Unitree Robotics: 5,215 Humanoids at a 63.2% Gross Margin Prove the Hardware, but CNY 488 Is 116x FY2025 Revenue and Sits Above the CNY 350 Base Value

The first fate-changing stage was quadruped productization: units rose from 3,121 in 2023 to 23,037 in 2025 as price fell and gross margin rose. Falling price with rising margin points to manufacturing learning, design-to-cost work and mix/scale economics rather than scarcity pricing, and it is Unitree's best historical proof of a real competitive advantage.

The second stage began with H1 in August 2023 and accelerated with G1 in 2024: humanoid sales went from five units in 2023 to 412 and then 5,215 in 2025, with 5,511 shipped. ASP fell from roughly CNY 593,400 on the tiny 2023 base to CNY 166,400, partly on a mix shift to lower-priced G1-type products, and gross margin from an unrepresentative 87.7% to 63.2%, still economically significant at five-thousand-unit scale. The prospectus cites third-party data ranking Unitree's 2025 humanoid shipments first globally.

The 2025 Spring Festival Gala appearance changed the brand, but it should not be confused with proof of industrial demand: the filing credits that exposure and AI/robotics policy support for part of the rise in domestic revenue from 44.26% to 56.35% of main-business revenue between 2024 and 2025, with overseas still at 43.65%. The more consequential decision was pricing humanoids low enough for developers, universities, AI laboratories and smaller commercial users to buy rather than merely observe.

The third stage, the 2025–26 scale-up into a broader platform, ran on a small organization: 516 employees at end-2025, 184 of them in R&D, generated almost CNY 1.7 billion of revenue, and Unitree disclosed dozens of open-source projects covering simulation, SDKs, motion control and embodied AI. That small internal organization plus extensive outsourced/non-core manufacturing helps explain why 2025 cash paid for fixed and other long-lived assets was only CNY 50.36 million (0.50 亿元); IPO proceeds change that at the margin by funding a manufacturing base alongside model, robot-body and new-product R&D.

The fourth stage is capital-market industrialization. After CSRC approved registration on July 1, 2026, Unitree sold 40.446 million new shares, 10% of enlarged capital, raising roughly CNY 6.10 billion against approximately CNY 4.20 billion of projects originally budgeted in the prospectus. The August 19 debut changed valuation rather than business economics: the CNY 1,100 opening implied roughly CNY 445 billion of equity value.

Financial trajectory

† FY2025 uses the final CNY 1,699.27 million figure; detailed historical mix comes from the May 25 prospectus version. The retrieved versions agree on the material 2025 operating picture.

The earnings-quality oddity matters: 2025 attributable profit was CNY 278 million against roughly CNY 591 million excluding non-recurring items, mainly because a large share-based-payment expense was treated as non-recurring. That expense is non-cash but still compensation and potential dilution, so I regard CNY 278 million as too punitive and CNY 591 million as too generous if equity compensation continues; a defensible normalized starting point lies between them.

Cash collection has beaten the income statement so far. Cumulative 2023–25 operating cash flow was about CNY 867 million against CNY 363 million of reported net profit, and 2025 OCF of CNY 670 million was 2.41x reported and 1.13x adjusted profit. Customer receipts exceeded booked revenue in every reported year, but the 2025 figure also leaned on roughly CNY 332 million of higher operating payables and a large share-compensation add-back, while inventory absorbed about CNY 242 million and operating receivables about CNY 99 million.

The balance sheet is capital-light so far rather than permanently so. 2025 capex was only 3.0% of revenue, leaving OCF less physical capex at about CNY 620 million, but the IPO manufacturing base and a larger R&D organization will raise investment needs. Normalized owner earnings should also remove the temporary working-capital benefit and include a recurring cost for equity compensation, so I use roughly CNY 400–500 million as the 2025 base rather than CNY 620 million of mechanical free cash flow, an analytical assumption grounded in the cash-flow reconciliation rather than a company metric.

Revenue structure, demand quality and moat

The product data are the core of the moat case: quadruped ASP fell about 21% from 2023 to 2025 while gross margin expanded 13 percentage points, and humanoids absorbed a larger deliberate price/mix reset yet still earned 63% gross margin in 2025. That combination is consistent with an advantage in integrated mechanical/electrical design, motion-control software and supply-chain cost rather than branding alone.

Unitree's proven moat is the ability to turn high-dynamic robot performance into comparatively inexpensive serial products; its embodied-AI moat is still unproven. Full-stack development reduces dependency on third-party robot-body components and speeds iteration while scale feeds cost-down, but the future profit pool may migrate from bodies and joints to manipulation models, task data, fleet software and integration. Unitree's open developer presence helps distribution, yet no network effect today locks customers into its intelligence stack.

Demand is diversified by customer name but opaque by economic purpose. The five largest customers were only 12.08% of 2025 revenue, the filing describes users spanning universities, institutes, technology companies and developers, and offline channels supplied 89.3% of main-business revenue (direct about 51.1%, dealers about 38.2%). That limits single-customer risk without showing how much demand recurs because a robot saves labor rather than because a laboratory, exhibition or developer needs a platform; the prospectus lists scientific research, application development, education, cultural performance and intelligent services as key humanoid applications without separately quantifying productive industrial deployment.

Supplier concentration is moderate rather than alarming: mechanical parts were about half of 2025 purchases, electronics and electrical materials roughly 22% each, and the five largest suppliers 22.5%, with complete-robot and core-module assembly kept in-house and non-core parts and processes outsourced. That structure has supported rapid volume growth without tying large amounts of capital to fabrication assets.

Management, control and future supply

Founder Wang Xingxing remains the central economic and governance variable: the offering materials give him and his controlled platforms about 31.3% of the equity but roughly 65.3% of voting power through ten-vote special Class A shares (hence the "-W" suffix). Transfers and other prescribed events can convert them into ordinary voting rights, but outside shareholders cannot realistically change control while Wang retains them.

Founder control cuts both ways: Wang's wealth is overwhelmingly tied to Unitree and the company has maintained unusual technical and product speed under his control, yet ordinary shareholders bear economic downside without proportionate governance power. Two IPO employee asset-management plans extend equity participation to 171 employees, sensible only if future grants are treated as a real cost rather than routinely stripped out of "adjusted" profit.

Lock-ups matter because the float was tiny. Of 404.464 million shares of enlarged capital, only about 30.09 million were unrestricted on day one, leaving about 374.37 million restricted; pre-IPO capital alone was 364.018 million shares. The retrieved prospectus records a 36-month lock on Wang's pre-IPO holdings, to August 19, 2029, while other ordinary pre-IPO holdings generally create a much earlier potential supply event around August 19, 2027, and IPO offline and strategic tranches carry shorter or differing locks.

† Derived as 40.446 million newly issued less roughly 30.09 million unrestricted shares. ‡ Approximation: the disclosed maximum CNY 271.5 million employee-plan subscription divided by the CNY 150.80 issue price; the exact split between the two plans should be checked against the final allocation schedule before trading the unlock. § The final tranche-level listing-announcement PDF was not machine-extractable in this session; the date reflects the disclosed sponsor co-investment lock convention and should not be treated as an exact tradable-share forecast.

What matters is narrower than an exact unlock forecast: the approximately 7% starting float is temporary, and a very large body of pre-IPO stock can become eligible for sale during or shortly after the 12-month horizon, a valuation-supply risk even if operations meet targets.

Industry, cycle, competitors and ecological niche

Humanoid robotics is still an adoption market rather than a mature replacement market. Industry estimates cited around the IPO put Unitree's 2025 humanoid shipments above 5,500 and at or near the global lead. JPMorgan estimates cited in financial media projected global humanoid shipments rising toward 60,000 in 2026 and 1.75 million by 2030, from a 2025 base that industry trackers put at roughly 18,000. The width of that range matters more than the point estimate: volume can grow tens of times while most manufacturers still fail to earn adequate returns, if hardware commoditizes or productive deployment arrives later than investors expect.

The current profit pool is unusually favorable to Unitree because customers pay for scarce robot-body capability, so an integrated platform can earn gross margins above conventional equipment manufacturing. Bargaining power may later shift to component suppliers with hard-to-replicate precision technology, foundation-model/data providers that control manipulation intelligence, and integrators that prove task-level ROI to factories. A robot that runs, flips or dances is technically hard; one that reliably replaces a worker over thousands of hours creates the economic switching cost, and that remains the industry bottleneck.

The relevant cycle is one of technology adoption and capital allocation: policy, venture funding, IPO liquidity and media attention can temporarily pull orders forward into laboratories, data-collection centers, exhibitions and pilot programs. Reuters reported on September 21 that Chinese regulators were becoming more cautious about humanoid-robot IPOs because high valuations and revenue tied to state-supported projects might outrun genuine commercialization. That shows the entire sector is moving from "can it be built?" to "who pays repeatedly for it?" Unitree's low customer concentration and large overseas share argue against reading its own revenue as simply subsidy-driven.

Policy is geographically asymmetric. Chinese AI and robotics policy has supported experimentation and procurement, which Unitree says helped 2025 domestic sales. Abroad, the U.S. FCC in July 2026 added foreign-made advanced robots to its Covered List, restricting new equipment authorizations: already authorized products can continue under existing approvals, which still face regulatory risk, but future products face a materially higher barrier, and Reuters reported that the Pentagon identified Unitree as linked to China's military industrial base. For a company whose differentiation depends partly on globally networked development, this is a structural constraint rather than a one-quarter tariff issue.

Horizontal comparison

UBTECH and Dobot are the most useful listed operating comparisons; Deep Robotics and Leju are closer in specific robot categories but were private/IPO-stage references in the prospectus. Tesla matters as a future humanoid scale benchmark but currently has a completely different earnings base, making its corporate valuation useless as a direct comparable.

The prospectus peer table also showed 2025 revenue of about CNY 337.5 million and 52.8% gross margin for Deep Robotics, and CNY 258.2 million and 40.8% for Leju. I leave same-date Hong Kong peer multiples blank rather than splice in a non-primary or stale quote, since robotics share prices have moved rapidly.

UBTECH became the enterprise-oriented humanoid integrator (customers are buying a broader solution, brand and industrial/service deployment program), with 2025 revenue slightly above Unitree's on a much weaker gross-margin profile; Unitree became the price-performance robot-body platform, where a developer or university can buy capable, relatively standardized hardware instead of funding a custom project; Dobot came from collaborative arms and industrial automation, with a more conventional installed base and less direct exposure to the legged-robot narrative. Those histories explain why margins and investor expectations should differ rather than converge mechanically.

Unitree's 60% 2025 gross margin deserves a quality premium to conventional robot assemblers if it persists as volume grows, but a 116x sales multiple needs a completely different standard of proof: even a company maintaining 50% gross margin can destroy equity value at that multiple if R&D, sales expense and price competition absorb the gross profit. Relative valuation cannot rescue the current price, because peers are themselves valued on an early-stage humanoid narrative and most are less profitable than Unitree.

Ecologically, Unitree is a leader in accessible high-performance robot bodies and a challenger in the far larger future market for productive robotic labor. It currently takes profit from bespoke robotics engineering by standardizing capability into products, and component commoditization, a competitor's manipulation-model breakthrough or integrators owning the customer could take that profit back. A broad price war favors Unitree over small startups given its cost-down record; a value shift from locomotion to task intelligence is more ambiguous.

Current fundamentals and what the market is trading

Reading 2026 cleanly means separating the headline profit recovery from the underlying trajectory. H1 revenue rose 48.54% to CNY 1.152 billion and reported attributable profit reached CNY 274 million against a CNY 32 million loss in H1 2025, yet adjusted profit fell 19.34% to approximately CNY 244 million. Large share-compensation effects in the base period explain the divergence; underneath, incremental R&D and selling investment is currently outrunning incremental operating profit.

The Q1 filing already showed it: revenue of CNY 422.84 million rose 68.49%, but reported profit fell to CNY 50.01 million and adjusted profit to about CNY 40.3 million, down 52.6% year on year. H1 improved sequentially, yet the 2025 pattern of explosive revenue and profit growth together has broken, which is normal for a business investing heavily in a new category and uncomfortable for an equity valued at several hundred times earnings.

There is little evidence yet of a balance-sheet demand problem: end-2025 receivables were only 3.6% of annual revenue, inventory rose but revenue grew even faster, and the IPO added approximately CNY 6.1 billion of gross proceeds. Unitree's near-term case is constrained by valuation and commercialization, not solvency.

Price-path attribution

The first trading day was almost pure scarcity and thematic price discovery: retail demand reportedly exceeded available supply by thousands of times, only about 7.44% of shares were freely tradable, and "embodied AI" was among the hottest narratives in the Chinese market. Unitree opened at CNY 1,100, 629% above the issue price, and closed at CNY 845, up 460%. The CNY 341.8 billion closing value was a multiple event created by scarce float and the market's readiness to capitalize an enormous future category at once; a one-day change in robot cash flows could not plausibly have produced it.

The retreat was also mainly a multiple event, though fundamentals gave the market reasons to question the narrative. Within weeks the shares had lost roughly 55% from the CNY 1,100 peak (CNY 477.12 on September 11, CNY 474.28 on September 16, CNY 514.98 on September 18, CNY 488 on September 24) as investors absorbed the H1 adjusted-profit decline, slower revenue growth versus 2025, new U.S. barriers, sector-wide debate over real versus policy-created humanoid demand and the realization that later lock-up expiries will eventually remove the initial scarcity. The business itself did not shrink by half.

At CNY 488 the stock is still 223.6% above the issue price, 42.2% below the debut close and 55.6% below the first-day peak: the collapse from CNY 1,100 moved the valuation from an extraordinary scarcity premium to a still-extraordinary growth premium without making the stock conventionally cheap.

STAR Market liquidity and margin financing can magnify such moves, but I did not retrieve a sufficiently reliable daily 688836 margin-financing balance to assign a numerical portion of the sell-off to leverage, so I treat financing flows as a volatility amplifier rather than a proven fundamental driver; the demonstrable drivers are the initial 7.44% float, extreme subscription demand, earnings deceleration and subsequent sector de-rating.

Bulls read the H1 expense surge as Unitree deliberately trading near-term adjusted profit for a larger platform while its hardware lead holds; bears read 2025 margins as unusually favorable early-adopter pricing that competitive R&D and selling costs are beginning to normalize before industrial volumes arrive. For valuation I lean toward the bear reading until productive deployments become separately measurable. The implication is that a CNY 197 billion equity value raises the burden of proof, not that the technology is weak.

Valuation analysis

Historical valuation and capital-market framing

With only 28 trading sessions by the base date, "historical valuation range" means IPO-to-present price discovery rather than a five-year multiple study. The CNY 150.80 issue price implied about CNY 61 billion of market value and an issuer-stated 219.23x 2025 issue P/E on the prescribed lower earnings measure, far above the 38.56x average static P/E of the reference C34 general-equipment group. The CNY 845 debut close implied about CNY 342 billion; the September 24 close of CNY 488 leaves CNY 197.38 billion.

Unitree can thus lose more than half from its first-day high and still trade above three times the IPO market capitalization. On FY2025 numbers that is approximately 709x reported profit, 334x adjusted profit and 116x sales, and a market-data service's approximately 338x TTM P/E on September 24, close to the adjusted figure, confirms that earnings growth since 2025 has not brought the stock into a remotely conventional range.

The reported last private valuation of roughly CNY 12.7 billion is a useful if imperfect anchor: the IPO value was almost five times that mark and the September 24 value about 15.5 times. Part of the gap is justified by 2025's extraordinary revenue growth, much better profitability, IPO liquidity and a large new cash balance; most of it is public-market option value on humanoid robotics, capitalizing an industry structure that does not exist yet.

Peer valuation

Peer earnings multiples are of little use as a primary instrument because most direct humanoid competitors are loss-making, privately held or mix humanoids with other robotics businesses. Unitree's 60.1% FY2025 gross margin stood above UBTECH, Dobot, Deep Robotics and Leju, which supports a premium but not, by itself, a 116x-sales valuation.

The question is how long Unitree can preserve that spread as ASP declines. Its quadruped history says cost reduction can outrun price cuts, but its humanoid history has only two commercial-scale years across very different products, so extrapolating the same learning curve for a decade is premature. The cleanest valuation is therefore absolute: how many robots Unitree must sell and what they must earn.

Cash-flow passthrough

Financial history covers three prospectus years rather than five. The 2.39x cumulative 2023–25 OCF/Net Income ratio and 2025 OCF of CNY 670 million against CNY 278 million of reported profit overstate steady-state owner earnings, because 2025 OCF benefited from payables and the non-cash share-compensation add-back.

I estimate CNY 15–25 million of 2025 capex as maintenance, though the split is not disclosed and future manufacturing investment will be far higher. On the CNY 400–500 million normalized owner-earnings base, CNY 197.38 billion is roughly 395–493x owner earnings: well below the 709x reported P/E, yet cash accounting cannot make today's price inexpensive.

Scenario valuation

The model asks what Unitree could earn in 2030 and discounts that back 4.25 years to September 2026, explicitly modeling humanoid and quadruped units, ASP, other/components revenue, normalized net margin and exit P/E, a deliberately higher bar than a multiple on next year's sell-side EPS because today's price is an option on several years of commercialization.

The 2025 starting point for the model is 5,215 humanoid sales at roughly CNY 166,400 ASP and 23,037 quadrupeds at CNY 30,300 ASP. Even the conservative scenario assumes 100,000 humanoids in 2030, more than nineteen times 2025 sales, so the scenarios differ less on whether robots succeed than on how large and profitable Unitree becomes if they do.

The conservative CNY 154 value assumes revenue near CNY 16.1 billion, almost ten times 2025, with ASP compression and a 15% net margin holding earnings to CNY 2.4 billion. The base case needs CNY 23.1 billion of revenue and CNY 4.6 billion of normalized earnings, the optimistic case CNY 39.5 billion and CNY 9.1 billion, several times the size of today's listed Chinese robot companies. Those outcomes are aggressive before any debate about exit multiples.

This is valuation-scenario analysis within a research framework, not investment advice.

Expectation gap

Reverse-engineering CNY 488 is revealing. At a 10% required return through 2030, today's CNY 197.4 billion becomes roughly CNY 296 billion of required terminal equity value. A 45x exit P/E then requires about CNY 6.6 billion of normalized 2030 earnings, or approximately CNY 33 billion of revenue at a 20% net margin, an approximately 81% five-year CAGR from FY2025; even 50x still requires close to CNY 30 billion of revenue. That is the growth embedded in the current price before any margin of safety.

The market may be underestimating Unitree's ability to cut hardware cost: the quadruped record is genuinely impressive. It is more likely overestimating how fast technical capability becomes productive, repeatable labor demand. The next decisive data are humanoid unit growth with disclosed factory/service deployments, repeat orders, stable gross margin and operating profit that begins growing again after the current R&D/sales investment phase, rather than another running-speed record.

Margin-of-safety recheck

CNY 488 is 218% above the CNY 154 conservative value, so the margin of safety against that scenario is zero. The most fragile base assumption is humanoid volume: cutting 2030 base-case units from 150,000 to 105,000, 70% of the assumption, lowers present value from about CNY 350 to about CNY 290 before any multiple cut for weaker adoption. With normalized earnings flat for three years and today's extreme multiple somehow unchanged, the annualized return before dividends would be approximately zero, and normal multiple compression turns it sharply negative: at 40x on CNY 500 million of flat owner earnings, for example, the resulting price is roughly CNY 49, about -53% annualized for three years. Even the zero-growth/unchanged-multiple case underperforms a positive 10-year government bond yield: there is no margin of safety at this buy price.

Margin-of-safety sufficiency verdict: none. This is presently a good underlying company at a price that requires a large amount of future success to become ordinary.

Risks, catalysts and tracking indicators

The highest-probability permanent-loss risk is demand quality. Probability: high; impact: high. The prospectus proves high shipment growth without disclosing the share of humanoid units whose end users earn measurable labor-cost savings, and research, education, AI development, performances and exhibitions are legitimate markets but do not support the same TAM or replacement cycle as factory labor. If unit growth slows before productive deployments broaden, revenue estimates and the multiple fall together. The observable indicator is repeat industrial/service orders, not aggregate shipment announcements.

The second risk is commoditization of robot bodies. Probability: medium-high; impact: high. Unitree's own price cuts bring it forward, and human-form robots are drawing Chinese startups, industrial-automation companies and globally capitalized technology firms. Humanoid ASP already fell from CNY 260,400 in 2024 to CNY 166,400 in 2025 as the mix moved toward affordable products, though gross margin has so far stayed above 60%. Permanent damage begins when price falls faster than BOM/manufacturing cost and consolidated gross margin moves sustainably below roughly 50%.

The third is geopolitics. Probability: high for continued restriction; impact: medium-high. With 43.65% of 2025 main-business revenue overseas and 13.3% from the U.S., the FCC Covered List action raises barriers to future foreign-made networked robots and the Pentagon designation raises the possibility of broader U.S. procurement, investment or technology restrictions, with second-order risk to parts, development ecosystems and compliance-conscious multinational customers. Unitree can redirect sales toward China, Europe and other markets, but that makes the valuation more dependent on domestic commercialization.

The fourth risk is valuation plus future share supply. Probability: high that more shares become tradable; price impact: medium to high. The stock began with only about 30 million unrestricted shares against more than 404 million outstanding, helping create extreme scarcity pricing. As offline, strategic and pre-IPO restrictions expire, more tradable supply can remove the scarcity multiple without any earnings miss, and the large August 2027 pre-IPO unlock window sits inside the requested 12-month horizon.

The fifth risk is earnings-quality normalization. Probability: medium; impact: medium. Share-based compensation opened an unusually large gap between reported and adjusted profit, and treating every future grant as non-recurring would flatter economic earnings, particularly for a 516-person technology business competing for scarce robotics and AI engineers. Cash flow is currently strong enough that this is a valuation-denominator concern rather than an accounting-solvency one.

Positive catalysts are unusually easy to observe: Q3 2026 revenue growth holding above roughly 40% with the adjusted-profit margin no longer deteriorating, which would indicate H1 spending is beginning to generate operating leverage; disclosure of repeat industrial or service deployments large enough to separate productive from research/display demand; and a 2027 volume ramp in lower-priced humanoids with consolidated gross margin above 50%, validating another quadruped-like cost curve. A material U.S. exemption or authorization route for future models would be a fourth, lower-probability catalyst.

Negative catalysts are the inverse: Q3 growth below about 30–35%; adjusted profit still declining despite rapid revenue growth; humanoid price cuts pushing group gross margin below 50%; another U.S. measure hitting already authorized models; or the market discounting the August 2027 unlock months before the shares become tradable. Sector-wide IPO tightening is already an early warning that tolerance for "embodied AI at any valuation" has weakened.

† The SSE runs the periodic-report reservation system, but I could not independently retrieve a company-specific date for 688836 at the research cut, so the dashboard uses the enforceable end-October Q3 deadline rather than inventing one.

Revenue growth, gross margin and industrial demand quality deserve the most weight. Unit growth without margin is commoditization; margin without unit growth means the TAM narrative is too large; both without productive deployment could still be a temporary research-and-demonstration boom. The thesis becomes materially stronger only when all three begin moving together.

Cross-synthesis conclusion, research uncertainties and sources

Vertically, Unitree has genuinely proven one capability: it can take difficult legged-robot technology, engineer it into standardized hardware, cut price rapidly and still earn high gross margin. Quadrupeds are the cleanest evidence because the result survived several product generations; humanoids show the beginnings of the same pattern at much greater speed. More than luck was at work: founder-led engineering, integrated hardware/control technology, Chinese supply-chain economics and an aggressive price-performance strategy all contributed.

Horizontally, Unitree is a better robot-body business than its spectacle-driven image suggests, with a 2025 gross margin materially above the disclosed margins of UBTECH, Dobot and Leju and somewhat above Deep Robotics. The unresolved weakness sits one level above the hardware: Unitree has yet to show that its embodied intelligence and customer ecosystem create durable switching costs, or that humanoids can deliver reliable labor ROI at scale, and the next battle is manipulation, autonomy, uptime and integration rather than running speed.

The share price pre-spends much of that success: CNY 488 needs roughly 81% annualized revenue growth from 2025 for five years, and Unitree can miss no major leg of the commercialization sequence for that to work comfortably.

Over one year the decisive variables are Q3/Q4 growth, adjusted-margin stabilization, U.S. restrictions and anticipation of 2027 share supply; over three years, unit economics and repeat industrial deployments. Over five years the question is where the industry's profit pool settles: Unitree is worth far more if its integrated robot-body platform keeps pricing power and captures models/data, and far less if bodies become standardized endpoints for intelligence owned elsewhere.

Bull and bear reasons

Core bull reasons:

• Unitree sold 5,215 humanoids and 23,037 quadrupeds in 2025 while retaining 63.2% and 56.7% respective product gross margins, evidence that scale has not yet destroyed unit economics.
• Quadruped ASP fell from CNY 38,300 in 2023 to CNY 30,300 in 2025 while gross margin rose from 43.7% to 56.7%, unusually clear evidence of cost-down capability.
• FY2025 revenue rose above CNY 1.69 billion and H1 2026 still grew 48.54%, so Unitree enters public markets with substantial commercial momentum rather than pre-revenue optionality.
• Customer and supplier concentration are moderate, cash collection has historically exceeded revenue, and the IPO added roughly CNY 6.1 billion, giving Unitree room to fund the next product cycle internally.

Core bear reasons:

• At CNY 488 the stock trades at roughly 116x FY2025 sales and more than 300x adjusted earnings, so even excellent operating execution can coexist with poor shareholder returns.
• H1 2026 adjusted profit fell 19.34% despite 48.54% revenue growth: R&D, selling and competition are already consuming incremental economics.
• Filings do not disclose how much humanoid revenue is productive industrial labor; with research, education, development and display still important, the long-run replacement TAM is unproven.
• The U.S. was 13.3% of 2025 main-business revenue before new FCC barriers to future foreign-made connected robots, a direct structural limit on an important overseas market.
• The tiny initial float will expand as lock-ups expire, and the large pre-IPO supply window around August 2027 can compress the scarcity premium even if operating results hold up.

Pre-mortem: where this could go wrong

Script one: by 2028 Chinese humanoid competitors match Unitree's mobility and cut entry hardware prices 30–40%. Unitree follows, but manipulation software stays undifferentiated and industrial customers refuse large deployments. Group gross margin falls below 45%, normalized profit reaches only CNY 1.5–2.0 billion, and the market rerates the company from several hundred times earnings toward 40–50x; the share price can fall well over 50%.

Script two: the large 2027 pre-IPO unlock arrives as revenue growth slips below 30% and U.S. restrictions eliminate most new-model sales there, while research/display demand stops compensating for missing factory orders. Even with positive profit and cash, the stock moves toward the conservative valuation range as scarcity disappears, a 60–70% capital loss from CNY 488.

Final research conclusion

Unitree is a rare case where I like the operating evidence materially more than the stock price. It has proven hardware commercialization, cost-down and high gross margins, and it may become one of a handful of globally important general-purpose robot platforms, but the valuation asks shareholders to pay today for industrial demand, embodied intelligence and scale economics the company still needs several years to demonstrate.

CNY 488 sits above my CNY 350 base fair value yet well below the CNY 759 value of a very successful 2030 outcome: real upside if humanoids become mass productive assets, but no conservative margin of safety. A materially lower price, or evidence that productive deployments are accelerating faster than my base assumptions, would change the balance.

【Investment rating】

• Rating: Watch
• One-line thesis: Unitree has proven hardware commercialization, but CNY 488 already prices a multi-year shift from research robots to scaled productive deployment.
• Ideal buy price: see dedicated line below.
• Acceptable hold price: CNY 300–400, approximately ±15% around the CNY 350 base value.
• Clearly overvalued price: CNY 840–950, beginning more than 10% above the CNY 759 optimistic value.
• Current-price classification: outside the three bands; above acceptable hold value but below the clearly-overvalued band.
• Whether to wait for a better price: yes; the preferred entry requires CNY 105–120 with operating fundamentals intact, or a materially higher conservative valuation after evidence of productive industrial demand.
• Target holding horizon and expected annualized return: 3–5 years; conservative -14.7%, base +1.8%, optimistic +22.0% annualized to the modeled 2030 terminal values.
• Max-loss risk: a demand-quality disappointment combined with hardware price competition and multiple normalization could plausibly destroy 60–80% of capital from CNY 488.
• Reassessment trigger: revenue growth below 30% YoY.
• Reassessment trigger: adjusted net margin below 15%.
• Reassessment trigger: consolidated gross margin below 50%.
• Reassessment trigger: inventory/revenue above 35% or inventory growing faster than revenue for two periods.
• Reassessment trigger: U.S. revenue collapses without replacement elsewhere, or restrictions expand to already authorized models.

【Ideal Buy Price】105–120 CNY Basis: 22–32% below the CNY 154 conservative scenario value, satisfying a minimum 20% margin-of-safety requirement while assuming the underlying commercialization thesis remains intact.

【Valuation Range】

• current: 488.00 (close as of 2026-09-24)
• bear (conservative · ideal buy zone): [105, 120]
• base (fair · acceptable hold zone): [300, 400]
• bull (optimistic · above the clearly-overvalued line): [840, 950]

Research uncertainties are material. First, the most detailed machine-readable primary document retrieved was the May 25, 2026 prospectus version; I reconciled its FY2025 totals with later offering/listing facts rather than substituting earlier draft figures, but a few final presentation details may differ.

Second, the aggregate restricted-share reconciliation matches disclosed capital and the initial float exactly, but I could not machine-extract the final listing announcement's complete sub-tranche allocation among offline, sponsor and individual strategic holders, so I did not fabricate those share counts; traders of specific unlock dates should use the final SSE allocation schedule.

Third, no sufficiently verifiable company-specific Q3 2026 appointment date was retrievable from the SSE reservation interface at the research cut, hence the end-October deadline in the dashboard.

Fourth, the largest fundamental blind spot comes from the company itself: neither the prospectus nor the operating data retrieved split humanoid revenue among productive industrial labor, R&D platforms, education, entertainment/display and other uses, which prevents a confident estimate of commercially proven TAM.

Fifth, Hong Kong robotics peer market capitalizations could not be reconciled to the same September 24 primary-source cut, so the peer section uses their disclosed 2025 operating economics and absolute valuation carries the conclusion.

Key primary sources are the Shanghai Stock Exchange IPO/prospectus materials and listing-announcement system, the CSRC registration approval, and Unitree's prospectus financial statements and operating disclosures; contemporary Reuters and AP reporting was used mainly for the IPO price path, U.S. regulatory developments and sector capital-market context.

Other tickers mentioned

• 9880.HK: UBTECH is the principal listed humanoid-robot operating comparison, with greater 2025 revenue but materially lower gross margin.
• 2432.HK: Dobot provides a listed industrial and collaborative-robot reference with a more established automation orientation.
• TSLA.US: Tesla's Optimus program is a potential long-run scale competitor, though Tesla's current corporate valuation is not a valid direct humanoid comparable.

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