RoboSense Deep Dive: A LiDAR Supplier Powered by the Twin Engines of Automotive and Robotics

The Company's Developmental History

RoboSense emerged in 2014 for a reason. At the time, global LiDAR was a market dominated by Western mechanical solutions, expensive, and oriented toward validation and research. The company was founded in Shenzhen in 2014, with a core founding team including chairman and CEO Qiu Chunxin, CTO Liu Letian, and scientific adviser Zhu Xiaorui. Both Qiu and Liu came from control and automation backgrounds, the former holding a PhD in control science and engineering from the Harbin Institute of Technology and the latter a master's in control science and engineering from HIT; this set the company up to start from an integrated "sensor plus algorithm plus systems engineering" capability rather than from pure hardware sales. The company completed development of its mechanical LiDAR product in 2016, launching the RS-LiDAR-16 for robotics and similar scenarios; in 2017 it released the RS-LiDAR-Perception software and unveiled a new-generation solid-state LiDAR, the RS-LiDAR-M1pre; in 2019 it launched its first ADAS-oriented M1. This path resembles a curve that gradually pushed cost down and moved closer to automotive-grade mass production, rather than betting on automaker orders from the start.

The company's first real strategic pivot came between 2019 and 2023. The prospectus shows that the M platform was defined as the company's flagship automotive-grade solid-state LiDAR platform, emphasizing a 2D MEMS scanning chip, modular transceiver design, and manufacturability suited to mass production; in parallel, the R platform was reserved for traditional mechanical LiDAR, mainly for robotics and other non-automotive fields. More importantly, by 2023 the company had explicitly stopped producing the low-margin RS-LiDAR-16, tilting resources toward automotive and higher-value-added products. The constraints in this phase were very clear: mechanical LiDAR is structurally complex, bulky, and costly, unsuited to mass vehicle adoption, while China's smart-driving penetration was beginning to climb, opening a window for the better cost-performance-balanced solid-state path. The company chose to bet on automotive because, at the time, autos were the only market large enough, fast enough, and capable enough of amortizing R&D, not because robotics did not matter.

The listing path fits this narrative too. The company listed in Hong Kong on January 5, 2024 at an IPO price of HKD 43.00, with net proceeds estimated at roughly HKD 877 million per the prospectus. The market first bought it because Hong Kong rarely offers such a pure play capturing China's smart driving, domestic substitution, and high-end hardware going global at once, not because it was already profitable. But soon after listing, the capital market taught it a lesson: when it entered Stock Connect on June 11, the stock was bid up to HKD 90.8; on July 5, under lock-up expiry, it fell intraday to a low of HKD 15.28. This sharp rise and fall left deep, lasting effects. It made management realize that concepts like "number of design wins" and "industry leader" alone cannot support a valuation, and it made investors weigh shipment delivery, cash flow, and profit quality more heavily.

If I divide the company's history into four stages, I would do it this way. The first is the "technology validation period" from 2014 to 2018. Here the core task was building mechanical products, pairing them with algorithms, and establishing the earliest customer scenarios. The second is the "automotive platform formation period" from 2019 to 2023. The M1 and the M platform pushed the company from a "robotics and validation-tool supplier" toward an "ADAS mass-production component supplier"; meanwhile, halting the low-margin legacy product showed the company starting to actively manage its product mix. The third is the "listing and narrative volatility period" of 2024. The capital market first gave it enormous imagination, then knocked the valuation back down on lock-up expiry and delivery pace. The fourth is the "digitalization and twin-engine restructuring period" from 2025 to now. The most critical thing in this phase is that the company used its in-house SPAD-SoC and digital architecture to reorganize both ADAS and robotics on the same underlying technology, not that shipments hit a record high.

2025 is the turning point of the whole story. The annual report puts it plainly: in Q1 2025, a major automaker and a tier-one supplier suspended cooperation with the company; management treated this shock as a window to change the business structure and avoid the risk of a single market and a handful of top customers, and therefore launched a full line of digital LiDAR products such as EMX, EM4, E1R, and Airy that year, pushing robotics hard as a second growth curve. This pivot was active diversification forced by automotive-customer volatility, not a temporary chase of a hot theme because robotics looked exciting. The result is that, even while the company remained loss-making for the full year, its revenue mix changed qualitatively: Q4 robotics and other revenue was about 347 million yuan, already close to ADAS's roughly 361 million yuan.

A longitudinal financial review makes this line clearer. The annual report's financial summary shows revenue growing from 331 million yuan in 2021 to 1.941 billion yuan in 2025, nearly a sixfold increase in four years; gross profit recovering from a gross loss of 39.295 million yuan in 2022 to gross profit of 514 million yuan in 2025; and operating loss narrowing from 941 million yuan in 2023 to 184 million yuan in 2025. The line that truly needs unpacking is net profit. The accounting losses from 2021 to 2023 were largely magnified by "fair-value changes on financial instruments issued to investors," reaching 1.488 billion yuan, 1.485 billion yuan, and 3.471 billion yuan respectively; after listing in 2024, this item nearly disappeared, and only then did the net loss begin to track actual operations more closely. In other words, judged only by the bottom line, pre- and post-listing comparability is actually poor.

But the signals beyond the income statement are not easy. In 2025 the company's net operating cash outflow was 582 million yuan, versus only 65.1 million yuan in 2024; the cash-flow notes show the drag came mainly from a 338 million yuan increase in receivables and notes receivable and a 229 million yuan increase in inventory. On fundamentals, this shows that although the company dressed up its income statement considerably in 2025, it also materially loosened working-capital usage to seize the volume window. This is normal for a growing hardware company, but it means "profit improvement" and "cash improvement" are not the same thing. The narrowing of net operating cash outflow to 89.2 million yuan in Q1 2026 is a positive signal, but it only proves the worst may be over, not that a cash-return model has been established.

The capital actions are also worth revisiting. After the IPO, the company bought back shares in both 2024 and 2025, and in February 2025 placed 22 million shares at HKD 46.15. Buybacks and placements look opposite in direction but in substance reflect two judgments by management: on one hand, stabilizing market expectations after the July 2024 lock-up plunge; on the other, the need for cash to fund the mass production of digital products and supply-chain and capacity buildout, while the company had yet to become self-funding. This capital allocation is hardly aggressive; it looks more like a growth-hardware company still buying time than a mature company's logic of "balanced returns to shareholders."

The share-price and valuation history can almost serve as a capital-market microcosm of this company. Early after listing, the market was willing to grant a scarcity premium; when it entered Stock Connect, capital was willing to price it as "industry leader plus scarce Hong Kong-listed AI hardware"; but after the lock-up, the market quickly scrutinized it by a harsher standard: where are the profits, where is the cash flow, did the design wins actually turn into revenue. Based on the June 11, 2024 close of HKD 90.8 and a market cap of HKD 40.945 billion, the market was then assigning a sales multiple far higher than today's; now at HKD 28.50 and HKD 13.46 billion, its valuation center has clearly shifted down from a "concept growth stock" to a "scale hardware stock awaiting validation." The valuation center changed because the market is no longer willing to pay as much for "might be profitable" as for "already profitable," not because the industry story has vanished.

Business Model, Moat, and Industry Cycle

RoboSense's real business engine can be summed up plainly: using increasingly cheap and increasingly standardized LiDAR hardware to fit as many vehicle models and robotics scenarios as possible onto the same technology and manufacturing curve. The company's revenue comes from three parts: products, solutions, and services and others. By 2025, products were the absolute majority; within that, ADAS products and robotics and other products are the two most critical revenue buckets. In the first half of 2025, ADAS revenue was 500 million yuan, down 17.9% year over year, while robotics and other revenue was 221 million yuan, up a sharp 184.8% year over year; through the first three quarters, ADAS revenue was 745 million yuan and robotics and other revenue was 363 million yuan; in Q4 of the full year, robotics and other product revenue was about 347 million yuan, nearly matching ADAS's roughly 361 million yuan. The way the company actually makes money today is by making enough unit economics replicable, not by selling software subscriptions or high-margin service fees.

This business model determines its cost structure. Fixed costs are mainly R&D, administration, and production-line preparation; variable costs are mainly raw materials, chips, optical components, packaging, and manufacturing. In 2025, R&D expense was 647 million yuan, still higher than full-year gross profit of 514 million yuan; this means the company already has operating leverage, but operating leverage has not yet fully overtaken R&D leverage. Capital expenditure in 2025 was not extravagant, with 85.53 million yuan to acquire property, plant, and equipment and 11.23 million yuan for intangible assets, well below R&D spending; what truly consumes cash is the working capital inflated to win orders and build inventory, not the factory itself. For investors, this has two implications: first, this is not a heavy-asset, factory-type enterprise, and long-term success or failure still hinges on technology roadmap and product definition; second, as long as competition persists, R&D cannot be casually compressed.

On the moat, I think only four lines genuinely hold. The first is the speed of cost reduction from chip-based and digital architecture. In 2025 the company emphasized that its in-house SPAD-SoC and 2D VCSEL chips completed automotive-grade certification, and launched digital products such as EMX, EM4, E1R, and Airy; as far back as the prospectus, the M-platform SoC and the E1 in-house chip path were already clear. For hardware like LiDAR, which is easily ground by price wars into "homogenized performance specs," the moat lies in whether you can thin the BOM faster than rivals, not in the spec sheet. The second is sharing an underlying platform across automotive-grade mass production and robotics. For RoboSense, robotics is a spillover of automotive-grade supply chain, chips, and algorithm capability into adjacent scenarios, not an entirely independent new business. The third is the scale economy that comes from penetration with domestic Chinese automakers. The fourth, and only then, is some degree of customer switching cost. After a vehicle's SOP, an automaker will not lightly change a core sensor mid-cycle; but before a design win, bargaining power is unstable.

Some "moats" often cited by outsiders I actually find not hard enough. Brand is not one. Automakers buying LiDAR look mainly at cost, stability, mass-production capability, and functional safety, not consumer brand preference. Patents are not decisive either. The company does hold a large patent pool, but the prospectus also disclosed that it once had a cross-license for mechanical LiDAR patents with Velodyne and paid licensing fees. LiDAR is an industry that cannot bypass process, system integration, and supply chain, and a single patent cannot lock out rivals. What is truly hard is running chips, algorithms, automotive-grade validation, supply chain, and customer projects smoothly together.

On governance, RoboSense's strengths and discounts are both typical. The strength is that the founding team is still on the front line, with Qiu Chunxin in charge of overall strategy and operations, Liu Letian leading technology, and Zhu Xiaorui serving as scientific adviser; this gives the company higher execution efficiency when switching roadmaps. The discount comes from strong internal binding within the founding team and governance traits common to growth companies: Qiu Chunxin and Qiu Chunchao are brothers, the company uses a Cayman structure and is Hong Kong-listed, equity incentives are heavy, and it is not averse to external financing during its expansion phase. So far, public disclosures show no major financial fraud or audit dispute, but that does not mean governance risk can be ignored. For companies like this, the governance question is often "when will it dilute shareholders again for growth," not "are the financials fake."

At the industry level, LiDAR has moved from a proof-of-concept phase into a mass-production phase. TrendForce projects the global LiDAR market to grow from USD 1.181 billion in 2024 to USD 5.352 billion in 2029, within which the automotive LiDAR market reaches USD 3.443 billion by 2029 and the industrial and logistics LiDAR market reaches USD 1.909 billion. Yole Group projects the global automotive LiDAR market to grow from about USD 859 million in 2024 to USD 3.56 billion in 2030, and notes that Chinese manufacturers control 88% of the market. TechInsights offers an even more direct judgment: of global high-resolution automotive LiDAR units in 2025, Hesai, RoboSense, and Huawei together will account for 83%. Taken together, these figures show the LiDAR story has become "who can run it into a commodity component at Chinese cost and global quality," no longer "whether to use it."

The industry profit pool is currently uneven. Installed volume of automotive LiDAR in China is already large, with Gasgoo counting 2.756 million primary LiDAR units installed on Chinese passenger vehicles in 2025, a penetration rate of 21% among new-energy passenger vehicles, crossing the 16% "chasm threshold"; but profit has not been broadly released in step, because ASP is falling extremely fast and the profit pool remains concentrated among leaders with stronger scale, supply chain, and chip capability. RoboSense is worth studying precisely because it is a participant on that profit-pool migration path, not a bystander. Only, going forward on this road, valuation will be decided by whether gross margin can hold amid shipment growth, not by the shipment curve itself.

In terms of cyclicality, this is more a superposition of "technology iteration cycle plus automaker capex cycle plus policy cycle" than a traditional macro-cyclical stock. The variable that benefits most in the up phase is rising penetration and more units per vehicle; the variable most fragile in the down phase is price. As long as the industry is still in the early stages of scaling, price cuts can easily look like growth, until they squeeze gross margin to the point where operating leverage fails. For RoboSense, 2025 and 2026 are proving one thing: its shipments have indeed broken out, and now the question is whether the margin can break out alongside them.

Policy and geopolitics are external variables that must be faced. The U.S. Department of Commerce finalized its "connected vehicles" rule in 2025, banning vehicle-network hardware and related software with sufficient ties to China or Russia from the U.S. connected-vehicle supply chain, with software restrictions starting from model year 2027 and hardware restrictions broadly taking effect at model year 2030 or in 2029. At the same time, the U.S. Department of Defense updated its Section 1260H list on June 8, 2026, including both Hesai and RoboSense. 1260H itself is not a comprehensive sanction, but it will restrict direct DoD procurement from June 30, 2026 and expand to indirect procurement in 2027; the larger effect is that it raises the compliance cost and psychological barrier for U.S. customers and some European customers when adopting Chinese sensors. On the other hand, Reuters reported in March 2025 that Mercedes-Benz still decided to use Hesai LiDAR in its global-market models, citing superior cost and scaling capability. This shows geopolitical friction will not make Chinese LiDAR immediately lose the global market, but it will force companies to turn overseas expansion from a "pure commercial question" into a "commercial plus compliance plus production-location" question.

Competitive Comparison

If we start only from the question "among listed companies, who most resembles RoboSense," the answer is actually very concentrated: the only directly comparable name is Hesai, and the others are more boundary references. Ouster is more an industrial and infrastructure perception company; Aeva is more a technology option betting on the FMCW path and L3 large customers; Luminar is more a cautionary counter-example, once overvalued by the capital market and later hit back by mass-production delivery and governance problems. RoboSense's direct battlefield is still contesting the leadership in Chinese automotive scaling and global robotics upside with Hesai.

Table note: RoboSense data is based on the 2025 results announcement, annual report, and Q1 2026 announcement; Hesai data is based on full-year 2025 and Q1 2026 results disclosures, plus market-cap data near the research reference date. Price-to-sales is estimated from each company's current market cap and 2025 revenue.

The most interesting thing about this table is that the two companies "have grown into different shapes," not who is bigger or smaller. Hesai is more a scale manufacturer of standardized automotive LiDAR. With 1.62 million units of total shipments and 1.38 million ADAS units in 2025, its ADAS share far exceeds robotics; a gross margin of 41.8% means it has formed a more pronounced cost and product advantage on large-scale automotive standard parts. Customers choose Hesai for two words at the core: scale and certainty. Especially when automakers want to push quickly down to the price band above RMB 200,000 or even RMB 150,000, whoever can offer a more stable BOM and supply has more say.

RoboSense is more a company with "a more balanced twin-engine structure but more fragile gross margin." Its total 2025 shipments were below Hesai's, and its ADAS volume also trails Hesai, but its robotics shipments are actually higher, reaching 303,000 units for the year. This is its biggest differentiator and its biggest valuation suspense. The market is willing to grant this business a premium because robotics looks like an earlier-stage, higher-elasticity incremental market; the market remains slow to grant too high a premium because this business's ASP is falling faster and its profit structure is not yet fully proven out. Why would a customer choose RoboSense? On the automotive side, because it has accumulated mass-production experience among domestic Chinese automakers and joint-venture brands, and can keep cutting cost with digital SoC solutions; on the robotics side, because it can turn automotive-grade reliability and cheaper digital products into "good enough and affordable" general-purpose perception parts. Why might a customer leave it? The answer is equally practical: automakers will switch suppliers over cost and internal tier-one systems, and robotics customers will press prices over low-cost substitutes, improvements in vision-based approaches, or demand seasonality.

Bringing in Ouster and Aeva helps clarify the boundaries of RoboSense's niche. Ouster posted USD 49 million of revenue in Q1 2026 at a 43% GAAP gross margin, with a pitch of an end-to-end physical-AI perception platform of "sensor plus camera plus AI compute plus software," with customers more from industrial, robotics, and infrastructure; its valuation logic is closer to a "high-margin perception platform" than a pure automotive component. Aeva posted USD 6.3 million of revenue in Q1 2026, growing fast year over year but very small, still in a phase driven by large projects and roadmap validation; customers buy it for FMCW's ability to deliver both velocity and distance, and for the imagination of L3 large customers. Compared with them, RoboSense's position is more grounded: it lacks Ouster's degree of software focus and Aeva's strong technology-option flavor, but it is closer to large-scale mass production and a larger revenue base.

This also explains why I define RoboSense as a "challenger-type leader in the industry." On the matter of Chinese passenger-vehicle LiDAR mass production, it is no longer a follower; among global listed peers, it has not yet reached Hesai's position of "earnings certainty clearly a tier higher." What it most directly competes for is the profit pool of Hesai, Huawei, and Seyond in China's automotive and robotics markets; and the players most likely to take its profit pool are exactly these names with greater scale and closer ties to automakers' core procurement systems. If the industry keeps advancing along technology substitution, price wars, and mid-tier model penetration, RoboSense's scale position may not weaken, but its valuation position may not strengthen, because scale itself will become more and more common while profit quality will become more and more scarce.

Current Fundamentals, Valuation, and Risks

Looking first at the operating cross-section of the most recent five quarters comes closer to the real picture than looking only at full years.

Table note: Q2 2025 is derived as H1 2025 minus Q1 2025, and Q3 2025 as 9M 2025 minus H1 2025; the rest are directly disclosed by the company. Amounts are in 100 million yuan, and shipments are in 10,000 units.

This table yields three conclusions. First, volume is already up. Q4 2025 shipments approached 460,000 units, and Q1 2026 still saw 330,000 units against a low-season and falling-price backdrop, which cannot be explained by one-off orders. Second, profit has not truly stabilized. The Q4 gross margin of 28.5% is good, but Q1 2026 returned to 21.7%; the Q4 net profit of 104 million yuan also did not continue into the following quarter. Third, price is falling faster than revenue is growing. In Q1 2026, product revenue was 441 million yuan and total shipments were 330,300 units, implying a blended ASP of only about 1,336 yuan; the corresponding blended ASP in Q1 2025 was about 2,784 yuan. The company's operating reality now is this: shipments grew sharply, true, and unit price fell quickly, also true.

Therefore, the theme the market is truly trading is three narratives holding simultaneously, not simply "rising autonomous-driving penetration": digital products can keep cutting cost, the robotics second curve can keep scaling, and the company can soon cross the full-year profit threshold. The first two already have clear evidence; the third does not yet. Supporters' strongest evidence comes from gross margin rising from 17.2% to 26.5% in 2025, and from the clear improvement in operating cash flow in Q1 2026. Skeptics' strongest evidence comes from two places: first, Q4 profit contains a relatively high proportion of other income and other gains; second, Q1 2026 gross margin slipped and the company returned to a loss, showing that profit elasticity remains highly sensitive to product mix and seasonality.

The bull logic is very concrete. First, in-house chips and digital architecture are indeed working. As early as Q1 2025, the company disclosed that after the MX product adopted an in-house SoC, ADAS product gross margin rose from 10.6% to 15.1%; the gross margins of the robotics E1R and Airy are also clearly higher than the old mechanical series. Second, the robotics business is not a slide deck. Full-year 2025 robotics LiDAR shipments were 303,000 units, with 221,200 units in Q4 alone, and there are real orders across lawn-mowing robots, autonomous delivery, and embodied-intelligence scenarios. Third, the cash side is at least improving. Q1 2026 operating cash outflow narrowed sharply year over year, showing the company beginning to shift from "using cash to grab growth" to "restraining cash burn within growth where possible."

The bear logic is equally concrete. First, profit quality is not clean. Within Q4 operating profit of 130 million yuan, if the roughly 95.38 million yuan fair-value gain recognized in 2025 is treated as a mainly one-off benefit, Q4 operating profit would fall back to the order of just over 30 million yuan; and if other income is separated out, core operations still hover near breakeven. This is my inference, but it is built on the company's disclosed details. Second, the price war is fiercer than the market imagines. In the first half of 2025, ADAS ASP had already fallen from 2,600 yuan to 2,300 yuan, and robotics from 8,700 yuan to 4,800 yuan; backing out from full-year and Q4 figures, prices are still falling. Third, working-capital pressure has not been lifted. In 2025, receivables and inventory together consumed more than 500 million yuan of cash, which for a company not yet stably profitable is a precursor of financing needs. Fourth, overseas expansion faces more than a pure sales problem. The U.S. connected-vehicle rule and the 1260H list will make overseas automakers clearly more cautious about adopting Chinese LiDAR.

On valuation, I think a P/E is still not the right tool. RoboSense was still loss-making for full-year 2025 and did not turn positive in Q1 2026; more importantly, the profit quality of Q4 2025 is not suitable to be annualized directly. Look first through cash flow. In 2025 the company's net loss was 145 million yuan, but net operating cash outflow was 582 million yuan; after deducting roughly 96.75 million yuan to acquire fixed assets and intangible assets, a rough "owner earnings" figure is still deeply negative. In other words, the apparent narrowing of the loss has not translated into distributable cash. Precisely for this reason, the current valuation can only be viewed through the sales multiple, gross-margin elasticity, and the operating-leverage path.

Based on the 2026-06-12 close of HKD 28.50, the company's market cap is about HKD 13.46 billion. Per Q1 2026 disclosures, cash and cash equivalents, time deposits, restricted cash, and restricted time deposits totaled roughly RMB 2.6 billion, and with year-end 2025 bank borrowings of RMB 447.2 million, net cash is roughly RMB 2.15 billion; on this basis, RoboSense's current price-to-sales is about 6.0x and EV/Sales about 4.9x. The most important anchor for this multiple is Hesai. Based on Hesai's market cap near the research reference date and 2025 revenue, Hesai's price-to-sales is about 5.4x, while its revenue is larger, its gross margin clearly higher, and it already booked net profit of 18.30 million yuan in Q1 2026. To put it bluntly: RoboSense is not currently treated by the market as the "clearly cheap number two"; what it receives is still part of a robotics premium.

Below are three valuation scenarios. This is a stress test that places revenue, margin, and multiple assumptions in the same table, not investment advice.

Converting from the assumptions above, the three price ranges I give are: bear HKD 19-23, base HKD 29-35, bull HKD 41-47. Behind them is "whether profit and cash can grow alongside revenue," not "whether revenue can grow." If it can only grow revenue, the multiple will rapidly converge toward a component company; if it can also stabilize gross margin and cash flow, the market will gradually accept a higher multiple.

On the margin-of-safety review, my conclusion is clear: there is none. The current price of HKD 28.50 is above the ideal buy range of HKD 19-23 that I give, and above the margin-of-safety price warranted for a company that has yet to form stable owner earnings for the full year. The most fragile assumption across the three scenarios is that the robotics business can maintain sufficient margin while continuing to scale. Once this assumption is cut by 30%, the base-scenario valuation drops back to roughly HKD 24-27. Looking again at a zero-growth case, if there is no substantive improvement in revenue and earnings capability over the next three years, then even pricing at roughly 3x EV/Sales puts the share price closer to HKD 19-20, with a negative three-year annualized return. Good company but bad price fits perfectly here.

On risks, I will expand on only five variables that could truly cause permanent capital loss. First, the price war and ASP keep falling, high probability, high impact. The most direct observation indicators are product ASP, blended gross margin, and the automotive/robotics revenue mix. If the company pushes prices all the way down below 20% gross margin to defend shipments, the market will find that revenue growth cannot translate into profit, and the valuation will naturally derate along component logic. Second, the robotics business has customer concentration and seasonality, medium-high probability, high impact. A large part of the 2025 robotics surge came from a few concentrated scenarios such as lawn-mowing robots, delivery, and embodied intelligence, and once a single large customer's stocking pace changes, inventory and revenue will be distorted at the same time. Third, working-capital deterioration, medium probability, high impact. The 2025 change in receivables and inventory has shown that a volume phase can easily convert income-statement improvement into cash-flow deterioration. Fourth, customer switching and in-house substitution, medium probability, medium-high impact. The company itself disclosed in its 2025 annual report that a major OEM and a tier-one supplier suspended cooperation; this shows the so-called "design-win moat" is far from as stable as the market thinks. Fifth, geopolitics and overseas compliance, medium probability, medium-high impact. The U.S. connected-vehicle rule and the 1260H list will not zero out overseas demand immediately, but they will raise the uncertainty of realizing overseas revenue and suppress the convergence of the valuation discount over the long term.

Catalysts, Tracking Indicators, and a Cross-Sectional and Longitudinal Synthesis

Over the next year, the positive catalysts most likely to push the share price up are, in order, three things. First, two consecutive quarters of pulling blended gross margin back into the 24%-26% range while operating cash flow keeps improving. Second, the robotics business keeping high growth while its quarterly ASP no longer falls steeply, proving this is not simply trading price for volume. Third, the automotive digital platform entering more mass-production models, truly turning "design wins" into "revenue." The negative catalysts are equally clear: gross margin continuing below 22%, with Q4-style profit unable to be replicated again in a high season; receivables and inventory continuing to exceed revenue growth in 2026; and geopolitical regulation expanding further to a broader set of civilian automotive perception components.

Among these indicators, I weigh the first three most. Because they directly answer the market's biggest doubt: whether Q4's profit is a replicable starting point. The tracking channels are simple. Gross margin, operating cash flow, receivables, and inventory come from the company's interim reports, annual reports, and voluntary quarterly disclosures; blended ASP can be backed out by dividing product revenue by shipments; overseas compliance requires continuous tracking of the U.S. BIS and DoD lists and company announcements.

Looking longitudinally, RoboSense has truly proven three capabilities. First, it can survive roadmap switches. The company moved from mechanical products to the solid-state M platform and then to the digital SoC path in 2025, without falling behind in any generational technology shift. Second, it can turn R&D into delivery. Full-year 2025 shipments of 912,000 units and Q4 single-quarter shipments of 459,600 units are no longer numbers a lab company would have. Third, when its customer structure took a hit, it at least had the ability to find a new source of growth. After losing some automotive customers in Q1 2025, the robotics business quickly filled the growth gap.

But looking horizontally, its weaknesses are equally concrete. Compared with Hesai, its scale is smaller, its gross margin is lower, and its current-quarter profit stability also trails its rival; compared with Ouster, its software and platform premium is not rich enough; compared with Aeva, it lacks as much room for roadmap imagination. Its strength is real-world mass-production capability and a twin-engine structure, its weakness is that profit quality is still thin. The place the market is most likely to misjudge is exactly here: treating high robotics growth simply as "a higher grade of growth" overestimates the valuation; treating the renewed loss in Q1 2026 as "Q4 was entirely invalid" underestimates the scale threshold this company has already crossed.

The most critical variable over the next 1 year is whether gross margin can hold steady while shipments keep growing. The most critical variable over the next 3 years is whether the robotics business can move from "explosive orders" to a stable revenue pool "with repeat purchases, with profit, and with customer diversification." The most critical variable over the next 5 years is whether LiDAR, like the camera and millimeter-wave radar, will become a standard part across a wider price band of models and more categories of service robots; if so, RoboSense will keep benefiting from the Chinese supply chain's reshaping of the global cost curve; if not, it will increasingly resemble a fiercely competitive supplier of optional components.

I give four bull reasons and four bear reasons.

• Bull reason 1: In 2025 gross margin rose from 17.2% to 26.5%, proving in-house chips and scale effects have begun to improve the unit-economics model.

• Bull reason 2: Full-year robotics shipments of 303,000 units and Q4 single-quarter shipments of 221,200 units show the second curve has moved from a theme to a real revenue source.

• Bull reason 3: Q1 2026 operating cash outflow narrowed sharply year over year, showing growth's drain on cash is easing.

• Bull reason 4: In the global high-resolution automotive LiDAR unit market, the Chinese supply chain is dominant, and RoboSense sits within the most-benefited leading bandwidth.

• Bear reason 1: The first profit in Q4 2025 is laced with a relatively high proportion of other income and fair-value gains, and profit quality still needs validating.

• Bear reason 2: ASP is falling fast, and the Q1 2026 blended ASP is already clearly below the same period in 2025, so volume growth may not cover price decline over the long term.

• Bear reason 3: Net operating cash outflow in 2025 was 582 million yuan, dragged mainly by receivables and inventory, with cash-flow performance weaker than the income statement.

• Bear reason 4: Relative to a peer like Hesai with larger scale and higher gross margin, RoboSense currently has no clear valuation discount, and the margin of safety is insufficient.

If I run a pre-mortem in which this investment is down 50% three years out, I think there are two most-likely scripts. The first script plays out in 2027: Hesai, the Huawei ecosystem, and automakers' in-house development bound to tier-ones together press automotive LiDAR unit prices down another 30%-40%, RoboSense is forced to follow on price to keep its installed volume, blended gross margin falls back to 18%-20%, and the robotics business fails to provide higher profit quality, so the market re-compresses it from the current near-5x EV/Sales back to 2.5x-3.0x, and the share price easily falls to HKD 15-18. The real basis supporting this script is precisely the ASP decline and customer switching that have already occurred over the past year.

The second script plays out between 2027 and 2028: the robotics surge comes mainly from a few highly concentrated scenarios and large customers, and as lawn-mowing robots, delivery vehicles, and some embodied-intelligence customers enter a destocking or roadmap-adjustment phase, robotics shipments no longer grow fast and even show clear quarterly swings; meanwhile, the receivables and inventory the company built up to scale in 2025-2026 do not recover in time, cash flow deteriorates again, and it has to extend its runway through refinancing. By then, the market will turn the "second curve" from a valuation plus into a source of volatility, and the share price may likewise be halved.

The final research conclusion is this. RoboSense is no longer a stock you can buy on story alone; it has reached the stage where it must speak with profit quality and cash flow. What is truly valuable about the company is that it has indeed completed a very hard pivot: from mechanical products, from a single automotive line, from a conceptual R&D company, into a hardware enterprise with large-scale delivery capability that has landed on both the automotive and robotics incremental markets. And the problem lies precisely here. The market has already seen that it can "sell"; what it now waits for is whether it can "stay." Staying does not only mean revenue keeps growing; it means gross margin does not fall, operating cash flow is no longer distorted, and Q4-style profit no longer relies on other gains to prop it up.

At the current price, I can accept the judgment that "the company is better than a year ago," but I cannot yet accept the judgment that "the stock is therefore already cheap." Compared with Hesai, RoboSense has not yet matched a near-equal sales multiple with higher profit stability; compared with itself, it has not yet turned the Q4 2025 profit inflection into a sustained profit range in 2026. What would truly change my view is two to three consecutive quarters of proof, not a few more design-win headlines: while ASP keeps falling, gross margin still holds, cash flow keeps improving, and the robotics business is still not low-price volume.

【Company Profile Score】

• Fundamental quality: Medium

• Growth: High

• Moat: Medium

• Financial soundness: Medium

• Management credibility: Medium

• Valuation attractiveness: Low

• Risk level: High

• Suitable investor type: Long-term growth

【Investment Rating】

• Rating: Watch

• One-line investment thesis: Robotics volume is materializing, but durable profitability still needs validating.

• Three-tier price signals: Ideal buy price: HKD 19-23

• Acceptable holding price: HKD 29-35

• Clearly overvalued price: above HKD 47

• Current price classification: outside the three tiers

• Whether it is worth waiting for a better price: Yes. The trigger is the share price returning below HKD 23, or the company pulling blended gross margin back above 24% for two consecutive quarters while continuing to improve operating cash flow. The opportunity cost of waiting is that if profit quality is suddenly proven in the second half of 2026, the share price may return to the base range ahead of fundamentals.

• Target holding period: 3-5 years

• Expected annualized return: bear about -10%, base about +4%, bull about +16%

• Maximum loss risk: about 50%. The trigger is another sharp ASP decline, gross margin falling below 20%, and robotics demand volatility compounded by refinancing.

• Signals that trigger a reassessment: If blended gross margin is below 22% for two consecutive quarters

• If full-year operating cash flow continues to be clearly weaker than the net-profit improvement

• If the robotics revenue share keeps rising but quarterly gross margin declines in step

• If a major customer suspends cooperation or a design win moves into mass production below expectations

• If the U.S. or Europe imposes broader restrictions on Chinese automotive perception components

【Ideal/Fair Buy Price】HKD 19-23 Basis: the price range corresponding to the bear scenario's 2026 revenue of 2.2-2.3 billion yuan, EV/Sales of 2.8x-3.2x, and retention of some net-cash cushion; it is the price at which "even if profit keeps wavering, the loss will not run out of control," not the price at which "the company will decline."

【Valuation Range】

• current: 28.50 (as of the 2026-06-12 close)

• bear (conservative · ideal buy zone): [19, 23]

• base (fair · acceptable holding zone): [29, 35]

• bull (optimistic · above the clearly-overvalued line): [41, 47]

The key data table is below.

Table note: amounts are in 100 million yuan. Net losses from 2021 to 2023 were materially affected by fair-value changes on financial instruments such as preferred shares, and cannot be simply compared horizontally with the post-listing basis.

I remain cautious on four points. First, the company does not disclose a complete quarterly split of ADAS and robotics shipments, so the segment ASP for Q1 2026 can only be estimated on a blended basis. Second, disclosure of robotics customer concentration and repeat-purchase structure is still limited, making it impossible to judge precisely how much of the 2025 surge came from one-off stocking. Third, the prices and orders of non-listed competitors such as Seyond, Huawei, and the DJI ecosystem are not transparent enough, so the horizontal comparison is inherently incomplete. Fourth, the quarterly contribution of other gains and fair-value changes to Q4 2025 profit can only be inferred with high confidence, not reconstructed line by line as in a review of working papers.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.

Other Tickers Mentioned in This Report

• HSAI.US — the most direct global listed comparison with RoboSense, comparing shipments, gross margin, the timing of profitability, and valuation.

• OUST.US — an overseas LiDAR comparable, representing the industrial and robotics perception-platform path.

• AEVA.US — a representative of the FMCW 4D LiDAR path, used to observe the valuation boundary of "high-expectation, low-scale" companies.

• LAZR.US — an early overseas automotive LiDAR star name, used as a counter-reference for mass-production delivery and governance risk.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.