Robotics
The clear global LiDAR volume leader, now GAAP-profitable and pivoting into physical-AI sensing — but the entire equity is hostage to a single binary it cannot control: the U.S. appeals-court ruling on its DoD "Chinese Military Company" designation, with a hard June 30 2026 contract-ban trigger.
Research
The verdict
The clear global LiDAR volume leader, now GAAP-profitable and pivoting into physical-AI sensing — but the entire equity is hostage to a single binary it cannot control: the U.S. appeals-court ruling on its DoD "Chinese Military Company" designation, with a hard June 30 2026 contract-ban trigger.
Hesai Group designs and manufactures LiDAR (3D laser ranging sensors) — the "eyes" that let cars and robots perceive depth. It is a Cayman-incorporated holding company operating through PRC subsidiaries, HQ in Shanghai, dual-listed: ADSs on Nasdaq (HSAI) and Class B shares on the Hong Kong Stock Exchange (2525.HK) . As of 2025-12-31 there were 156,145,167 ordinary shares (129.1M Class B + 27.0M Class A, dual-class with founder super-voting Class A) .
How it makes money. Two end-markets, one product family:
. The flagship **AT series** alone was **63.0% of total revenue** in 2025 (60.9% in 2024, 37.0% in 2023) — extreme single-product concentration.Revenue is overwhelmingly product (LiDAR products = RMB 2,973.3M of RMB 3,027.6M total in 2025, i.e. 98.2%) ``; the rest is engineering/NRE services (RMB 36.1M, shrinking) and a near-dead gas-detection legacy (RMB 3.2M).
Customers / suppliers / competitors. Customers are OEMs and Tier-1s, historically very concentrated (see Lens 4). Disclosed marquee names: a leading Chinese NEV maker (long Hesai's biggest, widely identified as Li Auto), plus design wins with top European automakers and, per Q1 2026, Mercedes-Benz as a strategic LiDAR partner . Suppliers: no single supplier ≥10% of purchases — supply is diversified, a genuine strength. Competitors: RoboSense, Seyond, Huawei, Valeo/Continental (ADAS); Ouster, Innoviz, Luminar (Western).
Contract structure. Purchase-order based, specifying volumes/prices/delivery; terminable by the customer on breach/insolvency. There is no take-or-pay; orders can be cancelled, and one was — the large U.S. OEM suspended a project in 2024, paying a one-off compensation (see Lens 5 / Lens 10) ``. Net revenue mix is volume × ASP, and ASP is falling fast (Lens 4).
Upstream → Hesai → end customer, named where the filing names them:
n/a — not named in 20-F).. Capacity is being scaled to **>4 million units/year in 2026** .Chokepoints / single-source risk. The 20-F's own risk language: shortages or price spikes in raw materials and "sole or limited-source components" could be material; developing alternative sources is "time-consuming, difficult, and costly" . **Tariffs** are an explicit cost line — 2024 cost-of-revenue fell partly because U.S. sales (and thus U.S. tariffs) dropped . The dominant chokepoint is not a physical input — it is U.S. policy (the 1260H designation, outbound-investment rules, potential export controls), covered in Lens 10.
. It also claims **>40% of the long-range automotive LiDAR market** and **No.1 across major robotics submarkets** (humanoid, quadruped, robotaxi, robovan, mower) .Bargaining power. Over suppliers: strong (diversified, no >10% supplier). Over customers: weak — top-5 customers were 55.8% of revenue in 2025 and OEMs hold the whip hand on price (the filing warns of "substantial continuing pressure from automotive OEMs" on pricing) ``. The moat is real on the cost/scale axis; it is thin on the customer-power axis.
By product type (RMB; ``):
| Line | 2023 | 2024 | 2025 | 2025 vs 2024 |
|---|---|---|---|---|
| LiDAR products | 1,735.3M | 1,946.8M | 2,973.3M (US$425.2M) | +52.7% |
| Engineering/NRE services | 100.5M | 100.3M | 36.1M | −64% |
| Gas detection | 26.9M | 15.4M | 3.2M | −79% |
| Other services | 11.6M | 10.8M | 8.5M | −21% |
| Total net revenue | 1,877.0M | 2,077.2M | 3,027.6M (US$432.9M) | +45.8% |
By application / units (``):
The story the segments tell: revenue +46% is the net of a brutal ADAS price war (ASP −51% YoY) overwhelmed by 3.2× unit volume. The high-margin NRE services line — which juiced 2024 — is collapsing, and management flagged that the 2025 gross-margin dip (42.6%→41.8%) was specifically "due to decrease in revenues from high-margin non-recurring engineering services" ``. So the mix is shifting away from soft, lumpy, high-margin service dollars toward hard, scalable, lower-margin hardware dollars — healthier long-term, optically dilutive to margin short-term.
By geography: the filing does not break out a clean revenue-by-region table in the extracted MD&A; it confirms revenue is "both in China and globally," that U.S. sales fell in 2024 (tariff-driven), and that cash is 69.4% Chinese Mainland / 26.9% Hong Kong ``. Precise geographic revenue split: n/a — not in extracted filing tables.
FY2025 — the year Hesai turned profitable ``:
Q4 2025 (the print): revenue RMB 1,000.5M (US$143.1M), shipments 631,095 units, net income RMB 153.2M (US$21.9M) ``.
Q1 2026 (latest actual): revenue RMB 680.6M (US$98.7M, +29.6% YoY), shipments 471,723 (+140.9% YoY), GAAP net income RMB 18.3M (4th consecutive profitable quarter; non-GAAP RMB 47.7M, +452.9%) ``. Growth decelerated from +46% to +30% and quarterly net income stepped down sharply from Q4 — partly seasonality, partly the absence of Q4's one-offs.
Balance-sheet flags (``):
Market reaction: despite the profitability inflection, HSAI traded ~$17–22 in June 2026, ~$17.15 on 2026-06-16 `` — below where it sat before the March-2025 short report. The market is not paying for the operational turn; it is discounting the 1260H overhang. That gap is the trade.
No transcripts on disk (transcripts/ empty). From ``:
. The recurring new phrase is **"physical AI"** / **"eyes and muscles"** — management now positions robotics components as a business that "could meet or surpass the LiDAR segment's scale over the next five years" .Net sentiment: management is increasingly confident and increasingly promotional — the pivot to "physical AI" is both a genuine strategic expansion and a narrative away from the ADAS price war and the DoD cloud. Treat the five-year "could surpass LiDAR" robotics claim as an aspiration, not a forecast.
LiDAR is pre-consolidation; multiples are noisy and the Western peers are loss-making, so EV/Sales is the only common yardstick. All multiples ``, June 2026:
| Company | Ticker | Mkt cap | 2025 revenue | EV/Sales | P/E | Notes |
|---|---|---|---|---|---|---|
| Hesai | HSAI | ~$3.0–3.5B `` | US$433M `` | TTM 5.6× / FWD 3.8× `` | TTM 44× / FWD ~29–44× `` | Only profitable scaled pure-play |
| RoboSense | 2498.HK | n/a | ~RMB 1.94B (~US$267M) `` | n/a | n/a (just hit 1st profitable quarter) | ~912k units; closest scaled rival |
| Innoviz | INVZ | n/a | US$55.1M `` | n/a | n/m (loss) | Sub-scale; BMW/VW programs |
| Ouster | OUST | n/a | ~US$260M `` | n/a | n/m (loss) | Industrial/robotics tilt |
| Luminar | LAZR | n/a | n/a | n/a | n/m (loss) | Restructured; cutting $50–65M cost |
5-yr avg ROE: n/a (Hesai was loss-making until 2025; a 5-yr average is meaningless). Read: Hesai is the only member of this peer set with a real P&L, scaled units, and a net-cash balance sheet; on EV/Sales (~3.8× FWD) it trades at a premium to the sector median (~1.2× ``) but a discount to where a profitable, 40%-share category leader growing 30–46% would trade absent the geopolitical tag. The multiple is a geopolitics-adjusted number, not a fundamentals number.
The tape says this stock trades on geopolitics and short reports, not earnings ``:
Pattern: the dominant variable is the U.S.-government designation; secondary is short-seller credibility attacks; fundamentals are a distant third. This is a policy-binary stock with an operating business attached, not the reverse.
— founders retain hard control. CFO is **Peng Fan**. Precise insider %: n/a — insider-transactions.csv not present.Income statement:
. **The 20-F discloses exactly such a payment and says it openly:** "a one-off project-based payment of RMB 203.3 million from a leading global OEM headquartered in the United States in 2024," booked in *other operating income*, following that customer's termination of a March-2023 contract . So the item is disclosed in the audited filing — the short thesis that it was hidden does not survive the 20-F; the legitimate residual question is whether one-off comp should be normalized out of "profitability" (it should). Note 2024 was a net loss anyway, so the fee did not manufacture a profit.Balance sheet:
; net cash from operations was only RMB 117M (US$16.7M) despite RMB 436M net income — **operating cash conversion is weak**, the classic "are the earnings real / are customers paying" question for a China hardware name. The encouraging counter-signal: the largest customer's share of receivables *fell* from 41.3%→23.5%→**5.3%** (2023→24→25) , i.e. receivables are diversifying, not concentrating.Structure: No VIE. Hesai is a Cayman holdco owning PRC operations through WFOE-style direct equity, not contractual-control entities — materially cleaner than the typical China ADR. Auditor is **Deloitte Touche Tohmatsu CPA (Shanghai), PCAOB ID 1113**; the PCAOB's 2021 inspection-denial determination was vacated in 2022, so HFCAA delisting risk is currently dormant (but the filing warns it could resurface) ``.
Regulatory findings (required sub-section):
; (2) a **new Rosen Law investigation** triggered by the March-2025 Blue Orca allegations . Both shareholder-driven, not agency enforcement; outcomes pending.Net forensic read: the accounting is more defensible than the short thesis claims — clean (no VIE), Big-4-audited, PCAOB-inspected, with the disputed one-off disclosed. The real risks are (a) weak operating-cash conversion / receivables, and (b) the off-balance-sheet political risk that no amount of accounting cleanliness can cure.
Anchored on management's own guidance + the latest actuals (NOT logged as a Brier forecast — --watchlist rule):
Three-year EPS path (``, ADS = 1 ordinary share; ~156M shares):
Bull: robotics/embodied-AI inflects faster, ASP stabilizes as the price war eases post-MIIT-mandate, global (non-China) design wins (Mercedes) ramp → FY2028 EPS >$1.80. Bear: the 1260H ruling goes against Hesai and/or broader sanctions follow, U.S./EU OEMs freeze Hesai out, ADAS price war compresses margin below 35%, robotics stays a rounding error → FY2026 EPS <$0.40 and multiple de-rates hard.
Inputs labeled; outputs ``. Every line is sensitive to the ASP-decline assumption and the binary policy outcome — the bear and base diverge less on operations than on whether the company keeps full market access.
Bull case. The global LiDAR volume leader just proved the unit economics work: first GAAP profit, ~40% share, 3.2× volume, net-cash US$1.1B balance sheet, R&D leverage from 41%→26% of revenue. China's MIIT mandate putting L2 ADAS (LiDAR-inclusive) into the standard from Sept 2025 is a structural demand tailwind, and Hesai is the cost leader positioned to capture it. The second act — robotics/physical-AI sensing (Unitree humanoids, robotaxis, the "eyes and muscles" thesis, NVIDIA-ecosystem robotics) — is a genuine new TAM where Hesai already ranks No.1 across submarkets and where ASPs are higher and the price war hasn't arrived. If the DoD overhang clears, the stock re-rates from a geopolitics-discounted ~3.8× EV/Sales toward a growth-leader multiple — that alone is a double, before any robotics optionality.
Bear case (permanent-impairment risks). (1) The 1260H designation becomes permanent and metastasizes. It already cost a >30% one-day drawdown and, per the company, has "disrupted existing and potential customer relationships globally, in particular in the United States." If the Court of Appeals affirms and broader sanctions/export-controls follow, Hesai is structurally locked out of the U.S. and possibly allied OEMs — a permanent TAM amputation, not a multiple wobble. (2) The ADAS business is a commodity price war — ASP fell 76% in two years; if volume growth ever stalls while ASP keeps falling, the model inverts fast (operating margin is only 5.6%). (3) Single-product / single-geography concentration — AT series is 63% of revenue and the customer base is overwhelmingly Chinese OEMs in a brutally competitive domestic NEV market.
Pre-mortem (18 months out, thesis broke): the Court of Appeals affirmed the DoD designation in 2026; the June-30-2026 contract ban and follow-on Western-OEM caution froze Hesai's ex-China expansion; a Chinese NEV demand air-pocket + continued ASP erosion pushed operating margin back toward zero; robotics revenue stayed sub-10% of total. The stock halved from an already-depressed base.
Multiples too high? No — if you believe the designation clears, ~3.8× FWD EV/Sales for a profitable 40%-share leader growing 30%+ is cheap. The multiple is low precisely because the binary is unresolved. You are not paying for growth; you are pricing a coin-flip.
Contrarian view (what the market refuses to see): the market treats Hesai as un-investable because of the military tag, yet the 1260H list carries no securities-trading or investment prohibition (it bars DoD/DHS procurement, a market Hesai barely serves) — and the District Court itself "recognized there is no evidence linking Hesai's products to military use." If the appeal even narrows the designation, the gap between the operating reality (profitable category leader) and the equity (priced as a fraud-suspect pariah) snaps shut. The asymmetry is the thesis.
Dismantling the bull case:
What must hold for today's price: that the appeal does not escalate into a trading/investment ban, that ADAS volume keeps outrunning ASP decline, and that robotics becomes a real margin contributor. Knock 20–30% off the growth and add an adverse ruling, and the equity halves.
The only profitable humanoid maker and #1 by volume — but the float is a Shanghai-only embodied-AI bet whose revenue is 74% lab demos, not labor replacement, and whose Western TAM is being legislated to zero.
A profitable EV maker priced as a solved-autonomy robotics company — the car business is shrinking, the GAAP profit prop (reg credits) is going to zero, and the entire ~190x multiple now rents on robotaxi + Optimus execution that is real but years behind the price.
A genuinely differentiated case-handling robotics moat wrapped around an ungovernable accounting-and-concentration core — adverse ICFR opinion + active SEC whistleblower probe + >90% Walmart revenue means the multiple is pricing a clean compounder that the filings say does not yet exist.