Shenzhen Dobot: Real Cobot Business, Unproven Humanoid Bet

This is why Dobot is best described not as “high-quality growth” and not as a “valuation bubble” either, at least not at today’s much lower price. The better label is a company in transition. It has already proved one thing that many robotics stories never prove: it can design, manufacture, ship, and globally distribute real products at meaningful scale in an economically relevant category. It has not yet proved the harder thing the current narrative now demands: that humanoid and quadruped products can become a durable, repeatable, and margin-resilient business instead of a small, fast-growing but still experimental adjacency.

At the current price, the stock no longer looks like a pure mania proxy. It also does not yet offer a clean margin of safety for a balanced investor. Using the 2025 year-end share count and the 24 July 2026 close, Dobot trades at about 19.5x trailing 2025 sales in HKD terms. Even if one gives credit for the 2026 revenue acceleration implied by management’s unaudited H1 range, the market is still capitalizing future success rather than current profitability. The share price is no longer pricing perfection. It is still pricing a lot of execution.

Vertical history and financial review

Dobot came into existence at the right moment and for a specific reason. The company was founded in Shenzhen in July 2015 by Liu Peichao, Lang Xulin, Wu Zhiwen, and three other shareholders. Liu’s training was in mechanical engineering at Shandong University, and the founding team’s original problem was to make robots small enough, safe enough, and cheap enough to pull automation out of the cage and into lighter industrial, educational, and prosumer settings, not to build a humanoid champion. The first public proof of demand was a successful Kickstarter campaign for the first-generation desktop-level cobot in 2015. This origin explains a lot about the business even now: Dobot’s DNA is productizing approachable robotics, not only serving blue-chip auto plants.

The company’s history falls naturally into four stages. The first stage was desktop proof-of-concept. In 2015 and 2016 Dobot launched its crowdfunding-backed first-generation product, then the Magician and the M1 SCARA product lines. The business model at that point was hardware-led and channel-led, aimed at lighter-use applications where price and usability mattered as much as payload. The second stage was industrialization. Between 2020 and 2023 Dobot launched the MG400, began manufacturing the six-axis CR series, added the Nova series for commercial settings, expanded the production base in Rizhao, and increasingly framed itself as a Chinese leader in cobots rather than an educational robot maker. The third stage was capital-market scaling. The company converted to a joint-stock company in late 2022, came to Hong Kong in December 2024, and then used a high-flying stock in 2025 to raise much more follow-on capital. The fourth stage, now underway, is the embodied-intelligence transition: the attempt to use the cobot base, customer access, and fresh equity to push into humanoids, quadrupeds, and data-collection systems before the field consolidates around a handful of better-capitalized players.

The listing path matters because it changed the company’s resources and its market identity. Dobot’s Hong Kong IPO issued 40.0 million shares at HK$18.80 on 23 December 2024, raising HK$752 million gross. The over-allotment option later added 4.20 million shares at the same price. It was only the beginning. In July 2025 Dobot sold 19.1 million new H shares at HK$54.30, and in November 2025 it sold another 16.66 million at HK$46.80. Net proceeds from those two placings amounted to about HK$1.79 billion. Management explicitly linked the placings to funding high-speed collaborative robots and humanoid and multi-legged embodied-AI robots. In plain English, the company used a thematic valuations window exactly as a rational founder-led growth company should. Existing shareholders got diluted. The balance sheet became much stronger. Both things are true.

The most consequential 2025 corporate event after the placings was smaller in size but larger in strategic meaning: the acquisition of Hangzhou INFFNI Robotics in July 2025. Dobot bought 100% of the company for total consideration of RMB24.24 million including contingent consideration tied to R&D targets in multi-legged bionic robots, and the acquired company brought patents, technology, and goodwill onto the balance sheet. This was a signal that Dobot was willing to buy capability where it sped up the quadruped roadmap, not a financially transformative deal. It fits the broader pattern of 2025: internal humanoid development, external option-building, and capital raised while the market was willing to pay.

Financially, the vertical story is one of better scale but still incomplete self-funding. Revenue rose from RMB373.7 million in 2024 to RMB492.2 million in 2025. The application-setting breakdown is more revealing than the headline. Industrial revenue grew 39.4% to RMB278.8 million and became 56.9% of product revenue. Education revenue still contributed RMB166.6 million, or 34.0%. Commercial revenue was smaller at RMB44.9 million but grew 75.7%. Geography was almost evenly split: RMB249.4 million from mainland China and RMB242.8 million from overseas markets including Hong Kong, Macau, and Taiwan. No single customer accounted for 10% or more of revenue, which reduces one of the common failure modes in young hardware companies.

Gross profit margin tells a subtler story. The group gross margin slipped modestly to 46.1% in 2025 from 46.6% in 2024. Excluding inventory write-downs, gross margin was 48.4%, down from 49.5%. Management attributed the decline mainly to a higher share of domestic revenue with lower gross profit margin. This is plausible and consistent with a company pushing harder into mainland industrial rollout. The more important detail is in the A-share prospectus: main-business gross margin stayed relatively stable over 2023-2025, but product-line margins diverged. In 2025 six-axis cobots carried a 45.4% margin, four-axis cobots 52.7%, composite robots 42.4%, and embodied intelligence 43.4%, down from 49.9% in 2024. The clearest available sign here is that the new segment is not yet a margin miracle. It is commercial enough to sell. It is not yet commercial enough to outrun the margin profile of the core cobot lines.

The expense structure remained the main reason the company stayed in the red. Selling and distribution expenses rose to RMB182.3 million, up 32.1%, as Dobot expanded sales and technical support teams and spent more on exhibitions and online marketing. R&D expense rose 59.7% to RMB114.7 million as headcount and materials spending increased for embodied AI and core technologies. Administrative expenses fell on paper because 2024 had listing expenses, but excluding that effect they also rose as the organization scaled. Employee wage and salary expense excluding directors climbed to RMB187.5 million, while share-based payment expense reached RMB34.4 million. These are the price of trying to scale distribution and frontier R&D at the same time, not cosmetic expenses.

Cash flow improved, but not enough to call the business cash generative. Operating cash outflow narrowed to RMB42.6 million from RMB91.7 million in 2024. Inventory increased by RMB74.6 million and receivables increased by RMB31.9 million, absorbing cash even as sales grew. This is common in hardware scale-up phases, but it does mean that accounting progress has not yet become owner earnings. Dobot also does not meaningfully capitalize development costs; the annual report says no material R&D costs were capitalized during the reporting period. For valuation purposes, that makes reported losses more honest than in some software or biomedical stories. It also means there is no hidden reservoir of near-term earnings waiting to surface through lower expensing.

The balance sheet, by contrast, is a real strength. Borrowings dropped to RMB71.8 million by year-end 2025 from RMB217.8 million a year earlier. Net current assets reached RMB2.57 billion, and total equity rose to RMB2.62 billion. Interest income of RMB50.2 million in 2025 makes sense only because the company was carrying a large cash and deposit balance after the IPO and placings. For a loss-making robotics company, that level of liquidity changes the risk profile. The permanent-loss risk is now much more about business-model disappointment and valuation compression than about a sudden funding crisis.

The price history reflects those stages very cleanly. Dobot listed at HK$18.80 in December 2024. By March 2025 it had become a humanoid proxy and hit HK$83.8 after the Atom launch. In that phase the market treated Dobot as a scarce asset: a listed robotics name with a credible industrial base and a tangible humanoid product. By the second half of 2025 the company was raising fresh equity twice at elevated prices, which helped fund strategy but also increased the share count from 400.0 million to 439.96 million by year-end. By July 2026 the shares had fallen back to HK$25.22. This left the stock roughly one-third above IPO but around 70% below the hype peak. The valuation center shifted first because the market switched categories, then because the business had not yet produced numbers big enough to hold the new category premium in place.

Business model, moat, industry, and competitors

Dobot’s revenue machine is narrower than the “embodied intelligence platform” label suggests. The real business still sits in collaborative robots and related integrated solutions. The A-share prospectus shows that, within main business in 2025, six-axis cobots accounted for 61.7% of revenue, four-axis cobots 19.0%, composite robots 13.8%, and embodied intelligence only 4.1%, with the balance in other items. Gross profit followed the same pattern: collaborative robots and composite robots together contributed more than 94% of main-business gross profit; embodied intelligence contributed only 3.8%. The market can choose to value the option. The profit pool is still old-fashioned enough to live in the established product lines.

The core economic engine is six-axis cobots. They grew quickly in 2025, with sales revenue up 44.7% and unit volume up 78.7%, even as average selling price per unit fell 19.1%. Unit cost also fell 19.7%, which kept gross margin essentially stable at 45.4%. This is what a real manufacturing learning curve looks like: more units, lower price, lower cost, steady margin. Four-axis cobots are now the slower and more mature product line; sales revenue fell 3.0% in 2025 while unit volume fell 7.0%, though price and margin remained healthy. Composite robots carry a lower margin and appear more project- and customer-mix-sensitive. Embodied intelligence is still too small for clean multi-year unit-economics confidence, but the disclosed 2025 margin decline from 49.9% to 43.4% says the early business is not yet enjoying scale benefits.

The cost structure has exactly the profile one would expect from a hardware company that is trying to become a platform company. Direct materials are still the largest cost item, including metal machining parts, harmonic reducers, and electrical components. This means Dobot cannot wish away supply-chain economics. At the same time, selling and R&D are climbing fast, which means operating leverage will not show up cleanly until either core cobot volumes rise much faster or embodied products begin to contribute meaningful gross profit. At present the business does have gross-margin resilience, but not enough scale yet to offset the fixed-cost load of global channel building and frontier R&D.

The moat is real in some places and promotional in others. The strongest real moat is product-market fit in accessible collaborative robotics. Dobot has spent a decade building a broad product matrix, spanning industrial, educational, and commercial use cases, and by its own account cumulative global shipments exceeded 100,000 units in 2025 while serving over 80 Fortune Global 500 companies. A young robotics company can exaggerate many things; it cannot fake ten years of installed-base learning, distribution habits, and application engineering. Dobot is not unassailable, but it is more credible than a pure demo-driven startup.

A second real moat is manufacturing and industrial-customer adjacency. Dobot’s 2026 reply materials argue that it can sell embodied-intelligence products into a pre-existing B-end customer base and that it already has nearly 100 industrial cooperation customers in that business line. This claim should be treated carefully because it comes from issuer materials. Even so, it aligns with a point the market often misses: humanoid and quadruped commercialization will probably go to whichever company gets factory managers, systems integrators, and distributors to tolerate risk one deployment at a time, not just the one with the most dramatic product video. Dobot’s cobot base is useful precisely because it lowers that tolerance threshold.

The weakest supposed moat is frontier-technology leadership. Dobot does have a credible R&D program. The prospectus lists ongoing work on the Atom project, robot software, and larger-payload CR products, and the company says the Atom project has already reached small-batch production and continuous iteration. But frontier robotics is one of the least forgiving domains in which to claim durable advantage before scale is proven. UBTECH, Unitree, and a long list of private Chinese peers are spending aggressively; global incumbents such as Fanuc and Universal Robots already own trusted positions in adjacent industrial automation. At this stage Dobot’s technology moat is promising, not yet proven through a full competitive cycle.

Management quality looks better than the stock’s volatility might suggest. Founder-chairman Liu Peichao has stayed in place since 2015, and CFO Wang Yong joined in 2022 with a finance and governance background. The most important capital-allocation evidence is not a dividend or a buyback but the willingness to sell stock when the market gave the company an unusually high valuation. This decision diluted existing holders, but it also sharply reduced financing risk and funded a costly strategic transition. On alignment, the founder-led structure still matters positively. On governance discount, the bigger watchpoint is not control abuse but the normal risk of issuer optimism in a hot sector. The annual report was audited by Ernst & Young, and I did not find evidence in the reviewed materials of auditor churn, major accounting disputes, or large legal overhangs.

The industry structure helps explain why Dobot can matter without being dominant. IFR says cobots were only 10.5% of industrial robot installations worldwide in 2023, which means the category is still in a penetration phase rather than a mature end state. Global industrial robot installations still totaled 542,000 in 2024, and Asia captured 74% of that market. China’s manufacturing base, automation push, and policy bias toward robotics give local players a large home field. But the profit pools are still split unevenly. Global incumbents capture trust, support networks, and installed-base recurring revenue; Chinese challengers capture faster growth, lower costs, and local manufacturing adjacency. The market gap Dobot filled was the gap between premium foreign cobots and simpler, lower-function domestic alternatives. The market gap it is now trying to fill is the one between pure humanoid vision stories and proven industrial deployment discipline.

The peer picture is clearer in prose than in a sprawling matrix. Universal Robots, inside Teradyne, became the premium reference brand in collaborative robots by making the category easy to buy, easy to deploy, and easy to integrate. Customers pick it for software ecosystem, integrator familiarity, and trust, not for nationalist technology symbolism. Fanuc became the opposite kind of comparator: a sprawling industrial-automation incumbent whose cobots are credible because everything else around them is credible. Customers pick Fanuc when downtime is intolerable and support depth matters more than novelty. UBTECH became the listed humanoid benchmark in Hong Kong, with vastly larger humanoid activity, larger losses, and much more direct exposure to the entire “China embodied AI” narrative. Customers and investors pick UBTECH for humanoid ambition first, while Dobot is still bought primarily for collaborative automation with optionality on embodied AI. Topstar sits even closer to the traditional factory-equipment cycle, where valuation is disciplined by project mix, manufacturing demand, and turnaround progress instead of frontier AI excitement.

The table above is a map of what each company became, not a valuation ranking. Dobot’s ecology is distinct: it is neither the safest incumbent nor the boldest humanoid pure-play. This niche can be attractive because it offers upside if embodied products scale without requiring the entire thesis to rest on humanoids. It can also be awkward because public markets often reward clearer identities. When the market wants “industrial quality,” Fanuc and UR set the benchmark. When it wants “humanoid national champion,” UBTECH and Unitree attract more direct attention. Dobot has to keep proving that the middle ground is not a muddle but a bridge.

Current fundamentals and valuation

The freshest data point in the file is management’s voluntary release of unaudited operating figures for the six months ended 30 June 2026, not an audited interim report. Dobot said it expected H1 operating revenue of RMB300 million to RMB330 million, up 94.65% to 114.12% year on year, and gross profit of RMB140 million to RMB170 million, up 84.73% to 124.31%. Just as important, it also said the first-half net loss attributable to the parent would widen to RMB90 million to RMB120 million, with the increased loss mainly due to provisional exchange losses and share-based payments, plus heavier investment in embodied-intelligence R&D and regional expansion. The headline 2026 story is “revenue acceleration with optically worse earnings because the company is spending and taking FX pain,” not “profit inflection.”

This distinction matters because the market can get lazy with growth names and assume faster top-line growth automatically means a cleaner earnings arc. Dobot itself says otherwise. The company is effectively telling investors that 2026 may be the year when commercial momentum improves faster than statutory profitability. Given the capital raised and the stage of the business, that is believable. It also means 2026 results will be unusually vulnerable to how investors choose to normalize exchange losses and stock compensation.

Embodied-intelligence disclosure improved materially in July 2026, and it gives a better read on what is really happening. By 30 June 2026 the company said it had expanded embodied-intelligence customers to 231 and had already covered the full commercialization chain from signed orders to POC validation, small-scale delivery, and scaled deployment. Shipment amount for embodied-intelligence products in H1 exceeded RMB40 million. In hand, embodied-intelligence orders including framework agreements exceeded RMB60 million, of which industrial-manufacturing orders exceeded RMB10 million. Those are meaningful numbers for a segment that booked only RMB20.0 million of revenue in all of 2025. They are not yet numbers that prove a mature recurring business.

The mix within embodied intelligence is still the key factual brake on runaway optimism. Dobot’s 2026 reply letter breaks 2025 embodied revenue into research and education at RMB13.46 million, industrial manufacturing at RMB4.67 million, and commercial retail at RMB1.91 million. Research and education therefore represented 67.1% of 2025 embodied revenue, industrial 23.3%, and commercial 9.5%. The fair criticism here targets any valuation framework that treats 2025 embodied revenue as if it were already dominated by recurring industrial fleets, not the company itself; nearly every early-stage robotics market starts with education, labs, pilots, and showcase deployments. It was not.

The market today appears to be trading a hybrid of real fundamentals and reopened narrative optionality. Real fundamentals include the sharp H1 2026 revenue acceleration, the growth in industrial customers for embodied products, and a balance sheet well-funded enough to finance aggressive R&D. Narrative optionality includes the ChiNext process, the possibility that A-share investors will price the company on a richer “future industry” template, and the hope that Dobot’s industrial DNA will make its humanoid program more monetizable than a typical robotics startup. The share price no longer reflects the panic-free exuberance of March 2025. It still reflects belief that 2026 and 2027 can be the first years when embodied products become large enough to matter in the consolidated model.

The ChiNext process is now more advanced than it was a month ago, but still incomplete. Dobot announced in March 2026 that it planned to issue up to 48,883,933 A shares, equal to about 11.11% of current share capital and about 10.00% of enlarged capital before any over-allotment option. The intended net proceeds were about RMB1.2 billion, earmarked primarily for multi-legged robot R&D and industrialization, humanoid technology enhancement, marketing capability enhancement, and working capital. The application materials were accepted by the Shenzhen Stock Exchange on 27 April 2026. On 22 July 2026 the SZSE Listing Committee approved the proposed A-share offering and listing, but the company said the deal still requires approval from the China Securities Regulatory Commission, and there was no disclosed final price or listing date as of the research date. The clean way to think about dilution is therefore about 10% on the base plan, rising to about 11.33% if the full over-allotment option is exercised.

For valuation, earnings multiples are the wrong tool. On a cash-flow passthrough basis, reported net income remains negative and operating cash flow was also negative in 2025, though less negative than in 2024. The operating-cash-flow-to-net-income ratio on an absolute basis was only about 0.5x in 2025, and Dobot does not separately disclose maintenance capex versus growth capex. Because the company also capitalized no material development costs, reported earnings are not hiding an easy adjustment to owner earnings. The practical conclusion is simple: P/E is not meaningful, owner earnings are still negative, and Dobot should be valued primarily on forward sales and on the credibility of eventual margin conversion.

At the 24 July 2026 close of HK$25.22, and using the 2025 year-end share count of 439.96 million, Dobot’s market capitalization is about HK$11.10 billion. Using 2025 revenue translated at about 1 RMB = HK$1.1575, trailing price-to-sales is about 19.5x. This is far below the euphoric peak valuation implied by the March 2025 high, but still rich for an industrial hardware company that remains loss-making. The reason investors pay it anyway is obvious: they are discounting future embodied-intelligence growth and possible mainland re-rating. The reason to be careful is equally obvious: if that future arrives slower than hoped, there is little support from current earnings.

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

The important reading of that table is the spread, not the optimistic cell. Dobot’s valuation is already leaning on 2027 rather than 2026. This is why the base case can be roughly fair while the conservative case still leaves substantial downside. It is also why the stock can feel cheaper than it was at HK$80 and still not cheap enough for a fresh, balanced-risk purchase.

On expectation gaps, the market is probably underestimating how hard the middle phase will be. The next upside surprise only needs to be audited evidence that the 2026 embodied ramp is translating from shipments and pilot funnels into recognized revenue, repeat orders, steadier margins, and less cash drag, not “humanoids win the world.” The main downside surprise would be the opposite: upbeat top-line guidance followed by weak cash conversion, still-thin gross margins in embodied products, and no audited proof that scaled industrial deployments are arriving on schedule. As of 24 July 2026, the most important missing data are the actual interim numbers and the final terms of the A-share issuance.

The margin-of-safety verdict is clear. Relative to the conservative scenario value of about HK$16.1, the current share price carries no margin of safety. Relative to the base case of about HK$26.0, the stock is near fair value. If revenue were flat for the next three years and investors eventually valued Dobot closer to a conventional automation multiple, returns from the current price would look poor. This is therefore a good-company-or-at-least-interesting-company, bad-price-for-new-money case more than a broken-story bargain. The disciplined answer for a new investor is to wait for a materially lower entry or for materially stronger proof.

Risk analysis and tracking indicators

The first serious risk is that embodied-intelligence commercialization remains real but narrower than the headline suggests. The probability is medium to high because the company’s own disclosures still show a 2025 embodied revenue mix dominated by research and education, while 2026 funnel data include everything from signed orders to POC validation. The impact is high because the stock’s valuation premium depends on investors believing the segment can become a scaled industrial business. The observable indicators are simple: the share of embodied revenue coming from industrial manufacturing, the proportion of customers graduating from POC to 10-plus-unit deployments, and whether gross margin in embodied products stops falling. If those indicators stall, revenue can still grow while the multiple contracts.

The second risk is dilution as strategy, not emergency. Probability is high because Dobot already used the hot 2025 market to issue stock twice, and the ChiNext plan would add about 10% new shares before any greenshoe. The impact is medium to high because dilution is manageable when projects earn good returns, but expensive when that capital only extends a long proving period. The observable indicators are the final A-share offer price, the actual number of shares issued, and whether management begins using the richer balance sheet to fund measurable revenue growth rather than only longer-dated R&D. A company can create value while diluting shareholders; it just needs returns on the new capital to exceed the per-share drag.

The third risk is margin disappointment hidden inside fast revenue growth. Probability is medium. Impact is high. The 2025 prospectus data already show that embodied-intelligence margin ran below four-axis cobots and below its own 2024 level, while H1 2026 guidance says losses widen mainly because of FX, share-based payments, and heavier spending. This combination can mask a deeper issue or simply reflect a temporary investment phase. The indicators to watch are gross margin by product line when available, R&D as a percentage of revenue, selling expense growth relative to revenue growth, and operating cash flow. If audited revenue doubles but operating cash flow remains deeply negative, the business may be scaling activity faster than economics.

The fourth risk is valuation compression from category slippage. Dobot currently sits in an awkward but valuable intersection between industrial automation and embodied-AI narrative. If investors decide it belongs more with traditional automation equipment names than with frontier-robotics names, the multiple can fall even if operations improve. Probability is medium. Impact is high because the stock at HK$25.22 still trades on forward hope rather than current earnings. The observable indicators include Dobot’s own numbers, plus sentiment and comparative developments in listed peers and private benchmarks such as UBTECH and Unitree. When the sector narrative cools, high-multiple small caps often lose valuation support faster than revenue can replace it.

The fifth risk is policy and capital-markets dependence. Probability is medium. Impact is medium to high. Chinese policy is supportive toward robotics, and that support is part of the bull case, but the next leg of Dobot’s funding and potentially its valuation frame still depends on CSRC approval and then on actual A-share pricing and listing conditions. If approval is delayed, if market conditions weaken, or if the A-share gets priced less aggressively than expected, Dobot would lose a significant source of narrative momentum. The business would still exist. The market’s willingness to capitalize its future could change quickly.

Positive catalysts are easier to identify than to time. The cleanest one is an audited H1 2026 report that confirms revenue near the top of the RMB300–330 million range while showing better-than-feared adjusted losses or better cash conversion. The next is evidence that the embodied segment is shifting from education and pilot revenue toward repeat industrial deployments. A third is final CSRC approval and a well-priced ChiNext issue that broadens the investor base without signaling desperation. A fourth is any disclosure that 2026 industrial embodied customers are moving from POC to double-digit-unit deployments.

Negative catalysts are equally clear. A miss against the H1 2026 range would matter because management pre-disclosed those figures unusually early. A visible decline in embodied gross margin or a further jump in share-based compensation would invite the market to question quality of growth. Weak A-share timing or pricing would puncture one of the market’s preferred supports. Another poorly timed or more dilutive follow-on financing would also hurt, because it would suggest the 2025 and prospective 2026 capital raises are not enough to fund the transition.

The dashboard is useful only if it stays grounded in what changes the payoff. Revenue alone is not enough; Dobot can grow fast and still disappoint if revenue carries too much low-conviction embodied mix or if expenses keep outrunning gross profit. Embodied shipment value and customer funnel matter because they are the earliest public read on whether the second growth engine is moving from showroom logic to production logic. Cash flow matters because it is the bridge between technological promise and shareholder value. The A-share timetable matters because it changes both dilute-share economics and the stock’s valuation environment.

Research uncertainties

The biggest blind spot is that Dobot had not yet published audited or reviewed H1 2026 interim results by the research date, so the report necessarily relies on management’s unaudited operating figures and the Shenzhen exchange reply materials for the latest momentum read.

The second blind spot is embodied-intelligence unit economics below the segment level. Public materials now show segment revenue mix, shipment value, and gross margin, but they do not give enough audited disclosed detail to cleanly separate recurring industrial deployments from showcase, education, retail, and data-collection sales at the level a long-only investor would ideally want.

The third blind spot is A-share pricing. The share count is disclosed; the eventual offer price is not. This means dilution can be estimated, but per-share value creation from the raise cannot yet be judged with confidence.

The fourth blind spot is peer comparability. Global cobot incumbents, Chinese humanoid leaders, and domestic automation companies are all relevant, but they operate under different accounting standards, product mixes, and market narratives. Any peer comparison in this space is useful but imperfect.

Sources

Primary sources were Dobot’s HKEX prospectus, 2025 annual report, 2026 voluntary operating-figures announcements, and the 2026 Shenzhen exchange A-share application materials and reply letters. Secondary sources were used mainly for market-data confirmation, peer context, and industry framing, with preference given to HKEX, SEC, company IR pages, IFR, Reuters, and Yahoo Finance quote pages.

Cross-synthesis summary

Looked at vertically, Dobot has already proved one hard capability: it can turn robotics engineering into shipped products and then into capital access. This sounds obvious. It is not. Robotics is full of companies that can build a prototype, win attention, and still fail at the brutal middle work of distribution, support, cost reduction, product iteration, compliance, and customer patience. Dobot’s decade-long path from Kickstarter desktop robot to a top Chinese cobot vendor, then to a Hong Kong listing and an accepted ChiNext application, says the company does have real execution muscles. Its past success came from building usable products in a category with legitimate demand and then broadening that offering over time, not from a single lucky year or a one-off policy grant.

This vertical record also tells you what Dobot has not yet proved. It has not yet turned embodied intelligence into the economic center of the company. The financial statements still show a cobot maker with a new option attached, not an embodied-AI platform with an old business behind it. The market sometimes reverses those nouns because it prefers the futuristic one. The filings do not. They show that collaborative robots and composite systems still create the overwhelming majority of revenue and gross profit, while embodied products are still early in mix, still softer in margin than investors often imagine, and still heavily tilted toward research, education, pilots, and early deployment.

Looked at horizontally, Dobot’s real advantage is that it is one of the few public robotics names trying to build a humanoid business on top of a real industrial-robotics business rather than on top of aspiration alone, not that it is the most advanced humanoid company in public markets. This gives it something valuable: customer adjacency, manufacturing know-how, and a reason for industrial buyers to return calls. Against Universal Robots and Fanuc, that advantage is not enough to claim premium-brand trust. Against UBTECH and Unitree, it is enough to claim a more grounded industrial entry path. Dobot’s weakness can fade only if the company uses its installed base and its balance sheet to turn more of today’s pilot funnel into repeat industrial spend; scale gaps and brand gaps do not disappear on their own in a quarter.

This is why the current valuation deserves restraint. Today’s price is pre-spending a credible part of what investors think the company may earn if three things happen, rather than mainly rewarding it for what it already earned: the H1 2026 acceleration shows up in audited numbers, the embodied segment moves visibly toward industrial deployment, and the ChiNext capital-markets path completes on acceptable terms. If any of those slip, the market only needs less certainty to re-rate the stock lower, not a disaster. The most likely market misjudgment right now is that the transition from “many pilots and growing shipments” to “repeat industrial economics” may take longer than the share price would like, not that Dobot is fake.

Over the next year, the decisive variables are audited H1 2026 numbers, the embodied revenue mix, A-share progress, and cash conversion. Over the next three years, the decisive variables are whether embodied robotics becomes large enough to matter in consolidated gross profit and whether dilution produces attractive returns on capital rather than just more time. Over five years, the question is much broader: does Dobot become a scaled industrial-intelligence vendor with several robot forms, or does it remain a good cobot company with a perpetually interesting but still not decisive humanoid side business. Those are different endings and produce very different valuation regimes.

The company would become a materially better investment under either of two conditions. The first would be price: a retreat into the low-teens HKD area would create a better margin of safety against execution risk. The second would be proof: audited evidence that embodied robotics can move from roughly 4% of main-business revenue to a level that clearly changes consolidated growth and gross profit, without requiring ever-larger share issuance. I would re-examine the judgment if audited interim numbers showed a much sharper cash-flow turn than expected, if industrial embodied deployments scaled faster than the current customer-funnel picture implies, or if the eventual A-share pricing demonstrated unusually deep mainland demand without punitive dilution. I would also overturn the judgment if embodied revenue remained small, margin kept slipping, and the market still insisted on valuing the stock like a frontier platform.

Bull and bear reasons

The bull case starts with the fact that Dobot already built a real global cobot business, with 2025 shipments said to rank first worldwide, cumulative shipments above 100,000 units, and a customer list that includes more than 80 Fortune Global 500 companies.

A second bull point is that the 2026 embodied ramp is no longer just a product-launch story: by 30 June 2026 the company reported more than RMB40 million of embodied shipments, more than RMB60 million of orders including framework agreements, and 231 customers across the commercialization chain.

A third bull point is financial survivability. The company entered this transition with far more cash support than most young robot makers because the IPO, greenshoe, and two 2025 placings sharply strengthened the balance sheet and reduced leverage.

A fourth bull point is strategic fit. Dobot’s industrial-customer base gives it a more plausible path into real manufacturing deployments than many humanoid stories that begin in exhibitions, labs, or public-relations showcases.

The bear case starts with scale reality: embodied intelligence was only about RMB20 million of 2025 revenue and about 4% of main-business revenue, so today’s market story still runs far ahead of today’s segment size.

A second bear point is economics. Embodied-intelligence gross margin fell to 43.4% in 2025 from 49.9% in 2024, which is not yet the profile of a segment proving superior profitability as it ramps.

A third bear point is dilution. Dobot already increased the share count meaningfully through a greenshoe and two placings in 2025, and the proposed A-share issue would add about another 10% before any greenshoe.

A fourth bear point is valuation support. At roughly 19.5x trailing 2025 sales, the stock is no longer euphoric, but it still needs future success rather than current earnings to justify itself.

A fifth bear point is that 2026’s apparent acceleration comes with a wider first-half loss, and management itself attributes that to exchange losses, share-based payments, and higher spending, which means the path from growth to profitability remains unproven.

Pre-mortem

One plausible 50% down script over the next three years looks like this. Dobot reports brisk top-line growth through 2026, completes the A-share issue, and keeps adding embodied customers, but most of those customers remain in validation, education, or small-batch deployment. Industrial orders scale slower than expected, embodied gross margin drifts from the low-40s into the high-30s because price and support costs stay heavy, and operating cash flow remains negative. The market stops valuing Dobot as a frontier-robotics bridge and instead values it as a small loss-making automation hardware company. A revenue multiple that now sits in the low double digits on forward numbers compresses toward the high single digits. The stock could halve without the operating business collapsing.

A second 50% down script is more capital-markets-specific. The ChiNext process drags or prices less well than hoped, the market cools on all listed humanoid names after a competitor such as UBTECH or Unitree posts stronger commercial evidence, and Dobot is forced to prove that its middle-ground identity is worth premium valuation. If, at the same time, a new round of equity financing arrives before the old capital has generated visible returns, per-share value could compress sharply even with revenue still growing.

Final research conclusion

Dobot is worth following because it has already crossed the first credibility hurdle that defeats most robotics stories: it has a real core business. It is not worth chasing at the current price simply because that core business is still being asked to underwrite a much more speculative second act. The cobot franchise is real, the balance sheet is in good shape, and the embodied pipeline is moving. What is missing is proof that the new segment can become large, repeatable, and economically attractive enough to deserve a lasting premium over ordinary automation names.

For existing investors with a high tolerance for volatility, the current price is not irrational. For new money with balanced risk tolerance, it is still a waiting game. Bankruptcy risk or product irrelevance worry me less than a more ordinary but more damaging possibility: that Dobot becomes a very busy company with heavy narrative traffic, solid sales growth, and slower-than-hoped per-share value creation because the second curve takes longer and costs more than the market wants to admit. I would change my mind on the upside with audited evidence that industrial embodied deployments are scaling and that cash conversion is beginning to follow revenue. I would change it on the downside if the company keeps raising capital while embodied economics remain mostly aspirational.

【Company-profile scores】

• Fundamental quality: medium
• Growth: high
• Moat: medium
• Financial soundness: strong
• Management credibility: medium
• Valuation attractiveness: low
• Risk level: high
• Suitable investor type: high-risk speculation

【Investment rating】

• Rating: Hold
• One-line thesis: A real cobot franchise supports the story, but the stock still charges investors in advance for unproven embodied-robotics economics.
• 【Ideal Buy Price】12–14 HKD Basis: at least a 20% margin of safety below the conservative scenario value of about HK$16.1 per share.
• Acceptable hold price: 22–30 HKD
• Clearly overvalued price: 47 HKD and above
• Current-price classification: acceptable hold
• Whether to wait for a better price: yes. New buying becomes attractive either below roughly HK$14, or at a higher price only if audited results prove industrial embodied deployments are scaling with better cash conversion. The opportunity cost of waiting is missing a narrative re-rating if ChiNext pricing turns out stronger than expected.
• Target holding horizon: 1–3 years
• Expected annualized return: conservative about -26%; base about 2%; optimistic about 42% on an approximately 18-month to 24-month framework
• Max-loss risk: about 50% or more if embodied commercialization remains pilot-heavy, margins slip, and the market re-rates Dobot toward a conventional small automation-equipment multiple
• Reassessment-trigger signals: if audited H1 2026 revenue misses the disclosed range; if group gross margin falls below 43% for two consecutive reporting periods; if embodied revenue mix remains dominated by education and pilot deployments through 2027; if operating cash outflow deteriorates materially despite revenue growth; if the A-share issue is delayed, weakly priced, or followed by another dilutive raise

【Valuation Range】

• current: 25.22 HKD (close as of 2026-07-24)
• bear (conservative · ideal buy zone): [12, 14]
• base (fair · acceptable hold zone): [22, 30]
• bull (optimistic · above the clearly-overvalued line): [47, 55]

Other tickers mentioned

• 9880.HK: listed humanoid-robot benchmark used to contrast Dobot’s smaller embodied business with a more direct humanoid market narrative.
• TER.US: owner of Universal Robots, the key global cobot reference point for ecosystem strength and deployment credibility.
• 6954.TSE: Fanuc, used as the industrial-automation incumbent benchmark for reliability, service depth, and manufacturing trust.
• 300607.SHE: Topstar, used as a domestic automation peer whose valuation is anchored more by factory capex and equipment economics than by humanoid narrative.

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