Author: LBank Research Analyst: Steven.fu
Disclaimer: This report is compiled and analyzed from publicly available information and is intended solely for informational and research purposes. It does not constitute investment advice, a securities recommendation, a trading instruction, or any guarantee of returns. The company's operations, valuation, market price, and consensus expectations discussed herein may change over time. Readers should independently verify the data and make their own decisions.
1. Core Conclusion: The Startup That Sells Robots Best Meets the Least Forgiving IPO Valuation
Overall view: Unitree Robotics is one of China's general-purpose robotics companies with the strongest commercialization quality. It has established genuine leadership in full-system engineering, motion control, cost, and distribution. However, the ¥ 150.80 offer price implies a market capitalization of approximately ¥ 60.993 billion, 103.25x 2025 adjusted net profit and 35.89x revenue. The valuation has already priced in several years of high growth, leaving the current risk-reward skewed to the downside. The primary variable is whether adjusted net margin can stabilize after price cuts used to drive volume, with industrial applications generating sustainable repeat purchases.
-
The company is high quality, and the evidence of leadership comes from deliveries, profit, and cash rather than stage demonstrations. In 2025, Unitree shipped 5,511 humanoid robots and recognized sales of 5,215 units, while quadruped robot sales reached 23,037 units. Humanoid revenue was ¥ 868 million, or 51.78% of core-business revenue. The company's 60.44% consolidated gross margin, ¥ 591 million adjusted net profit, and ¥ 670 million operating cash flow were all materially better than the still-loss-making UBTECH, DOBOT, and Leju Robotics. The top five customers accounted for only 12.08% of revenue, indicating low dependence on any large customer and demonstrating standardized, global sales capabilities.
-
The essence of growth is “lower prices x higher volume x a broader product portfolio,” with volume contributing far more than price. In 2025, humanoid robot sales volume increased 1,165.78% year over year, while the average selling price fell from ¥ 260,400 to ¥ 166,400; revenue still grew 708.81%. Quadruped robot sales increased 222.83%, while average selling price declined 6.27% and revenue rose 202.60%. The transmission chain is clear and reproducible: mid-sized products such as G1 lowered the entry barrier -> humanoid sales increased by 4,803 units -> humanoid revenue increased by ¥ 761 million -> humanoid gross profit contribution rose to ¥ 548 million, or 54.40% of core-business gross profit -> group adjusted net margin rose to 34.77%. The low-price strategy expands the ecosystem and market share, but it also means future profit will increasingly depend on scale-driven cost reduction rather than high prices for scarce products.
-
Profit and cash flow were both strong in 2025, but the profit trajectory weakened ahead of revenue in 2026. Based on operating cash flow minus cash paid to acquire long-term assets, simplified free cash flow was approximately ¥ 620 million in 2025, or about 104.9% of adjusted net profit. Cash, financial assets held for trading, and debt investments totaled ¥ 2.584 billion at year-end 2025, with almost no interest-bearing borrowings. The contrary evidence is equally clear: in 1Q26, revenue increased 68.49% year over year, but adjusted net profit fell 52.55% and operating cash flow declined 85.65%. The company estimates that 1H26 revenue will grow 35.62%-45.41%, while adjusted net profit will still fall 6.43%-21.97%. Revenue growth no longer translates automatically into higher profit, which is the most dangerous fundamental change for the current valuation.
-
The offer price pays too much of the industry-leader premium up front. The ¥ 60.993 billion IPO valuation is 4.80x the ¥ 12.7 billion post-money financing valuation from June 2025. It corresponds to 103.25x adjusted net profit, 219.23x attributable net profit, and 35.89x sales based on 2025 figures. The sales multiple is 81.8% above the 19.74x average for UBTECH and DOBOT. Even mechanically annualizing the 1H26 revenue and adjusted profit ranges results in a price-to-sales ratio of approximately 27.0x-29.0x and an adjusted P/E of approximately 107.8x-129.2x. The online subscription's effective oversubscription multiple of approximately 8,288.82x on August 10 and final allocation rate of 0.01809759% prove scarcity and investor enthusiasm, not that future cash flow can support this price.
-
The threshold for reversing the negative-leaning view is high but quantifiable. If full-year 2026 revenue exceeds ¥ 3.0 billion, adjusted net profit exceeds ¥ 800 million, consolidated gross margin remains above 58%, and at least one manufacturing or service application produces a disclosed, repeat bulk order, faster growth could absorb the 103x baseline earnings multiple. If humanoid average selling prices fall by more than another 20%, consolidated gross margin drops below 55%, full-year adjusted net profit fails to grow, or the embodied foundation model remains limited to pilots and demonstrations, the current IPO valuation will move from “an expensive industry leader” to “using mature earnings to pay for a long-dated technology option.”
2. Company Overview, Business Mix, and Core Operating Metrics: Quadrupeds Provide the Manufacturing Base, While Humanoids Have Taken Over the Profit Pool
Founded in 2016, Unitree Robotics is a high-performance, general-purpose robotics company spanning humanoid robots, quadruped robots, robotic components, and embodied-intelligence models. It primarily earns revenue from selling complete robots and components rather than subscriptions or per-task fees. The company develops robot architecture, core components, motion control, and final assembly in-house. It procures or outsources non-core activities including mechanical parts, electronic components, some surface-mount assembly, injection molding, and machining, then sells standardized products to research and education, commercial and consumer, inspection and rescue, and application-development customers through direct sales, distributors, e-commerce, and its developer ecosystem.
Quadruped robots were Unitree's engineering starting point. The lower-priced Go series serves education, research, consumer entertainment, and smart services, while the B and A series and wheeled-legged products target electric power, energy and chemicals, firefighting, surveying, and public services. The significance of quadrupeds extends beyond their ¥ 698 million of revenue: motors, reducers, joint modules, energy management, and motion control can be reused in humanoid products. This both shortens product iteration cycles and helps Unitree extend humanoid pricing from the early H1's premium range down to the developer market accessible to G1 and R1.
Humanoid robots have become the largest business. H1 and H2 are positioned as full-size, high-performance platforms, while G1 and R1 cover mid-sized and small-to-mid-sized platforms. EDU versions support secondary development, whereas basic versions walk, run, dance, and interact according to preset instructions. At this stage, customers are still primarily buying a “mobile, programmable robot body” for research, teaching, algorithm validation, performances, data collection, and application pilots. Unitree has not yet deployed its proprietary general-purpose embodied foundation model across products at scale, so not all humanoid revenue can be treated as mature autonomous-labor revenue.
Robotic components include joint modules, dexterous hands, robot arms, lidar, high-compute kits, batteries, navigation systems, and control terminals. This business is small but constitutes an important part of the cost moat. In-house development and production of core components enabled Unitree to retain a 63.18% humanoid gross margin in 2025 even as humanoid average selling prices fell 36.10%. It also creates a positive feedback loop of lower hardware prices -> higher sales volume -> more data and developers -> lower costs for the next product generation.
|
2025 business
|
Products, customers, and pricing/contract model
|
Revenue
|
YoY
|
Share of core-business revenue
|
Gross margin/key KPI
|
Research view
|
|
Humanoid robots
|
H1/H2 and G1/R1; universities, developers, technology companies, education, performances, and application-validation customers; primarily sold as complete units and configurations
|
¥ 868 million
|
+708.81%
|
51.78%
|
63.18% gross margin; sales of 5,215 units and shipments of 5,511 units
|
Has taken over the revenue and profit pools, but commercial demand still skews toward development and pilots
|
|
Quadruped robots
|
Consumer-grade Go and industrial-grade B/A series; research and education, inspection, firefighting, surveying, and consumer entertainment customers; sold as units, configurations, and solutions
|
¥ 698 million
|
+202.60%
|
41.62%
|
56.72% gross margin; sales of 23,037 units
|
Foundation for scaled manufacturing and component reuse; industrial-grade mix improves gross margin
|
|
Robotic components
|
Joint modules, dexterous hands, robot arms, lidar, high-compute kits, and supporting components; sold separately or configured with complete robots
|
¥ 104 million
|
+132.92%
|
6.19%
|
59.36% gross margin
|
Small revenue base, but determines cost, iteration speed, and developer-ecosystem stickiness
|
|
Other
|
Other products and services, including fitness pumps
|
¥ 7 million
|
+30.67%
|
0.41%
|
33.53% gross margin
|
Non-core and almost irrelevant to valuation
|
|
Total core business
|
General-purpose complete robots, components, and related products
|
¥ 1.676 billion
|
+332.35%
|
100%
|
60.13% core-business gross margin
|
Humanoid and quadruped growth engines support industry-leading earnings quality
|
Note: Core-business revenue is below total revenue of ¥ 1.699 billion, with the difference representing other-business revenue. Year-over-year rates are recalculated from figures disclosed in the preliminary prospectus; rounding may produce minor differences.
The sales mix combines standardization with industry coverage. In 2025, offline sales represented 89.28% of core-business revenue, comprising 51.10% offline direct sales and 38.18% distribution. Online direct sales and sales into e-commerce inventory totaled 10.72%. Overseas revenue accounted for 43.65%, down from 55.74% in 2024, mainly because stronger brand exposure and interest in embodied intelligence drove faster domestic demand growth. The top five customers accounted for only 12.08%, and the largest customer represented 3.11%. Customer diversity supports pricing and collections and also shows that revenue was not built on a single large factory order.
The profit pool is moving from quadrupeds toward humanoids. Quadrupeds contributed 74.91% of core-business gross profit in 2023, while humanoids contributed only 3.73%. By 2025, humanoid gross profit reached ¥ 548 million, accounting for 54.40% of core-business gross profit, while the quadruped contribution fell to 39.26%. This does not reflect a decline in quadrupeds. Instead, G1 transformed humanoids from a small number of expensive prototypes into a standardized platform capable of volume sales. The group's most valuable asset is its low-cost engineering system across the “body + cerebellum + core components.” Its most important weakness is that the “brain” has not yet been deployed at scale, while repeat purchases by industrial customers, task success rates, and actual work hours per robot still lack continuous disclosure.
Core Operating Metrics and Changes
The current catalyst is that Unitree's STAR Market IPO registration has taken effect and the online subscription is complete. The company uses security code 688836 and is issuing 40.446434 million shares, equal to 10% of post-offering shares, at an offer price of ¥ 150.80. As of the report date, there was no public trading price suitable for fundamental analysis. This report therefore discusses only the IPO valuation and does not treat subscription demand or potential first-day price action as evidence of value.
|
Metric/event
|
Latest value
|
YoY/historical comparison
|
Implications for revenue, profit, cash flow, and valuation
|
|
2025 revenue
|
¥ 1.699 billion
|
+332.64%
|
Humanoid and quadruped volumes both scaled, so growth is supported by actual deliveries
|
|
Humanoid sales/average selling price
|
5,215 units / ¥ 166,400
|
Sales +1,165.78%; average selling price -36.10%
|
Lower prices materially expanded demand; future gross profit depends more on scale-driven cost reduction
|
|
Quadruped sales/average selling price
|
23,037 units / ¥ 30,300
|
Sales +222.83%; average selling price -6.27%
|
Product maturity is higher, and the effect of volume growth still far exceeds the price decline
|
|
Gross margin/adjusted profit
|
60.44% consolidated gross margin; ¥ 591 million adjusted net profit
|
Gross margin +3.22 percentage points; adjusted profit +652.78%
|
Scale and mix created genuine profit, but post-share-based-payment expense efficiency cannot be extrapolated linearly
|
|
1Q26 / 1H26 estimate
|
1Q revenue ¥ 423 million and adjusted profit ¥ 40 million; 1H revenue ¥ 1.052-1.128 billion and adjusted profit ¥ 236-283 million
|
1Q revenue +68.49% and adjusted profit -52.55%; 1H revenue +35.62%-45.41% and adjusted profit -21.97% to -6.43%
|
R&D and selling expenses are growing faster than revenue; the earnings inflection has shifted from improvement to pressure
|
|
IPO and subscription
|
Offer price ¥ 150.80; post-offering market capitalization ¥ 60.993 billion; final allocation rate 0.01809759%
|
4.80x the ¥ 12.7 billion post-money valuation from June 2025
|
Capital resources and scarcity increase materially, but so does the burden of growing into the valuation
|
Note: The 1H26 estimate does not constitute an earnings forecast or performance commitment. The IPO valuation is calculated as ¥ 150.80 multiplied by 404,464,340 post-offering shares, consistent with the offering announcement.
The first quarter of 2026 revealed a change in the growth mix. Revenue still increased 68.49%, primarily due to humanoid demand, but R&D expense increased ¥ 38.328 million year over year, alongside higher brand-promotion and sales spending. Adjusted net margin fell from 34.77% for full-year 2025 to 9.52%. The first-half estimate implies a profit recovery in the second quarter versus the first, but reaching ¥ 800 million of full-year adjusted profit would still require simultaneous volume expansion, stable gross margin, and a lower expense ratio in the second half.
The IPO will expand the balance sheet materially. Unitree originally planned ¥ 4.202 billion of investment projects, of which intelligent-robot models, robot bodies, and new-product R&D totaled ¥ 3.578 billion, or 85.15%, with ¥ 624 million for a manufacturing base. Estimated net proceeds at the offer price are ¥ 5.917 billion, 1.41x the original plan. Prioritizing the “brain” and robot body is the right direction. What matters for valuation is whether these investments create billable industrial-task capabilities, not simply how much R&D expense increases.
3. Fundamental Quality: Cash Flow Leads Peers, but Front-Loaded Spending Is Ending Exceptional Margins
|
Item
|
2023A
|
2024A
|
2025A
|
1Q26 / 1H26E
|
Research view
|
|
Revenue
|
¥ 159 million
|
¥ 393 million
|
¥ 1.699 billion
|
¥ 423 million / ¥ 1.052-1.128 billion
|
Breakout growth in 2025; growth normalizes in 2026 but remains high
|
|
Consolidated gross margin
|
44.75%
|
57.22%
|
60.44%
|
Not disclosed
|
Full-stack in-house development and scale-driven cost reduction are working, but humanoid price cuts narrow further upside
|
|
Attributable net profit
|
¥ -11 million
|
¥ 95 million
|
¥ 278 million
|
¥ 50 million / ¥ 258-306 million
|
¥ 349 million of share-based payment depressed 2025 profit; 1H attributable profit turns positive year over year
|
|
Adjusted net profit
|
¥ -18 million
|
¥ 78 million
|
¥ 591 million
|
¥ 40 million / ¥ 236-283 million
|
Strong 2025 earnings quality; 1H26 expected to decline year over year
|
|
Operating cash flow
|
¥ 5 million
|
¥ 192 million
|
¥ 670 million
|
¥ 34 million / not disclosed
|
Strong cash collections in 2025; 1Q down 85.65% year over year
|
|
Simplified free cash flow
|
Approximately ¥ 0 million
|
¥ 183 million
|
¥ 620 million
|
Not disclosed
|
Positive and above adjusted profit in 2025, but affected by working capital and share-based payment
|
|
Capital expenditure
|
¥ 5 million
|
¥ 10 million
|
¥ 50 million
|
Not disclosed
|
Current model is asset-light; the planned manufacturing base will increase future depreciation and capital intensity
|
|
Cash-like assets/liability ratio
|
Cash ¥ 222 million / 23.57%
|
Cash ¥ 566 million / 16.19%
|
Cash, financial assets held for trading, and debt investments ¥ 2.584 billion / 18.82%
|
1Q total assets ¥ 3.459 billion; estimated IPO net proceeds ¥ 5.917 billion
|
Very strong net cash and low financing risk, with a materially higher required return on capital
|
Note: Simplified free cash flow equals net operating cash flow minus cash paid to acquire fixed assets, intangible assets, and other long-term assets. The 2023 figure was approximately ¥ -0.17 million and rounds to approximately ¥ 0 million in the table. The 1H26E figures are management estimates that have not been audited or reviewed.
Growth and margins. Revenue increased 332.64% in 2025 and consolidated gross margin rose to 60.44%, creating the appearance of hardware growth with software-like margins. A closer look shows that the gross-margin improvement came from in-house components, lower production and procurement costs, and a higher mix of industrial-grade quadrupeds. Humanoid gross margin, however, fell from 69.26% in 2024 to 63.18% as lower average selling prices and a higher G1 mix began to dilute per-unit profit. Scale can still offset lower prices, but not without limit. If sales volume grows more slowly than unit pricing and gross margin decline, revenue and profit will downshift together.
Cash flow and capital expenditure. Cash received from customers was ¥ 1.835 billion in 2025, equal to 107.98% of revenue, indicating strong collection quality. Accounts receivable was only ¥ 59 million, and the top five customers were diversified. Operating cash flow of ¥ 670 million nevertheless reflected the combined effects of ¥ 349 million of non-cash share-based payment, a ¥ 332 million increase in operating payables, and a ¥ 242 million increase in inventories; it cannot be treated directly as stable distributable cash. Operating cash flow was only ¥ 34 million in 1Q26, showing that changes in spending, inventory preparation, and supplier-payment timing can make cash conversion more volatile than it was in 2025.
Balance sheet and capital allocation. At year-end 2025, Unitree held ¥ 1.419 billion of cash, ¥ 282 million of financial assets held for trading, and ¥ 883 million of debt investments, with no short-term borrowings. Liabilities consisted mainly of accounts payable, contract liabilities, and lease liabilities. Once the estimated ¥ 5.917 billion of IPO net proceeds arrive, liquidity will no longer be a strategic constraint. The capital-allocation test therefore changes: the company must convert the approximately ¥ 1.715 billion of incremental capital above its original project requirements into stronger model, product, and manufacturing returns. Otherwise, excess cash will only reduce return on equity and will not support a high sales multiple.
Fundamental conclusion. Unitree's fundamentals are stronger than those of its main domestic robotics peers, and it has already proven that hardware can be profitable and generate cash. The single variable that will determine future earnings and valuation is whether industrial and service applications can upgrade low-priced robot bodies into a high-repeat-purchase, high-utilization task platform. If investment in foundation models and applications only increases expenses without increasing customer repeat purchases, the 34.77% adjusted net margin in 2025 will mark a cyclical peak.
4. Industry and Competitive Landscape: Unitree Wins in the Body and Cerebellum, While the Brain and Real Labor Remain Scarce
The upstream high-performance general-purpose robotics value chain includes motors, reducers, chips, batteries, sensors, and structural components. Midstream platform manufacturers integrate the robot body, motion control, and systems. Only downstream come research and education, commercial and consumer use, industrial manufacturing, inspection and rescue, logistics, and home services. The industry's largest measurement trap today is treating “shipments” as equivalent to the “number of robots performing productive work.” Research platforms, performance equipment, and factory pilots can all count as shipments without producing recurring task revenue or economic returns. This report therefore evaluates competitiveness through sales volume, revenue, gross margin, customer repeat purchases, and deployment status rather than substituting distant market-size forecasts for company earnings.
|
Competitive dimension
|
Unitree's verifiable position
|
Market-share/scale measure and limitations
|
Main competitors or alternatives
|
Implications for growth, margins, and valuation
|
|
Engineering and volume production
|
5,511 humanoid shipments and 23,037 quadruped sales in 2025; humanoid shipments disclosed as the world's highest
|
Shipments include research, education, performances, and pilots; no standardized disclosure of productive task hours or repeat-purchase rates
|
UBTECH, Leju, AgiBot, and Figure; Deep Robotics and ANYbotics in quadrupeds
|
Volume-production leadership is a genuine advantage, but only repeat purchases and task efficiency can turn unit volume into long-term cash flow
|
|
Robot body and motion control
|
Full stack across humanoids, quadrupeds, core components, and motion algorithms; rapid iteration across multiple products
|
Difficult movements demonstrate control and robustness, not autonomous success rates for complex tasks
|
Boston Dynamics, Tesla, Figure, and AgiBot
|
Unitree is the net winner in the “body + cerebellum,” supporting value for money and gross margin
|
|
Embodied foundation models
|
WMA and VLA in parallel; proprietary models are open-source and in pilots including factories
|
The prospectus states clearly that they have not been deployed in products at scale; no series for paid deployments or task success rates
|
Tesla, Figure, and domestic foundation-model/robotics companies
|
The “brain” remains a valuation option and should not be priced as a mature software platform
|
|
Cost and supply chain
|
Humanoid average selling price of ¥ 166,400 with 63.18% gross margin; core components developed and manufactured in-house
|
Average price reflects the H1/G1/R1 mix and does not represent like-for-like model pricing
|
UBTECH, Leju, Deep Robotics, automakers, and consumer-electronics entrants
|
Current cost-reduction capabilities lead; scaled manufacturing by automakers could compress the price advantage
|
|
Distribution and ecosystem
|
Overseas revenue 43.65%; top five customers 12.08%; official GitHub presence continues building the developer ecosystem
|
Open-source attention does not equal commercial conversion; customer names and depth of applications are disclosed only to a limited extent
|
UBTECH's industrial customers, DOBOT's collaborative-robot channels, and Deep Robotics' inspection channels
|
Unitree wins on breadth, while some peers win on depth in vertical applications
|
|
Capital and valuation
|
Estimated IPO net proceeds of ¥ 5.917 billion; IPO price-to-sales ratio of 35.89x
|
Listed peers remain loss-making, so P/E ratios are not comparable across companies
|
UBTECH and DOBOT average 19.74x sales; Tesla has stronger capital resources and internal use cases
|
Capital can address the brain deficit, but the 81.8% sales-multiple premium requires continued share gains and sustained high gross margins
|
Note: The “world's highest” humanoid ranking is the company's 2025 assessment based on third-party shipment data; a consistent global denominator was not disclosed. Cross-company unit sales, product sizes, revenue-recognition methods, and applications are not fully comparable.
|
Company
|
Comparable 2025 operating scale
|
Growth/gross margin/adjusted net margin
|
Robot deliveries or business status
|
Capital investment and strategy
|
Competitive conclusion
|
|
Unitree Robotics
|
Revenue ¥ 1.699 billion
|
+332.64% / 60.44% / 34.77%
|
5,511 humanoid shipments; 23,037 quadruped sales
|
Full stack across body, cerebellum, and components, with increasing investment in embodied models
|
Net winner in engineering volume production, cost, and profitability
|
|
UBTECH
|
Revenue ¥ 2.001 billion
|
+53.29% / 37.67% / -35.71%
|
1,079 full-size humanoid sales, alongside education, logistics, and consumer robots
|
More focused on industrial applications, with heavier sales and R&D investment
|
Slightly larger revenue, but materially weaker delivery efficiency and profitability
|
|
DOBOT
|
Revenue ¥ 493 million
|
+31.65% / 46.49% / -25.05%
|
¥ 20 million of humanoid and multi-legged-related revenue; collaborative robots remain the core business
|
Extending into humanoids through established collaborative-robot channels
|
Industrial channels have value, but humanoids are not yet a scaled threat
|
|
Deep Robotics
|
Revenue ¥ 337 million
|
+227.02% / 52.83% / 4.48%
|
Cumulative 4,195 quadruped and wheeled-legged units and four humanoids from 2023-2025
|
Focused on vertical applications including electric power, firefighting, and inspection
|
Local winner in industrial quadruped applications, but scale and profit remain below Unitree
|
|
Leju Robotics
|
Revenue ¥ 258 million
|
+365.20% / 40.78% / -30.07%
|
Three-year cumulative sales of 609 full-size humanoids and 301 small-to-mid-sized units
|
Pursuing research and education, data collection, and industrial production in parallel
|
Fast growth, but behind in volume production, gross margin, and cash generation
|
|
Tesla Optimus
|
No comparable external robot revenue
|
Not disclosed / not disclosed / not disclosed
|
Gen 3 announced to have begun low-volume trial production
|
Advantages in AI, automotive supply chains, capital, and deployment in its own factories
|
Not currently the volume leader, but the most likely potential rival to rewrite cost and application rules
|
Note: UBTECH, DOBOT, Deep Robotics, and Leju data follow the regular-report or filing measures cited in Unitree's preliminary prospectus. Different product scopes and revenue-recognition rules cannot be compared by simple division. Growth rates are recalculated from disclosed 2024-2025 revenue.
The winners are already clear. Unitree leads in robot-body performance, product iteration, standardized volume production, unit cost, and cash generation. Deep Robotics has established a vertical advantage in high-value quadruped inspection applications, while UBTECH has invested more deeply in onboarding industrial customers. DOBOT's and Leju's humanoid businesses have not yet reached Unitree's scale. The losers are equally clear: models dependent on high-priced customization, low volume, and persistently high selling expenses are being squeezed by Unitree's standardized platforms and developer ecosystem.
However, the profit pool could ultimately migrate from “selling robot bodies” to “models, data, task software, and maintenance services.” There is currently no auditable winner on this longer track. Unitree has the largest installed hardware base and developer entry point, making it a potential net beneficiary of the data flywheel. It may also fail to obtain sufficiently deep, high-quality data from any single industrial application because its products are dispersed across research, education, performances, and consumers. The company is currently a net industry beneficiary, but the IPO valuation requires it to win four contests simultaneously: the body, cerebellum, brain, and industrial commercialization. The margin for error is narrow.
5. Key Risks
-
Demand shifts from enthusiasm to real use. If research procurement, entertainment performances, and short-term rentals cool while manufacturing, logistics, inspection, and service customers have yet to generate repeat bulk orders, humanoid sales growth will decline ahead of the industry narrative. Fixed R&D and selling investments would then raise expense ratios, causing slower revenue to flow rapidly into adjusted profit.
-
Price competition outpaces scale-driven cost reduction. Humanoid robot average selling prices already fell 36.10% in 2025, and gross margin declined 6.08 percentage points. If like-for-like prices fall by more than another 20% in 2026 without commensurate declines in materials, chips, and outsourced processing costs, consolidated gross margin could drop below 55%, weakening the high margins on which the IPO valuation depends.
-
Investment in the “brain” fails to create a product. Unitree's proprietary general-purpose embodied foundation model has not yet been used at scale. The R&D expense ratio fell from 31.39% in 2023 to 8.53% in 2025, below the 24.70% average for peers in 2025. If IPO investment only raises spending on compute, personnel, and data collection without improving task success, generalization, and customer repeat purchases, profit will fall first while the platform premium remains unrealized.
-
Cash-flow normalization could exceed the decline in profit. Operating cash flow in 2025 benefited from share-based payment, higher payables, and contract liabilities, then fell 85.65% year over year in 1Q26. If inventory continues rising, supplier payment terms shorten, or customer prepayments decline, simplified free cash flow could fall materially below adjusted profit, further reducing the approximately 1.02% free-cash-flow yield at the IPO valuation.
-
Cross-industry competitors rewrite the cost curve. Tesla, domestic automakers, and consumer-electronics companies have greater capital resources, supply-chain capabilities, and internal factory applications. Once their humanoid robots achieve stable low-volume production and become available to external customers, Unitree could face simultaneous competition for talent, components, pricing, and customer validation, pressuring both market share and gross margin.
-
Intellectual-property, export, and supply-chain constraints. As of January 31, 2026, Unitree held 262 registered patents in China and overseas, including 20 domestic invention patents. The relatively small number of invention patents may make core technology more difficult to protect. The latest U.S. FCC policy does not currently affect certified core models, but if new models cannot obtain certification or exemptions, the 43.65% overseas revenue base would face regional market-access pressure. Price increases for memory and other chips would also raise costs.
-
Governance risk combines with a high valuation. Before the offering, Xingxing Wang directly held 23.82% of shares and, through the special voting-rights structure and employee platforms, controlled a combined 68.78% of voting rights. External shareholders have limited ability to constrain major decisions. The IPO price-to-sales ratio is 81.8% above the listed-peer average. If enthusiasm fades after listing or restricted shares are gradually released, multiple compression could dominate shareholder returns even if the company continues growing.
6. Monitoring Checklist
-
Official 1H26 results: Revenue should fall within the ¥ 1.052-1.128 billion range, and adjusted net profit should at least approach the ¥ 260 million midpoint. Results below the low end of guidance would directly reinforce the negative-leaning view.
-
Humanoid volume and price: Full-year 2026 recognized sales should exceed 8,000 units, while the decline in average selling price should remain within 20%. Insufficient volume growth or faster price declines would indicate weakening price elasticity.
-
Gross margin and expense ratios: A consolidated gross margin of at least 58%, an R&D expense ratio rising to approximately 15%, and adjusted net margin of at least 20% would show that Unitree can preserve economics while investing in the “brain.” Gross margin below 55% would weaken the view.
-
Cash conversion: Trailing-12-month operating cash flow divided by adjusted net profit should remain above 80%, with simplified free cash flow staying positive. If the ratio remains below 50% for two consecutive reporting periods, the sustainability of 2025 cash flow should be reassessed.
-
Industrial repeat purchases and task metrics: Unitree should disclose at least one deployment of 100 or more units in manufacturing, logistics, inspection, or services, along with a repeat order and task-success or productive-work-hours data. Performances and prototype orders alone would not change the valuation conclusion.
-
Model commercialization: Movement of the UnifoLM series from open source and pilots into paid deployments, together with disclosure of model-driven feature revenue, customer counts, or repeat-usage data, would weaken the “brain as an option only” view. Continued absence of quantified KPIs would reinforce the negative-leaning conclusion.
-
Competitor progress: Monitor whether UBTECH can exceed 2,000 full-size humanoid sales, Deep Robotics' quadruped revenue and net margin, and Tesla Optimus external sales and actual factory deployment. If any competitor closes the loop between cost and applications, Unitree's leadership premium will narrow.
-
Growing into the valuation: Without relying on a share-price target, continue comparing market capitalization with trailing revenue, adjusted profit, and free cash flow. If fundamental improvement lowers adjusted P/E below 80x and industrial repeat purchases are established, the negative-leaning view could move to neutral. Otherwise, the high multiple will continue magnifying execution risk.
7. Research Sources