independent financial adviser recommending approval, and a vote from which the interested parties are excluded. That is a stronger check than most related-party revenue gets.
- The cap is a ceiling rather than an order, and RMB42M was actually transacted in the first three months - so the RMB300M figure describes headroom management asked for, not revenue it has booked.
- 2026-09-27editedC23's meeting was held: the poll result (S17) replaces 'subject to approval', and the filed interim report (S18, note 15) adds the half-year revenue from Sharpa and the receivable due from it at 30 June 2026
Three of the four inputs are mine. At a 25 times exit the required growth falls to 8.5% and at 15 times it rises to 20.2%; at a 10% required return every figure drops further. What does not move is the comparison that matters: the same arithmetic applied to Ouster on the same day needs 38.4% compound growth, because Ouster starts from about a third of the revenue at more than twice the enterprise value.
- The share count is the one at 31 March and predates the July subdivision, whose ADS ratio change leaves the traded unit unchanged; if any shares have been issued since, the market value is understated.
- The four quarterly revenue figures are convenience translations struck at four different exchange rates, so the $480.1M is not a figure any single document states.
It is the company relaying a vendor's tally, and the same paragraph cites four different research houses across different submarkets, which is the pattern of a company shopping for the ranking that suits each claim. RoboSense's own announcement claims the global number one position in robotics, which Hesai also claims for itself on GGII data; both cannot be right.
Where we disagree with the market
A Chinese national-security problem with collapsing margins, covered by only three analysts.
It won the volume war, but price fell faster than volume rose. The cheapness is real, and so is its reason.
The profit that made headlines is interest and grants, and in 2025 also RMB148.0M from selling a stake in Sharpa. The shares still look cheap once all that is taken out.
The case against us is that the real losses are closing fast. Without grants, the yearly operating loss narrowed from RMB598.1M in 2023 to RMB12.7M in 2025, so the business may be closer to paying for itself than we allow.
What analysts say, and the evidence
The evidence under this - 6 claims
Interest on a genuinely owned RMB7.05B of cash is real income that accrues to shareholders, not an accounting trick, and a company with net cash worth a third of its market value should be expected to earn it. What it is not is evidence about the lidar business.
- 2026-09-27editedAdded the 2025 composition from the 20-F's other operating income note, read while tracing C35: the 'patent transfer' gain was RMB37.0M of know-how sold to Sharpa under a May 2025 agreement
- The share count is the one at 31 March and predates the July subdivision, whose ADS ratio change leaves the traded unit unchanged; if any shares have been issued since, the market value is understated.
- The four quarterly revenue figures are convenience translations struck at four different exchange rates, so the $480.1M is not a figure any single document states.
- The trend is the point that runs the other way: the loss excluding grants narrowed from RMB598.1M to RMB12.7M in two years, which is a RMB585M improvement and is not an accounting effect.
- Q4 2025 did earn RMB41.8M excluding the line, on revenue of RMB1,000.5M, so at the scale the second half now reaches the business does produce an operating profit without help.
- 2026-09-27editedCorrection: the 20-F does explain most of the line - Item 7.B records the RMB148.0M net gain on selling the Sharpa stake in August 2025 - and the thesis had said it explained none of it; found reading note 15(b) of the 2026 interim report (S18)
What could go wrong, and how we would know
The risks
- The price war goes on. Gross profit on each sensor fell about 36% in a year; if that continues, volume never becomes profit.
- Washington. The Pentagon can relist Hesai as a Chinese military company once it fixes the process the court faulted.
- A deal with the boss's company. Sharpa, co-founded by the chief executive, could buy essentially all of the raised 2026 guidance for new lines.
- Customers building their own. Some carmakers and robot makers have announced their own sensors or bought their way in.
- Outside holders have no say. The founders hold 67.6% of the votes, and every figure arrives translated from renminbi.
The evidence under this - 6 claims
- The mix explains part of it by construction: driver-assistance sensors are the cheaper product and they grew 60.1% while robotics grew 193.4%, so a falling average price is partly the arithmetic of a shifting mix rather than price cuts on any one product.
- Hesai's own guidance says robotics and other non-automotive revenue should approach half of total in Q3, which would push the average the other way if the mix explanation is right.
- A cap is set above expected volume by design - it carries an explicit RMB24M buffer for demand, mix and delivery timing - so the two numbers coinciding is what a cap sized to a forecast looks like rather than evidence that the forecast is the related party.
- The company also says Kosmo begins contributing in Q3 and will be low eight-digit revenue this year, which is third-party revenue inside the same guidance range.
- The governance around it is real and documented, not nominal: cost-plus pricing benchmarked by a Big Four transfer-pricing study, an independent board committee, an independent financial adviser recommending approval, and a vote from which the interested parties are excluded. That is a stronger check than most related-party revenue gets.
- The cap is a ceiling rather than an order, and RMB42M was actually transacted in the first three months - so the RMB300M figure describes headroom management asked for, not revenue it has booked.
- 2026-09-27editedC23's meeting was held: the poll result (S17) replaces 'subject to approval', and the filed interim report (S18, note 15) adds the half-year revenue from Sharpa and the receivable due from it at 30 June 2026
What we still do not know
- Whether the fall in profit per sensor is the price war or simply more cheap car sensors in the mix.
- Whether Sharpa, which owed Hesai RMB50.2M at the end of June, pays like any outside customer.
- How much of the new robot business will be sold to the chief executive's own company.
The evidence under this - 4 claims
- The mix explains part of it by construction: driver-assistance sensors are the cheaper product and they grew 60.1% while robotics grew 193.4%, so a falling average price is partly the arithmetic of a shifting mix rather than price cuts on any one product.
- Hesai's own guidance says robotics and other non-automotive revenue should approach half of total in Q3, which would push the average the other way if the mix explanation is right.
- The governance around it is real and documented, not nominal: cost-plus pricing benchmarked by a Big Four transfer-pricing study, an independent board committee, an independent financial adviser recommending approval, and a vote from which the interested parties are excluded. That is a stronger check than most related-party revenue gets.
- The cap is a ceiling rather than an order, and RMB42M was actually transacted in the first three months - so the RMB300M figure describes headroom management asked for, not revenue it has booked.
- 2026-09-27editedC23's meeting was held: the poll result (S17) replaces 'subject to approval', and the filed interim report (S18, note 15) adds the half-year revenue from Sharpa and the receivable due from it at 30 June 2026
- A cap is set above expected volume by design - it carries an explicit RMB24M buffer for demand, mix and delivery timing - so the two numbers coinciding is what a cap sized to a forecast looks like rather than evidence that the forecast is the related party.
- The company also says Kosmo begins contributing in Q3 and will be low eight-digit revenue this year, which is third-party revenue inside the same guidance range.
What would change our view
- Neutral→AttractiveVolume converts into real profit
It needs all of:
- ○Full-year sensor shipments reach 2.7 million or more, with gross margin staying above 35%.checked at the next results, due in Nov 2026
- ○Operating profit without government grants turns positive through 2027, so the profit comes from the business rather than the state.checked at the next results, due in Nov 2026
- Neutral→UnattractiveThe economics erode or the listing risk returns
It needs any of:
- ○The gross profit on each sensor keeps falling after sales outside car sensors pass half of revenue.checked at the next results, due in Nov 2026
- ○The Pentagon relists Hesai after a fairer process, or related-company sales top a quarter of the new lines' revenue in 2027.watched as it happens
We would also be wrong if
- Gross margin falls below 35% for two quarters running.
What is coming
- 10 November 2026 Q3 2026 results: Revenue against the RMB1,100M to RMB1,150M guide, gross margin against 40.1%, whether revenue outside driver-assistance lidar nears half, and the first Kosmo revenue.
Every dated event ahead
- 10 November 2026 Q3 2026 results: Revenue against the RMB1,100M to RMB1,150M guide, gross margin against 40.1%, whether revenue outside driver-assistance lidar nears half, and the first Kosmo revenue.
- December 2026 Great Wall Motor start of production: Whether a Q2 2026 design win turns into shipped volume on time, the first test of design wins that carry no disclosed volume or price.
- March 2027 Q4 and full-year 2026 results: Shipments against the 3 to 3.5 million unit guide, Strategic Growth Initiatives revenue against RMB200M to RMB300M, and operating income excluding grants.
- July 2027 China's mandatory L3 and L4 safety standards take effect: Whether factory-fitted lidar becomes a requirement that lifts sensors per car, or the standards leave room for camera-only systems.
Since we wrote this
- 5 OctThesis reviewThe bet now names what it ridesThe owner asked that every bet name the macro trend behind it, after BWXT's bet was missing its chief executive's case for nuclear over gas.
- 4 OctThesis reviewOwner's checklist: revenue streams, cost of growth, management, catalystsThe owner's checklist found the method silent on who runs the company, what growth costs, where the money comes from and what is dated ahead; this pass fills those four blocks from the 20-F and the 2026 releases.
- 27 SepThesis reviewSharpa cap approved; 2025's unexplained profit was a Sharpa stake saleShareholders approved sales of up to RMB300M to the chief executive's robot company. Most of 2025's 'other income' turned out to be a gain on selling a stake in that same company.
- 25 AugThesis writtenNeutral, medium convictionIt had won the lidar price war, but its profit was interest and grants, and gross profit per sensor had fallen 36% in a year.
Also checked 5 times with nothing new to act on.
Sources 28
Sources - 28
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-04-24Three years of income statements, unit volumes and average selling prices, the customer and supplier concentration, the dual-class voting structure, the composition of other operating income, the share count, and the Department of Defense designation and its litigation history.backsC01C03C04C05C06C11C13C25C29C31C35C38C39C40C41C44C46C49C62C65C68C69C70C71
- S02primarycompany filing2026-08-18The quarter's statements and the earnings bridge below the operating line, the ADAS and robotics unit split, the balance sheet and borrowings, the cash reserve, the Q3 revenue guidance, the raised Strategic Growth Initiatives guidance and the July share subdivision.backsC01C02C07C10C18C20C21C24C26C27C29C31C33C36C37C38C39C63C66
- S03primarycompany filing2026-03-24The 2025 unit split of 1,381,133 driver-assistance and 239,273 robotics sensors, the full-year statements, the first full-year profit claim, the 2026 shipment outlook of 3 to 3.5 million units and capacity above 4 million.backsC03C06C12C27
- S04primarytranscript2026-08-18The related-party character of the new growth business, the raised guidance and its counterparty, the composition of robotics volume, the penetration and regulatory dates, and the Q3 revenue and unit guidance.
Demand from Sharpa has been the major source of our revenue for robotic actuation. … I myself am also a co-founder of Sharpa, a humanoid robotics company … we are seeking to increase the annual cap under the supply of products framework agreement with Sharpa from RMB 100 million to RMB 300 million, subject to shareholders' approval at the EGM … We actually raised the full year 2026 SGI revenue guidance from RMB 100 million to a range of RMB 200 million to RMB 300 million. For 2027, we are now looking at roughly USD 100 million … Most of our robotics lidar volume this year is robotic lawn mowers. Those customers are largely Chinese manufacturers focused on Europe as their key overseas markets. … Lidar penetration on EVs was close to 20% in 2025. We think it could get to roughly 30% to 40% this year. … China's mandatory L3 and L4 safety standards came out in August and should take effect in July 2027. … we've already secured multi-lidar design wins with Li Auto, Xiaomi and Changan with 3 to 6 lidars per car. … For Q3, we expect that the total revenue of between RMB 1.1 billion to RMB 1.15 billion with lidar shipments around 800,000 to 850,000 units. … Q3 is expected to be the quarter where revenue from outside ADAS lidar gets close to or even more than half of total revenue. … We are still expecting the full year outlook for our robotics lidar over 500,000 units. Last year, it was around 240,000.
backsC15C19C20C21C22C26C28C60 - S05primarycompany filing2026-05-19The second-quarter revenue guidance of RMB850M to RMB900M the quarter was then measured against, and the 31 March cash reserve that shows the six-month decline.backsC37C63
- S06secondaryvendor data2026-08-25Research pack for HSAI, frozen at content/HSAI/research-pack.jsoncontent/HSAI/research-pack.jsonPrice, the 20-day average daily range, the 252-day return and range, the analyst panel and estimates, and the filed-competitor cohort multiples. Its enterprise value, market value and revenue series all mix renminbi with dollars and are not used - see the data notes.backsC23C47C48C49C50C51
- S07primaryregulator2026-08-18Hesai Technology Co., Ltd v. Department of Defense, No. 25-5256, United States Court of Appeals for the District of Columbia Circuit United States Court of Appeals for the D.C. CircuitThat the appeal was decided on 18 August 2026. The court reversed on Hesai's due-process claim and remanded so the Secretary of Defense may give Hesai notice of the unclassified materials and an opportunity to respond before finalising the designation; it did not hold that Hesai is not a Chinese military company. The quote is the docket line from the court's own August 2026 opinions index, which is machine-readable where the opinion PDF is not.
25-5256 Hesai Technology Co., Ltd v. DOD 08/18/2026
backsC45 - S08primarycompany release2026-03-25RoboSense Technology full year 2025 results announcement RoboSense Technology Co., Ltd.The nearest rival's own unit volumes, revenue and gross margin - the only outside measurement of Hesai's cost position, and a robotics leadership claim that contradicts Hesai's.
For the full year of 2025, total LiDAR sales volume reached approximately 912,000 units, representing a year-over-year increase of 67.6%. In 2025, the Company ranked No. 1 in the robotics sector with total LiDAR shipments reaching 303,000 units, securing the top position across major segments. The annual revenue was approximately RMB 1.94 billion, with the overall gross margin rising to 26.5%.
backsC08C12 - S09primarycompany filing2026-03-02A competitor naming Hesai in its own competition section, and the other side of the arbitration that produced part of Hesai's other operating income.backsC16
- S10primarycompany filing2026-03-27Luminar Technologies Form 10-K for the fiscal year ended December 31, 2025 Luminar Technologies, Inc.What happened to the largest Western automotive lidar pure-play while Hesai took share: chapter 11 in December 2025, delisting, the lidar business sold to MicroVision, and a plan of liquidation cancelling all equity.backsC14
- S11tertiarynews2026-07-07The date of a national-security story the price reacted to. Recorded for the measured reaction and the date only: the substance is a think-tank commentator's argument, and the checkable facts behind it are the Department of Defense designation in the 20-F and the court record.
Chinese lidar maker with Nvidia ties accused of being cyber risk for U.S.
backsC47 - S12primarycompany filing2026-03-04Innoviz Technologies Form 20-F for the fiscal year ended December 31, 2025 Innoviz Technologies Ltd.The nearest like-for-like listed comparison - a Tier-1 automotive lidar supplier selling to the same kind of buyer, admitted to the cohort on that basis.read, but no claim rests on it
- S13primarycompany filing2026-03-04MicroVision Form 10-K for the fiscal year ended December 31, 2025 MicroVision, Inc.A cohort member's own description of the same product sold into automotive, robotics, warehouse, agriculture and mining.read, but no claim rests on it
- S14primarycompany filing2026-03-20Aeva Technologies Form 10-K for the fiscal year ended December 31, 2025 Aeva Technologies, Inc.A cohort member's own description of its product and its admission that it has primarily sold prototypes and engineering services.read, but no claim rests on it
- S15primarycompany filing2026-03-18A cohort member confirming the same product and the same physical-AI framing.read, but no claim rests on it
- S16primarycompany filing2026-07-31The terms of the Sharpa transaction that the earnings call did not: the actual amount transacted, the component breakdown of the cap, the cost-plus pricing policy and the transfer-pricing benchmark behind it, and the independent board committee's and independent financial adviser's recommendation.backsC22C60
- S17primarycompany filing2026-08-28The vote on the revised Sharpa supply agreement and annual cap: passed as an ordinary resolution, 581,636,027 Class B votes for (99.9796%) and 118,872 against, with the three co-founders and their associates abstaining on 217,311,136 shares.backsC22
- S18primarycompany filing2026-09-17Note 15, related party transactions: RMB44.9M of first-half 2026 revenue from Sharpa, RMB50.2M due from Sharpa at 30 June 2026 aged within six months, and a restatement of the 2025 Sharpa stake purchase and disposal; the subsequent-events paragraph confirming the 28 August 2026 approval of the revised cap.backsC22C35
- S19primarytranscript2026-02-05Regal Rexnord fourth quarter 2025 earnings call Regal Rexnord CorporationA listed motion-control maker selling joint subassemblies to humanoid robot makers, and how many of its targeted makers have put it on their platforms.
We continue to gain traction with humanoid OEMs and selling our subassembled [ axis ] solutions. … There's 10 key OEMs that we're targeting in the U.S. 3 of them, we are on their platforms, and we're selling subassemblies. 7 of them, we're still working on getting integrated.
backsC53 - S20primarytranscript2026-05-07Regal Rexnord first quarter 2026 earnings call Regal Rexnord CorporationThe size of that rival's humanoid business in orders: about $40M in 2025 and a little over $1M in the first quarter of 2026.
Humanoid in the quarter, we only saw a little bit over $1 million of orders. But I remind you that last year, we saw $40 million of orders. What gives us confidence in our position in humanoids is we're seeing more and more cross-sell opportunity.
backsC53 - S21primaryindustry data2026-09-30International Federation of Robotics press release - Global Sales of Professional Service Robots Surge 24% (World Robotics 2026 Service Robots) International Federation of Roboticsthe robotics industry body's count of full-size humanoids sold in 2025 and of professional service robots (almost 250,000, up 24%); read through ir_fetch
In 2025, about 7,000 units of full-size humanoids (above 140 cm) were sold globally. These robots are intended for commercial and professional applications beyond R&D and entertainment.
backsC54 - S22primaryindustry data2026-01-15IDC Market Perspective - Worldwide Humanoid Robotics Market Analysis 2026 (CHE54229626), abstract IDCan independent tracker's 2025 humanoid shipments and revenue, and the two leading makers at about 5,000 units each; the free abstract of a paid report (publication month January 2026, day not shown), read through ir_fetch
In 2025, the global humanoid robot market experienced rapid growth, with total shipments reaching approximately 18,000 units and market revenue of around US$440 million, representing a year-over-year (YoY) increase of roughly 508%.
backsC55 - S23primaryindustry data2026-09-29IDC blog - 半年出货 2.5 万台，同比增长432.1%：全球人形机器人商业化探索持续升温 (Global Humanoid Robot Tracker, first half 2026) IDCfirst-half 2026 humanoid shipments and revenue, China's 77.9% share, the falling share of research, education and display uses, and the raised 2030 forecast; read through ir_fetch
IDC数据显示，2026年上半年，全球人形机器人出货量近2.5万台，同比增长432.1%；市场规模超过7.4亿美元，同比增长322.7%。……预计2030年全球人形机器人出货量将超75万台左右，较此前预测上调约50%。
backsC55 - S24secondaryanalyst2024-02-27Goldman Sachs Insights - The global market for humanoid robots could reach $38 billion by 2035 Goldman Sachs Researcha bank's base case for 2030 humanoid shipments and its estimate of the fall in humanoid manufacturing cost to $30,000-$150,000 a unit; read through ir_fetch
The team’s base case is for more than 250,000 humanoid robot shipments in 2030, almost all of which would be for industrial use.
backsC56 - S25primarycompany release2025-11-17UBTECH press release - UBTECH Humanoid Robot Walker S2 Begins Mass Production and Delivery, with Orders Exceeding 800 Million Yuan UBTECH Roboticsa listed Chinese humanoid maker's own capacity targets, the makers' side of the shipment forecast; read through ir_fetch
Amid rising market demand, UBTECH has outlined a clear production ramp-up plan, targeting an annual capacity of 5,000 industrial humanoid robots by 2026, scaling to 10,000 units by 2027.
backsC57 - S26primarycompany release2026-10-01Unitree Robotics - G1 humanoid product page Unitree Roboticsthe joint count and list price of the highest-volume humanoid maker's main model, and its own joint motors offered for sale; read through ir_fetch
Extra large joint movement angle space, 23~43 joint motors ... Price from $13.5K
backsC58 - S27primarycompany release2026-05-13Humanoid press release - Humanoid Secures Landmark Deal with Schaeffler to Deploy Thousands of Humanoid Robots Humanoid (SKL Robotics Ltd)a rival actuator supplier's signed volume with a humanoid maker, from the buyer's own release; Schaeffler's own release answered 403 to ir_fetch
Under this agreement, Schaeffler will become Humanoid’s preferred supplier, covering more than 50% of the company’s demand for joint actuators for its wheeled-based platforms through 2031. This partnership is expected to translate into the supply of a seven-digit number of actuators.
backsC59 - S28secondaryother2026-01-06CES 2026 Innovation Awards - SharpaWave Consumer Technology Association (entry submitted by Sharpa)the joint count of the dexterous hand Hesai's actuators go into; the text is Sharpa's own entry, so it is the related party's description, read through ir_fetch
SharpaWave is a dexterous hand designed for robotics companies, academic institutions, and research labs. SharpaWave has 22 active degrees of freedom (DOF), at the 1:1 human size and scale.
backsC58