Unitree Robotics
Vertically integrated robot hardware sales (quadrupeds, humanoids, actuators/components) at aggressive price points; volume-led cost curve; nascent software/embodied-AI layer
~3.5x step-up from the Jun 2025 Series C mark to the Jul 2026 approved IPO minimum in ~13 months; debut may price above the minimum given A-share humanoid scarcity dynamics. Series C mark variance (CNY 12B vs 12.7B) flagged per sourceNote.
Earnings, margins, COGS & capex
Prospectus (audited) shows revenue of CNY 159M (2023), CNY 392M (2024), CNY 1.70B (2025, +333%), with main-business gross margin expanding 44.2% -> 56.4% -> 60.1% as volume scaled and humanoids mixed in. FY2025 reported net profit CNY 278M (Caixin; some accounts 288M - the gap to adjusted is a one-time CNY 349M share-based-compensation charge); adjusted net CNY 591M (some coverage: 600M, +674% YoY). Q1 2026 revenue CNY 422.8M (+68% YoY) but adjusted net profit fell 52% YoY to CNY 40.3M (vs CNY 84.8M Q1 2025) as Unitree front-loads embodied-AI model R&D and brand/sales spend. H1 2026 guide: revenue CNY 1.05-1.13B (+35.6-45.4%) with adjusted net CNY 236-283M (-6.4% to -22% YoY). Humanoids became the largest segment - 51.8% of FY2025 revenue (1.9% in 2023, ~30% in 2024), quadrupeds ~30% (down from ~65% in 2024), components/accessories the balance.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~40¢ is cost of goods and ~0¢ operating expense, leaving ~60¢ of operating profit (~16¢ net).
Revenue trend
Margins
Expanding: 44.2% (2023) -> 56.4% (2024) -> 60.1% (2025)
First clearly profitable year; depressed by one-time CNY 349M SBC charge
+674% YoY; Q1 2026 -52% YoY on R&D/brand spend
Compression phase; H1 guide (-6.4% to -22%) implies sharp sequential recovery
COGS structure
Vertically integrated - in-house joint motors/actuators, reducers, controllers and sensors keep BOM cost industry-low and underpin disruptive pricing (Go2 quadruped from ~$1,600; G1 humanoid from ~$16k). Humanoid ASP fell ~72% from ~CNY 593k (2023) to ~CNY 168k (2025) while gross margin still expanded to 60.1% - costs falling faster than prices. Key bought-in items include compute (NVIDIA-class edge modules) and semiconductors.
Capex
Stepping up from asset-light roots: IPO proceeds (CNY 4.2B) earmarked for four programs - intelligent robot model R&D (~CNY 2B / ~$300M over three years), robot body R&D, new product development, and a smart-manufacturing base build-out.
Latest earnings
n/a - no public consensus pre-listing
H1 2026: revenue CNY 1.05-1.13B (+35.6-45.4% YoY); adjusted net CNY 236-283M (-6.4% to -22% YoY, narrowing from Q1's -52%)
- FY2025 humanoid shipments
- 5,511 units
- Cumulative quadrupeds (2022-Sep 2025)
- 30,000+ units
- FY2025 gross margin
- 60.1%
- IPO raise / min listing valuation
- CNY 4.2B / CNY 42B (~$6.2B)
Growth drivers
- Humanoid mix shift — 1.9% of revenue (2023) -> ~30% (2024) -> 51.8% (FY2025); 5,511 humanoids shipped in 2025 - top global shipper
- Quadruped franchise (60-70% global share; 30,000+ units cumulative 2022-Sep 2025) funding the humanoid ramp
- Price-down/volume-up strategy opening education, research, inspection, entertainment and developer markets
- Embodied-AI model R&D (UnifoLM and successors) to move up from hardware into the software/autonomy layer
- Components/actuator sales to third parties as a picks-and-shovels line
- IPO capital (CNY 4.2B) removing manufacturing-capacity and R&D funding constraints
Bull & bear
The only profitable, volume-shipping legged-robot company on earth, entering the humanoid super-cycle with the lowest cost curve, 60% gross margins, a fortress cap table, and a freshly approved CNY 4.2B war chest - the Ford of humanoids at the Model T moment.
- Real, audited hyper-growth: revenue +147% then +333%, with gross margin expanding to 60.1% - growth is not bought with losses, unlike every Western humanoid peer
- Cost leadership is structural (in-house actuators/sensors): humanoid ASP fell ~72% in two years while GM expanded - Unitree prices rivals out (G1 at ~$16k vs six-figure Western BOMs) and still earns a ~16% reported net margin
- Shipment scale (30,000+ quadrupeds, 5,511 humanoids in 2025) creates a fleet-data and manufacturing-learning flywheel competitors cannot quickly replicate
- Humanoid revenue already >50% of mix and growing - a humanoid pure-play with a profitable quadruped cash engine, a unique combination
- IPO proceeds (~CNY 2B to embodied-AI models) attack the one acknowledged gap (software/autonomy) from a position of hardware strength
- Scarcity value: first A-share humanoid-core listing; fastest STAR review under the pre-review mechanism (73 days from acceptance to committee approval; 104 days to effective registration) signals state-level strategic backing
A ~$6B+ minimum valuation on ~$240M of revenue whose margin is already rolling over, in a category where the killer app is still demos - with Tesla, subsidized domestic clones, and geopolitics all pointed at its margin pool.
- Valuation: CNY 42B minimum is ~25x FY2025 sales and ~150x FY2025 reported net profit (~71x adjusted) - priced for flawless multi-year hyper-growth before the first trade
- Profit inflection is negative right now: Q1 2026 adjusted net -52% YoY and the H1 guide confirms a down year for earnings even as revenue grows
- Unitree's own prospectus flags intensifying competition/price-war risk and names Tesla Optimus plus Chinese carmakers/electronics giants as incoming competition - 60% GM is a magnet
- Demand quality: 74% of humanoid revenue is research/education and much of the rest entertainment (dancing/boxing robots), only ~9% of humanoid shipments industrial - not ROI-proven labor; replacement cycles unknown
- Thin IP moat (262 global patents, 20 domestic invention patents) invites both litigation and fast-follow cloning at Shenzhen speed
- Geopolitical ceiling: Western markets (a large share of long-run TAM) may be closed by security-driven procurement rules, leaving Unitree fighting the price war at home
- US investors largely cannot own the A-share; access is indirect (ETFs/index vehicles), and no ADR exists - liquidity and governance disclosures follow CSRC, not SEC, standards
What it is worth
Reverse-DCF sanity + pre-IPO marks + scarcity-adjusted comps (no trading price exists yet)
CNY 20-30B (~$3-4.5B)
if the earnings down-year extends into 2027, the price war compresses gross margin below 50%, or Western market access closes - still ~2x the Series C mark, showing how much hope is already in the private price
CNY 42-60B (~$6-8.5B)
prices at/modestly above the approved minimum; ~25-35x trailing sales digested by 35-45% guided growth and the first clean public-company prints
CNY 80-100B+ (~$11-14B)
if the debut catches the A-share humanoid frenzy and 2026-2027 revenue compounds >80% with humanoid mix >60% - scarcity premium on the only profitable humanoid pure-play
Marks to anchor on: Series C post-money ~CNY 12B (~$1.7B, Jun 2025, per Forbes; some accounts cite CNY 12.7B with secondary trades above CNY 15B); approved minimum listing valuation CNY 42B (~$6.2B, Jul 2026, implied by >=40.45M new shares at a >=10% float raising CNY 4.2B) - a ~3.5x step-up in ~13 months. At CNY 42B the stock starts at ~25x FY2025 sales and ~150x FY2025 reported net profit (~71x adjusted). That price implies roughly 40-50%+ revenue CAGR for 5+ years with gross margin holding near 60% despite prospectus-flagged price-war risk - achievable only if the humanoid TAM inflects on schedule AND Unitree keeps share and margin simultaneously. Comps: UBTech (9880.HK) trades as the loss-making public benchmark; Western privates (Figure) carry richer sales multiples on far less revenue, so Unitree is arguably the cheapest real revenue in the category even at the minimum - but A-share scarcity/frenzy dynamics (30+ robot-concept stocks hit daily limit-up on its approval) could push the debut well above fundamental support. Not financial advice; US investors have no direct access (no ADR), only indirect ETF/index exposure.
SWOT
Strengths
- Only legged-robot maker at scale that is already profitable (FY2025 reported net ~CNY 278-288M) - peers burn cash
- Vertical integration (motors, reducers, controllers, sensors in-house) -> 60% gross margin at disruptive price points
- Dominant quadruped share (60-70% global) and #1 humanoid shipment volume (5,511 in 2025)
- Founder-led engineering culture — Wang Xingxing holds 23.8% directly (~33% economic interest) with 68.8% of pre-IPO voting rights
- Blue-chip cap table — Meituan (9.65%, largest external holder), HongShan/ex-Sequoia China (7.11%), Matrix Partners China (5.5%), plus Tencent, Alibaba, Ant, Geely, China Mobile's fund
Weaknesses
- Small absolute revenue base (~$235-250M) versus the valuation being asked (CNY 42B+ minimum)
- Profit compression underway: Q1 2026 adjusted net -52% as model R&D and brand spend ramp
- Thin patent moat by its own admission (262 global patents, only 20 domestic invention patents) - prospectus flags IP-infringement-claim exposure
- Software/foundation-model capability unproven vs compute-rich rivals; today's humanoids are mostly teleop/choreographed use cases
- Customer base skews research/education (74% of humanoid revenue) and entertainment - repeatable industrial ROI deployments still early (~9% of humanoid shipments)
Opportunities
- Humanoid TAM inflection — Morgan Stanley-class forecasts of ~$5T annual humanoid revenue by 2050 scenarios; first-mover volume advantage compounds via data + cost curve
- STAR Market IPO (CNY 4.2B) funds manufacturing base + ~CNY 2B embodied-AI model program
- Component/actuator supply to the broader robot industry as a picks-and-shovels line
- Services, logistics, inspection and eventually consumer/household segments as autonomy matures
- First A-share humanoid pure-play scarcity premium; index/ETF inclusion flows (e.g. STAR-tracking vehicles)
Threats
- Tesla Optimus scaling (named in Unitree's own risk factors) plus deep-pocketed Chinese entrants (carmakers, electronics majors) -> prospectus-flagged price-war risk
- US-China tech tension — export controls, potential US procurement bans on Chinese robots, tariff walls limiting Western expansion
- Security/trust concerns in Western markets (government scrutiny of Chinese-made robots) capping overseas mix
- Humanoid hype cycle deflating if commercial ROI lags - sector re-rate would hit the multiple hardest
- Chip supply dependence (advanced compute) exposed to sanctions regimes
Moats, dependencies & bottlenecks
Moats
motors, reducers, sensors) 60.1% GM at the lowest price points in the industry despite a 72% humanoid ASP cut in two years; hardest moat to copy quickly
Manufacturing scale + volume learning (30k+ quadrupeds cumulative, 5,511 humanoids in 2025) Only player shipping legged robots at four-digit annual volume profitably
durable in niche 60-70% global share; SDK/community lock-in in research and education
Real distribution asset but bought partly with marketing spend that dented Q1 2026 profit
262 global patents, only 20 domestic invention patents; prospectus itself flags infringement-claim exposure
~CNY 2B of IPO proceeds aimed here; currently behind compute-rich rivals
Dependencies
e.g. NVIDIA-class SoCs) Export-control exposure for training and possibly edge compute under US sanctions regimes
STAR listing and 'hard-tech' policy tailwind cut both ways - support now, direction risk later
74% of humanoid revenue is research/education; industrial ROI demand must materialize to sustain growth
~33% economic interest, 68.8% pre-IPO voting rights) Concentrated control post-IPO; succession and governance untested as a public company
Deep local supply base is actually an advantage; specific vendors not disclosed
Security scrutiny of Chinese robots could cap the highest-value overseas TAM
Advantages
- Lowest-cost producer in legged robotics with proof at 60% gross margin
- Only profitable pure-play at shipping scale - self-funding before the IPO even prices
- First-mover A-share scarcity + fastest regulatory path under the STAR pre-review mechanism (104 days acceptance-to-registration) implies strategic-asset status
- Dual engine: quadruped cash cow + humanoid growth spearhead + components line
- Blue-chip strategic investors (Meituan, HongShan, Matrix Partners China, Tencent, Alibaba, Ant, Geely) as channel and ecosystem allies
Weaknesses
- Earnings quality deteriorating near-term (Q1 2026 adjusted net -52%; down-year guided for H1)
- Thin patent estate for a company inviting global IP fights
- Unproven software/autonomy stack vs Tesla/Figure compute advantage
- Revenue concentration in early-adopter segments — industrial repeat-purchase evidence thin (~9% of humanoid shipments)
- No US listing/ADR - governance, disclosure cadence and investor access all A-share-bound
Bottlenecks
- Embodied-AI autonomy: hardware outruns software - robots still need teleoperation/scripting for most tasks
- Manufacturing capacity ahead of the humanoid ramp (the IPO's smart-manufacturing base addresses this)
- Dexterous manipulation (hands) - industry-wide unsolved constraint on real labor substitution
- Talent competition for robotics/ML engineers vs deep-pocketed domestic AI labs and carmakers
- Export controls on high-end training compute for foundation-model ambitions
Top signals & trends
Top signals
Watch whether it prices above the CNY 42B minimum and how the A-share humanoid complex re-rates around it
First test of guidance credibility as a public company
Mix shift continuing = thesis on track
negative near-term · Q1 2026 -52% YoY; guide implies narrowing - confirm the trough is H1 2026
Named in Unitree's own risk factors; the reference competitor for Western TAM
Determines whether overseas TAM stays open
Prospectus-flagged price-war risk; watch G1/Go2 price cuts and GM trend
Trends
strong tailwind · Unitree is the shipment-volume leader into the cycle
Fastest IPO review under the STAR pre-review mechanism; state alignment lowers domestic funding cost
Compute import limits + Western market access limits, both directions
Expands TAM but compresses ASPs (Unitree's own humanoid ASP -72% in two years) - favors the cost leader, punishes margins sector-wide
tailwind if caught, threat if missed · The ~CNY 2B model program is Unitree's bet not to be commoditized into a body-shop
Structural demand pull for labor-substituting robotics in Unitree's home region
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Edge AI compute (Jetson-class) and Isaac simulation stack used across the legged-robot industry; also training-compute dependency
Upstream fab for the advanced silicon in robot compute (indirect)
Depth cameras widely used in legged-robot perception stacks
Chinese LiDAR maker; Unitree also builds in-house LiDAR (L1/L2) - sensing is partially internalized
Representative China-domestic battery supply base; specific battery vendor not disclosed in prospectus excerpts
Research universities & national labs (global) Core buyer base - research/education is 74% of humanoid revenue (prospectus); platforms Go2, B2, G1, H1
Spring Festival Gala performances, robot boxing/dancing showcases drive brand-led sales
Quadruped patrol/inspection deployments (power, construction, security); ~9% of humanoid shipments are industrial
Low-price Go2 (~$1,600) seeded a long-tail developer install base
Named in Unitree's prospectus risk factors; vertically integrated with massive compute + manufacturing scale; the benchmark Western humanoid program
US private humanoid leader (multi-billion valuations; Microsoft/NVIDIA-linked backing); targets commercial/industrial deployments
Atlas humanoid + Spot quadruped incumbent; strongest legacy engineering brand; Hyundai-owned
US private; Digit humanoid in warehouse pilots (GXO, Amazon trials); ahead on US logistics deployments
HK-listed Chinese humanoid peer (Walker S); the other public-market benchmark for the category; loss-making at scale
Mainland-China private rival scaling humanoid shipments fast; context only, not a coverage/ownership call
Shanghai-based humanoid maker (GR series) named among Unitree's humanoid competitors; context only
Consumer-electronics giant with robotics program and BOM-cost DNA
Swiss private quadruped maker focused on industrial inspection; premium-priced Western alternative in Unitree's quadruped core
US-based, Korean-owned (LIG Nex1) defense quadruped maker; wins the security-sensitive Western government segment closed to Unitree