RoboSense

Phase A — Understand the business

Company Overview

RoboSense (速腾聚创 / Suteng Innovation Technology) is a Shenzhen-based LiDAR and robot-perception company — the world's volume leader in automotive LiDAR that is, in 2025-26, aggressively repositioning itself from "ADAS sensor supplier" into a full-stack robotics technology platform. Founded 2014; IPO'd on the HKEX Main Board 5 Jan 2024 at HK$43/share (first HK IPO of 2024, ~HK$20bn / ~US$2.6bn debut valuation).

How it makes money — two engines:

1. ADAS LiDAR (automotive). Solid-state / MEMS + now "digital" LiDAR sold to OEMs and Tier-1s for L2+/L3 driver assistance. This was ~100% of the story at IPO; it is now roughly half.
2. Robotics LiDAR + components. LiDAR and a widening stack of robot parts (vision, actuators, controllers, a dexterous hand, a planning model) sold to lawn-mower, delivery, cleaning, industrial, logistics, inspection and early-humanoid makers.

Scale (FY2025): revenue RMB 1.941bn, +17.7% YoY (~US$270m TTM); 912,000 total LiDAR units, +67.6% YoY. Customer breadth: 310+ automotive brands/Tier-1 partners and 3,400+ robotics customers.

Contract structure: design-win → mass-production model typical of auto components — long qualification cycles, then multi-year per-model volume once "designated." ADAS order backlog >9 million units. No take-or-pay; revenue is unit-shipment-driven and therefore highly ASP-sensitive (the central risk — see Lens 5/13). Key backers on the cap table double as customers: Cainiao/Alibaba (biggest pre-IPO holder, ~10.46%) and BYD among 30+ strategic investors across 12 rounds.

Supply Chain

The robotics commercial-layer map (kb/robotics/wiki/supply-chain.md) places RoboSense at the Sensor node, alongside Hesai/Luminar — but that map is humanoid-centric and understates RoboSense's real position: it is both a component supplier into robot/car builders AND, increasingly, a vertically-integrating chip designer.

Upstream → RoboSense → end customer, named:

• Upstream inputs: photodetectors/lasers historically bought-in (Sony/onsemi-class SPAD & VCSEL parts, plus Chinese optics/PCB/assembly). RoboSense's strategic move is to replace this with in-house silicon — its own SPAD-SoC receiver chip and 2D-VCSEL emitter chip ("digital LiDAR") — claimed to cut BOM cost by ~50% vs discrete architectures. This is the moat bet (Lens 3).
• RoboSense (integrator): designs the chips, LiDAR (EM4, Airy, E1R, EMX, Fairy), plus new robot parts — Active Camera (AC2), Papert 2.0 dexterous hand (20-DOF), FS-3D force sensor, LA-8000 linear motor, DC-G1 domain controller, and the VTLA-3D planning model.
• Downstream — automotive: BYD (and sub-brands), Geely (Galaxy, Zeekr — incl. Zeekr 9X on a 520-beam digital LiDAR), IM Motors, plus the China JVs of Toyota, Nissan, VW, GM; 70%+ share of JV-brand LiDAR. Rivian R2 is a US customer (Rivian declines to name the supplier amid US-China tension).
• Downstream — robotics: 3,400+ buyers; named strength in commercial cleaning (71% share, GGII Q1-26), robotic lawnmowers, delivery robots; an Active Camera order from a "leading humanoid robotics firm".

Chokepoints / single-source risk: (1) the in-house chip ramp is the whole margin thesis — if SPAD-SoC/VCSEL yields disappoint, the cost-down evaporates; (2) wafer/fab access for those chips (foundry dependency, plausibly SMIC-class given US-China constraints) — undisclosed, a real single point of failure; (3) demand-side concentration in BYD (Lens 13).

Competitive Advantages (moats)

What's genuinely durable:

• Volume leadership → cost curve. Yole ranked RoboSense No.1 in global passenger-car LiDAR by annual and cumulative unit volume (2024). Most units shipped = best learning-curve economics in a business where price falls every year. That is the closest thing to a moat in commodity-tending hardware.
• In-house digital-LiDAR silicon. Moving from discrete parts to proprietary SPAD-SoC + 2D-VCSEL chips is a genuine vertical-integration edge — it is what let Hesai (the rival) reach profit first, and RoboSense is now running the same playbook with a claimed ~50% cost reduction.
• Design-win lock-in. Once qualified into a car model, a LiDAR supplier is sticky for that model's life (re-qualification is costly). A 9M-unit backlog is contracted future revenue.
• Robotics land-grab. No.1 in robotics LiDAR by units (2025) and a broadening component catalogue create a platform option — if humanoids inflect, RoboSense sells the eyes, a hand, motors and a controller, not just one sensor.

Bargaining power — weak on both sides, and that's the problem. RoboSense sells into BYD/Geely-scale buyers who dictate price (auto OEMs are brutal on component margin) and, on robotics, into a fragmented long tail of cheap-robot makers. Its ASP fell ~52% YoY in Q1-26 — the market, not RoboSense, sets price. The moat is relative cost leadership within a deflating category, not pricing power.

Segments

No our figures on the shelf — all.

By product line (FY2025):

The tell — a mix inversion in real time: in Q4-25, robotics was RMB 347m = 49% of product-sales revenue; by Q1-26, robotics LiDAR units surpassed ADAS units for the first time. Robotics units +1,142% in FY25 (303k units) and +1,458.8% YoY in Q1-26.

Geography: China-dominant; overseas is an emerging edge — 33 design wins with 14 overseas/Sino-foreign-JV brands across APAC/Europe/North America. Overseas expansion is now structurally capped in the US by the 1260H designation (Lens 10/13).

Why it matters: revenue grew +18% while volume grew +68% → blended ASP fell ~30% for the full year, and the robotics units driving the volume are the cheapest products (lawnmower/blind-spot LiDAR at ~RMB 1,336 blended in Q1-26). The company is trading price for share and category leadership — deliberately. Whether that is empire-building or value-destruction hinges entirely on the chip cost-down (Lens 11/12).

---

Phase B — Measure performance

Earnings Result (latest prints)

FY2025 (reported ~2026-03-25) — the inflection print:

• Revenue RMB 1.941bn, +17.7% YoY (FY24 ~RMB 1.649bn implied).
• Gross margin 26.5% (FY24 17.2%); gross profit RMB 514.2m, +81.3% — the standout line; proprietary-chip adoption is the driver.
• Net loss narrowed to ~RMB 145.9m (FY24 loss RMB 482m; ~70% reduction); LPS −CNY 0.32 vs −CNY 1.11. Conflict flag: one secondary summary cited a FY25 net loss of "RMB 180m" — I lead with RMB ~146m because it reconciles arithmetically (482 × 0.30 ≈ 145) and with the −0.32 LPS; the 180m figure is unreconciled and treated as an aggregator error.
• Q4-25 = first-ever quarterly profit: revenue RMB 751m (+46.1% YoY), operating profit ~RMB 130m, net profit ~RMB 104m. This is the event that reframed the stock.

Q1-26 (the caution): revenue RMB 460m, +40% YoY; robotics LiDAR out-shipped ADAS for the first time; but gross margin fell to 21.7% (−6.7pp QoQ) and blended ASP −52% YoY to RMB 1,336 — lowest since inception, on mix shift to cheap lawnmower/blind-spot units. The maiden profit did not hold into Q1-26 at the margin line. That is the single most important fact in this dossier.

Balance-sheet flags: well-capitalised post-IPO (HK$985m) plus a Feb-2025 top-up placement of up to 22m new shares — dilutive, but funds the chip/robotics build. Exact cash, receivables, inventory and FCFFree cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices. not cleanly sourced on the empty shelf — n/a; flagged as an open item. R&D is the swing factor: Q3-25 R&D alone RMB 179.7m (+18.5% YoY), implying FY25 R&D on the order of **RMB 650–720m ** — i.e. R&D exceeds full-year gross profit, which is exactly why the company is still loss-making despite 26.5% gross margin.

Earnings Calls (sentiment trend)

No transcripts on the shelf; sentiment read from results releases and call summaries (Alpha Spread Q2/Q3-25; Futu/Gasgoo Q1-26) — all.

• Tone trajectory: from "LiDAR market-share leader" (FY24 release) → "dual-engine ADAS + Robotics" (through 2025) → "global leader in robotics technology platforms" (2025-26). Management has consciously re-narrated the company from an auto-parts supplier into a robotics/embodied-AI platform.
• Recurring phrases: "digital LiDAR," "full product matrix," "in-house chip / SPAD-SoC / VCSEL," "cost reduction ~50%," "No.1 in robotics," "embodied intelligence," "Active Camera / eyes of robots."
• What they've started saying: humanoid, dexterous hand, VTLA planning model, "4 million units" capacity.
• What to watch for them to stop saying: if "path to sustained profitability" language softens after the Q1-26 margin relapse, that is the sentiment tell.
• Net: promotional and momentum-forward — consistent with a founder-led hardware company riding a robotics-hype tape. Discount accordingly; weight the printed margin over the narrated TAM.

Comps

LiDAR pure-plays. Multiples are from market caps ÷ revenue — no clean sourced EV/Sales line exists; shown with arithmetic. Market caps are date-sensitive (Hesai executed an 8-for-1 split effective ~2026-07-10, adding noise).

The comp story writes the thesis: the LiDAR shakeout is over and the Chinese duopoly won. Western pure-plays are dead or dying (Luminar bankrupt; Ouster/Innoviz/Aeva marginal). The mass-market car LiDAR the world buys "comes almost entirely from two Chinese firms, RoboSense and Hesai". Within that duopoly, Hesai leads on revenue, ASP, long-range/robotaxi share (>40%) and — decisively — profitability, and the market pays it a richer P/S. RoboSense leads on unit volume and (now) robotics-LiDAR share, trades at a discount, and is the higher-beta bet on catching Hesai to profitability. Yole put third-party auto share ~Hesai 33% > RoboSense > Huawei > Seyond.

Stock-Price Catalysts (moves that mattered)

All; the price path is the story of hype → de-rating → re-rating attempt.

• IPO 5 Jan 2024, HK$43 — flat/weak debut (closed unchanged, −2% intraday), ~HK$20bn valuation.

---

Phase C — Judge people & books

Management

• Founder & Chairman — Dr. Qiu Chunxin (邱纯鑫). PhD in outdoor-robotics perception, Harbin Institute of Technology; founded RoboSense 2014; MIT TR35 (2017), Guangdong Tech Invention Award (2022). Deep technical founder. Stepped down as CEO effective 27 Mar 2024 — right at the IPO — to Chairman.
• CEO — Qiu Chunchao (Mark Qiu). Brother of the founder; MBA, CUHK; CEO from 27 Mar 2024; 8+ yrs in LiDAR; runs strategy/BD/ops.
• Co-founders: Liu Letian (CTO) and Dr. Zhu Xiaorui (Chief Scientist / non-exec director) — the founder's former PhD supervisor. A tight academic-lineage founding team.
• Track record: built the global unit-volume leader in automotive LiDAR from a 2014 start and took it public — genuinely impressive execution in a category that killed most entrants. Delivered the gross-margin turn (17%→26.5%) and the maiden quarterly profit.
• Skin in the game / capital allocation: founder-controlled (exact insider % n/a on the empty shelf); reinvesting hard into chips + robotics rather than buybacks/dividends — appropriate for the stage. The CEO handoff to the founder's brother concurrent with the IPO is a governance note: founder-family control persists; watch related-party and comp disclosures.
• Red flags: promotional cadence (constant milestone PR, humanoid/embodied framing); heavy share-based comp inflating opex; a dilutive placement within 14 months of listing. None disqualifying — but this is a momentum-narrating, family-controlled Chinese hardware issuer; underwrite the prints, not the press releases.
• Archetype: technical founder-operators, early-cycle, empire-building for platform scale. Implication: expect continued reinvestment/dilution over near-term profit maximisation.

Forensic Red Flags

Income statement · balance sheet · cash flow — constrained by the empty shelf (no filings to tie out); flagged as elevated-diligence, low-verifiability.

• Non-GAAP / SBC: share-based payments are a named driver of opex growth. The maiden "profit" (Q4-25 net RMB 104m) should be checked against SBC add-backs and any one-offs once the annual report ties out — verify against 2025 annual report (HKEXnews, filed 2026-04-24).
• Revenue vs cash: auto/robotics component sellers into Chinese OEMs carry long receivables and channel/inventory risk; with volume +68% and ASP −30%, watch receivables and inventory growth vs revenue — not sourceable on the stub; open item.
• ASP/mix optics: headline "revenue +18%, gross profit +81%" is real but flatters a business whose units grew 4× faster than revenue — read GM and ASP together, never GM alone.
• Capitalised R&D / dilution: confirm R&D is expensed (not capitalised) and track the diluted share count post-placement.

Regulatory findings (required sub-section). Per regulatory/regulatory-findings.md (2026-07-10): RoboSense has no SEC CIK — 0 EDGAR LR/AAER findings; not applicable to a non-US filer. Web scan for non-SEC actions:

• Pentagon Section 1260H "Chinese Military Company" designation — MATERIAL, June 2026. RoboSense was added to the DoD 1260H list (Hesai was listed Jan 2024). Cited as a "military-civil fusion contributor" with alleged PLA affiliation; described as "BYD-backed". Not a sanction and not illegal for civilian use, but it restricts future US-government procurement and complicates US OEM/capital-markets relationships (Rivian already declines to name RoboSense as its R2 supplier). This is the defining legal/geopolitical overhang.
• No SEC/FTC/DOJ financial-fraud action found. HK-listed → HKEX/SFC continuing-obligations regime, not SEC.
• Conclusion: No accounting-fraud findings surfaced; the material regulatory fact is the 1260H national-security designation (June 2026), verified via CNBC/law-firm alerts, with SEC EDGAR n/a (no CIK) as of 2026-07-10.

---

Phase D — Project & stress-test

Forward Projection

EPS is negative and the shelf is empty — I do not fabricate an EPS line. I project revenue + the path to sustained profit bottom-up from disclosed actuals (FY25 RMB 1.941bn; Q1-26 +40%). All, arithmetic shown; no our model logged (watchlist rule + task instruction).

Swing variables, ranked: (1) in-house SPAD-SoC/VCSEL yield & ramp — the entire margin thesis; (2) robotics ASP/mix — do cheap lawnmower units stay a majority, or do higher-value humanoid/Active-Camera products scale; (3) ADAS attach rate — China L2+/L3 LiDAR penetration vs camera-only pressure and Huawei in-house; (4) US 1260H drag on overseas expansion; (5) opex discipline — R&D currently > gross profit. Base call: RoboSense reaches sustained full-year profitability in FY2026–2027, one to two years behind Hesai, contingent on the chip ramp beating the price curve.

Bull vs Bear

Bull case. The LiDAR wars are over and RoboSense is one of two survivors of a global cull (Luminar bankrupt, Western pure-plays gone). It ships more LiDAR units than anyone on earth, just printed its first profit, and is early to the next S-curve — robotics — where it is already No.1 by units and is building beyond the sensor into a components platform (eyes + hand + motor + controller + planning model) exactly as humanoids move from demo to deployment. The in-house digital-LiDAR chip is a credible ~50% cost-down that mechanically lifts margin as it ramps — the same lever that carried Hesai to profit. A 9M-unit auto backlog underwrites the base, and the stock trades below its IPO price at a discount to Hesai with a consensus target ~2× spot. Buy the volume leader before the margin inflects.

Bear case (permanent-impairment risks). (1) Deflation outruns the cost-down: ASP fell 52% YoY in Q1-26 and GM relapsed to 21.7% one quarter after the maiden profit — if commodity robotics LiDAR (lawnmowers, blind-spot) stays the majority of volume, RoboSense grows units forever and never makes durable money. (2) Concentration + geopolitics: heavy BYD dependence (BYD is both top customer and a 1260H-listed shareholder) and RoboSense's own June-2026 1260H designation structurally cap the US/Western TAM and invite bans/decoupling — a Chinese sensor in "critical infrastructure" is a policy target. (3) ADAS commoditisation / substitution: camera-only autonomy (Tesla-style) and Huawei's in-house LiDAR could compress the auto attach-rate and price simultaneously. Pre-mortem (18 months out, thesis broken): the chip ramp slipped, robotics stayed low-ASP, a US action forced Rivian/JV OEMs to drop RoboSense, another placement diluted holders — the "profit" was one quarter of mix luck, and the stock re-rated to a hardware-commodity multiple.

Are multiples too high? At ~4.6–5.2× sales for a barely-profitable, ASP-deflating hardware maker, the multiple already discounts a lot of skepticism — it is not obviously expensive versus Hesai's 6–8×; the risk is earnings/margin, not a bubble multiple. Contrarian view the market is missing: the bears fixate on ASP collapse as decline, but RoboSense is deliberately dumping price to buy irreversible category leadership in robotics LiDAR while the in-house chip resets its cost floor — if the chip ramp lands, today's "margin relapse" is the trough, not the trend.

Devil's Advocate (short-seller)

Dismantling the bull case.

• Where the money really comes from: a deflating, OEM-dictated component sold to Chinese automakers who exist to crush supplier margin, plus a long tail of cheap-robot makers. RoboSense has no pricing power — the tape sets ASP (−52% YoY) and RoboSense follows.
• Concentration: BYD is the load-bearing customer and a strategic shareholder — if BYD in-sources, dual-sources to Hesai/Huawei, or is itself constrained by its own 1260H listing, a large revenue slug is exposed. The named US customer (Rivian) already won't say its name out loud.
• The moat is thinner than bulls think: "most units shipped" is a cost-curve edge, not a switching-cost moat; Hesai runs the same in-house-chip playbook and got to profit first; Huawei has scale, capital and state backing. This is a 3-way margin war, and RoboSense is the least profitable of the three.
• Worst capital-allocation optics: a dilutive placement 14 months post-IPO, heavy SBC, R&D running above gross profit, and a CEO handoff to the founder's brother at the IPO — family-controlled, momentum-narrated.
• What must hold for today's price: that the chip cost-down beats the ASP collapse and robotics mix shifts up-value and 1260H doesn't escalate into procurement bans. If FY26 revenue disappoints 20–30% (chip slip + ADAS softness), the profit story evaporates and it re-rates to a commodity multiple — plausibly −40–50%.
• Single permanent-impairment scenario, and its plausibility: a hard US decoupling (export-control escalation off the 1260H base) that forces Western OEMs out and caps RoboSense in China against Hesai/Huawei on price — medium-plausibility, rising, and the one that would structurally cap the equity.

Management Questions (ordered by information value)

1. What is the current in-house-chip (SPAD-SoC/VCSEL) mix of shipped LiDAR, the realised BOM reduction vs discrete, and the quarter you expect 50%+ of units on in-house silicon? (This single answer decides the margin thesis.)
2. Q1-26 gross margin fell to 21.7% one quarter after your first profit — was Q4-25 profitability a durable structural step or a mix/one-off? What GM floor do you underwrite for FY26?
3. What is BYD's % of revenue today, the trajectory, and your contingency if BYD in-sources or dual-sources?
4. What is the concrete operational and revenue impact of the June-2026 1260H designation on Rivian and other US/overseas programs — and your base case if export controls escalate?
5. Robotics units are +1,000%+ but blended ASP is ~RMB 1,336 — what is the ASP and gross margin by robotics sub-segment (lawnmower vs cleaning vs delivery vs humanoid), and when does mix move up-value?
6. When do you reach sustained (multi-quarter) full-year profitability, and what revenue/GM/opex bridge gets you there without another equity raise?
7. R&D appears to exceed gross profit — what is the R&D plan as a % of revenue over three years, and where is the operating-leverage inflection?
8. Beyond LiDAR, what is the realistic 3-year revenue contribution from robot components (Active Camera, dexterous hand, motors, controller) — real orders vs showcase?
9. Who fabricates your chips, and how exposed is that supply to US-China semiconductor controls?
10. In the humanoid value chain, are you a preferred multi-part supplier or one of many to the top programs — name the design-in depth.
11. On the 4-million-unit 2026 capacity — what utilisation and take-or-pay/backlog coverage underwrites it, and what is the impairment risk if robotics demand disappoints?
12. Where do you lose to Hesai and to Huawei, respectively, and what is your durable edge against each?
13. How should investors read share-based comp and the post-IPO placement — should we expect further dilution to fund the robotics build?
14. As a founder-family-controlled company (Chairman and CEO are brothers), what related-party controls and independent-board safeguards are in place?
15. If camera-only autonomy scales in China, what is your ADAS LiDAR attach-rate downside, and how fast can robotics offset it?

---