Deep Robotics

1.1 Identity, footprint, and stage

Deep Robotics presents itself as an embodied-AI robotics company focused on industrial and hazardous-environment deployments rather than a pure research lab. Across official Chinese and US-facing pages, the company describes a product set spanning quadruped robots, humanoid robots, and core robotic systems, while the US site highlights current quadruped use cases in industrial inspections, emergency response, underground navigation, and scientific research. The assigned sources support a 2017 founding date and a Hangzhou base, but they do not support reader-grade precision on customer count, headcount, revenue, or locations beyond the company-reported footprint. The strongest scale statement in the retained pack is management's December 2025 claim that the company had already reached all 34 provincial-level regions in China and 44 countries/regions overseas. That is useful as a commercialization signal, but it is still company-issued evidence and should not be treated as the same thing as independently benchmarked installed-base data. On stage, the assigned evidence supports "private / pre-IPO marketplace tracked" more than it supports any active public-listing conclusion: EquityZen offers Deep Robotics as a pre-IPO name with shareholder-liquidity infrastructure, but the pack does not confirm a filed prospectus, a completed exchange process, or a current valuation.[CO001, CO002, CO003, CO006, CO015]

1.2 Leadership disclosure, founder-market fit, and governance gaps

Public leadership disclosure in the assigned evidence is narrow. The clearest named executive is Zhu Qiuguo, identified in the financing announcement as founder and CEO, while the other retained materials describe the founding team and wider technical bench more generically. EquityZen says the founding team is rooted in Zhejiang University, and the US about page says the disclosed core team spans universities including Zhejiang University, Shanghai Jiao Tong University, Beijing Institute of Technology, Wuhan University, NYU, UIUC, Georgia Tech, and Dortmund University of Technology. That is supportive of founder-market fit in legged robotics and autonomy, but it is not a substitute for a full executive roster or board map. Governance is therefore a real diligence gap at this stage. None of the assigned sources disclose a board list, committee structure, or named non-founder executives beyond the founder/CEO reference, which increases key-person concentration in the public record. The same gap matters for financing diligence because the company is clearly scaling into a more institutional shareholder base, yet the retained evidence does not show who currently governs investor rights, how oversight is structured, or whether post-Series-C reporting discipline has matured alongside commercialization.[CO001, CO003, CO004, CO010]

1.3 Capital formation and stakeholder map

The best-supported financing event in the pack is the December 2025 Series C. Deep Robotics said it raised more than RMB 500 million, with CMB International and ChinaAMC as joint lead investors and strategic participation from China Telecom- and China Unicom-affiliated funds. The same release named a long list of new and existing backers, which matters because it suggests the round was not just financial but also ecosystem-oriented: telecom-linked funds, industrial investors, and prior robotics-focused capital all appear in the roster. Even so, the evidence still stops well short of a full cap-table view. The assigned sources do not disclose post-money valuation, preference terms, liquidation protections, secondary volume, debt facilities, or whether any investor has governance or distribution rights beyond ordinary equity ownership. That means the stakeholder map is directionally useful for understanding who matters around the company, but it is not yet a substitute for a diligence-grade ownership model.[CO003, CO007, CO008, CO009]

1.4 Recent milestones and the main adverse signal in the pack

The retained milestone evidence clusters in 2025-2026 and shows a company trying to expand beyond a single robot form factor. Management said 2025 included the launch of the Lynx M20 wheeled-legged platform, the unveiling of the DR02 industrial humanoid, the establishment of an embodied-AI pilot production base, and completion of shareholding-system reform by the time of the December 2025 financing announcement. A July 2026 paid WAIC release then claimed that DR02 had already passed substation precision-operation tests and entered commercial deployment for power inspection, hazardous chemical operations, and disaster rescue. Together, those milestones imply an active move from quadruped inspection roots toward broader embodied-AI system commercialization. The main adverse signal in the assigned pool is not litigation or regulator action; it is perception risk. The retained Reddit thread repeatedly frames the robot platform as a likely military or social-control tool, showing how advanced mobility demos can trigger dual-use anxiety even when official materials emphasize industrial inspection, rescue, and research applications. That does not prove a commercial problem today, but it is a credible reputational watchpoint because dual-use perception can shape customer, regulator, and export-control scrutiny faster than it shows up in formal legal disclosures.[CO011, CO012, CO013, CO014, CO015, CO016]

1.5 Exhibits

2.1 Market boundary and substitutes

The assigned pool supports a clear market boundary before any sizing exercise: Deep Robotics is not best understood as a generic "humanoid" or broad embodied-AI vendor, but as a legged-robotics company whose monetized wedge is industrial field work. Hello China Tech's prospectus-based analysis says industry applications generated 79.33% of 2025 main business revenue, and it lists power-grid inspection, emergency firefighting, police and security operations, industrial monitoring, infrastructure patrol, and logistics as the paid use cases. Company-issued pages expand the adjacency set to energy, emergency response, industrial operations, manufacturing, infrastructure, education, and research, which is consistent with a product stack spanning industrial quadrupeds, wheeled-legged systems, and early humanoids. The main substitutes are therefore status-quo human patrols and hazardous-site inspection workflows, plus adjacent research/education platforms, not the entire robotics hardware market. Research and education remain relevant because the Lite3 line keeps a non-industrial entry point in the portfolio, but the evidence does not justify treating that adjacency as the same budget pool as industrial deployments.[CM001, CM002, CM003, CM014]

2.2 Buyer map and adoption path

The buyer map in the assigned pool is more concrete than a generic robotics narrative. Utilities and infrastructure owners appear as core buyers for inspection workflows; public-safety and emergency organizations appear as buyers for hazardous reconnaissance; and industrial operators or security teams appear as buyers for patrol, monitoring, and logistics-adjacent use cases. The company repeatedly frames adoption around places where human entry is difficult, dangerous, or slow, while independent reporting adds evidence that the installed base is moving beyond one-off demos into repeatable procurement. Hello China Tech cites cumulative deployment counts above 300 in both power inspection and police/security, more than 150 in fire rescue, and about 300 in industrial settings, while the company says its systems now package six sector solutions refined from more than 1,200 scenarios. That combination implies an adoption path in which a buyer first approves a risk-reduction use case, then standardizes procurement once a robot proves mobility, sensing, and reliability in a narrow workflow rather than across a broad labor category.[CM002, CM005, CM006, CM007, CM008]

2.3 Sizing lenses, drivers, and constraints

The strongest sizing evidence is company-specific and must be treated as a bounded lens rather than a market-total claim. The pool shows RMB 267.3 million of 2025 industry-application revenue, 681 X-series industrial quadrupeds sold at roughly RMB 287.5k each, and a 2026 Q1 snapshot of more than RMB 110 million in revenue with RMB 235 million of orders on hand. Those figures demonstrate a real, monetized wedge, and they explain why the company can pursue an IPO and keep funding humanoid expansion. But the same evidence also highlights the constraints that matter for valuation: the buyer base is narrower than consumer or research hardware, sales cycles are longer, safety-critical deployments require reliability in harsh environments, and the company's move into humanoids increases capital needs before that segment is equally proven. Tech in Asia's procurement examples help show that public and utility buyers can carry meaningful unit budgets, yet the assigned pool still does not provide a defensible third-party TAM/SAM/SOM or an independently reconciled spend model by geography and segment. Adoption is therefore best viewed as a hazardous-workflow automation market with credible proof of demand, but incomplete evidence on how large the reachable budget pool becomes outside today's proven niches.[CM001, CM004, CM007, CM009, CM010, CM011, CM012, CM013]

2.4 Exhibits

3.1 Competitive landscape and alternatives

The available evidence places Deep Robotics in a broad legged-robotics contest rather than a single-product race. Its public portfolio spans Jueying quadrupeds, LYNX wheeled-legged systems, DR humanoids, and joint components, while its own materials position those products across power inspection, emergency response, tunnels, construction, and research. The clearest named direct peer in the assigned independent evidence is Unitree, whose IPO application was described as running in parallel with Deep Robotics'. Public discussion also invokes Boston Dynamics and Unitree as reference points, but that discussion is anecdotal and cannot establish comparative performance. Incumbent fixed inspection equipment, human inspection teams, and customer-built integrations remain economically relevant solution paths, yet the assigned pool does not quantify their cost or performance. Accordingly, the landscape table distinguishes evidenced company facts from alternatives that still require primary diligence.[CP001, CP004, CP006, CP011, CP012]

3.2 Capability and portfolio positioning

Deep Robotics' differentiation claim rests on a vertically integrated perception-decision-execution stack and a product ladder that reaches from an ROS2-compatible research platform to industrial quadrupeds, wheeled-legged robots, and early humanoids. The official US site describes Lite3 as supporting an open SDK, LiDAR, depth cameras, and NVIDIA Jetson Xavier NX, which may lower prototyping friction for research buyers. Company materials also claim autonomous navigation, SLAM, real-time adaptation, and deployments in difficult industrial environments. These are meaningful scope signals, but most are company-issued rather than independent benchmark results. The positioning figure therefore scores only Deep Robotics' own portfolio on ordinal commercial-maturity and industrial-specialization axes; it does not manufacture unsupported rival coordinates. The central competitive question is whether industrial deployment know-how transfers into repeatable product advantage faster than component and locomotion capabilities commoditize.[CP004, CP005, CP006, CP007, CP008, CP009]

3.3 Pricing, distribution, and switching dynamics

No retained source supplies a verified public price, realized contract value, discount schedule, service attachment, or comparable total cost of ownership for Deep Robotics, Unitree, or Boston Dynamics. That prevents a defensible price-performance ranking. Deep Robotics does disclose an open research-oriented software posture for Lite3 and a modular autonomy suite across its broader platform. Those interfaces could reduce initial integration friction, while customer-specific mapping, perception, safety, workflow, and fleet-management work could create switching costs after deployment; the latter is an inference, not a disclosed retention metric. The company also claims field relationships with major industrial organizations, but the evidence does not reveal exclusivity, channel economics, repeat purchases, or whether customers multi-home across vendors. Buyers should therefore treat integration depth and service performance, not unavailable list prices, as the near-term diligence focus.[CP007, CP008, CP013, CP014]

3.4 Moat durability and displacement risk

The best evidence of durability is commercial rather than theatrical: third-party reporting attributes RMB 196 million, 58.11% of 2025 revenue, and 681 units to the Jueying X quadruped line. It also reports that the newer Shanmao M wheeled-legged line contributed RMB 74 million, while humanoids generated only RMB 823,000. This mix implies that Deep Robotics' present moat depends on industrial quadruped deployment, product reliability, and customer integration more than humanoid optionality. At the same time, anonymous discussion of the LYNX demonstration highlights both perceived mobility strength and concerns about battery life, tangling, and militarization; these comments are useful adverse signals but not engineering evidence. A parallel Unitree IPO and the absence of independently tested head-to-head metrics increase displacement uncertainty. Durable differentiation would require repeat-order cohorts, reliability data, service economics, and validated performance against direct peers, none of which appears in the assigned pool.[CP001, CP002, CP003, CP011, CP012, CP015, CP016]

3.5 Exhibits

4.1 Revenue model and public traction

Deep Robotics is monetizing industrial robotics rather than a pure software subscription. Company-issued materials describe a business spanning the R&D, production, sales, and service of quadruped robots, humanoid robots, and core robotic components, with deployments in power inspection, emergency response, industrial operations, security patrol, and education/research workflows. Independent summaries then fill in the commercial picture: 2025 revenue reached roughly RMB 337 million, gross margin moved above 52%, and embodied-intelligence robots accounted for the vast majority of main business revenue. Product-level evidence suggests the current revenue base is still anchored in quadruped and wheeled-legged systems rather than humanoids, so the right financial lens is industrial hardware monetization plus attached deployment/service work, not SaaS-style recurring revenue. [CI001][CI002][CI004][CI006][CI007]

4.2 Pricing signals and unit economics

The retained public record is strongest on realized product economics, not public list pricing. Robotichina reports that the X-series generated RMB 196 million from 681 units in 2025, implying roughly RMB 287.5 thousand per unit with a 54.4% gross margin, while the Lite series generated RMB 51.3 million from 1,850 units, implying roughly RMB 27.7 thousand per unit. Those disclosed economics are useful because they show Deep Robotics already sells premium industrial systems at much higher monetization than its lighter research/field platforms. But they do not answer the underwriting questions that matter most for forward revenue quality: contract duration, implementation fees, service attach, warranty reserve, realized discounting, or customer concentration. The company therefore has enough public evidence to show product-market monetization, but not enough to show how durable or repeatable the current margin profile will be. [CI008][CI009][CI014]

4.3 Capital needs and financing dependency

The latest clearly disclosed financing event is the December 2025 Series C round of more than RMB 500 million, which management said would support R&D, production-capacity expansion, and market development. Tech in Asia also reported that Deep Robotics had entered IPO guidance with preparatory work expected to finish by June 2026, while later independent summaries and company-linked releases described a STAR filing, new humanoid/embodied-AI ambition, and early 2026 commercial deployment efforts. That combination means Deep Robotics is not simply funding working capital for an already-stable business: it is trying to finance a broader scale-up that spans industrial quadrupeds, wheel-legged systems, humanoids, and production infrastructure. Because no retained public source discloses cash on hand, monthly burn, debt obligations, or runway, capital adequacy still depends more on financing access than on public operating cash evidence. [CI003][CI010][CI011][CI015]

4.4 Financial verdict and diligence blockers

The core positive is that Deep Robotics has crossed the threshold from robotics narrative into documented industrial revenue, first profitability, and credible product-level ASP evidence. The core negative is that the next phase requires investors to underwrite capital intensity and execution with incomplete disclosure: one independent analysis argues humanoid sales were still de minimis, the same source flags sub-80% utilization against an aggressive expansion plan, and EquityZen explicitly markets deeper cap-table, revenue/financial, and risk-factor materials only to accredited private-market users. In practical terms, that leaves public investors able to see growth and product economics, but not the cash conversion, concentration, preference stack, or support burden that determine whether today’s gross-margin improvement can survive a capacity and humanoid buildout. The right financial verdict is “commercially real, but not yet transparently underwritable.” [CI005][CI012][CI013][CI016]

4.5 Exhibits

5.1 Product scope and workflow fit

Deep Robotics is not selling a single showcase robot; the assigned evidence points to a multi-line embodied-robotics portfolio built around industrial field workflows. Official and third-party profile material consistently describe quadrupeds, wheeled-legged robots, humanoids, and standalone joints/components rather than a pure demo platform. The workflow emphasis is also concrete: the official surfaces repeatedly anchor use in power inspection, emergency response, underground/tunnel work, logistics, research, and education. That makes the product story more industrial-operations-led than consumer robotics-led, with a module map that already spans buyer types from utilities and rescue teams to research institutions.[CE001, CE003, CE004, CE013]

5.2 Architecture and deployment model

The strongest technical narrative in the pool is the company’s “Perception–Decision–Execution” stack paired with hardware designed for uneven, hazardous terrain. On the product side, the LYNX M20 page discloses autonomy, sensing, environmental hardening, compute, and I/O details at a level that is materially better than generic marketing copy, while the US surface also shows a research/developer-oriented Lite3 configuration with ROS2 and open SDK positioning. At the operating-model layer, Deep Robotics appears to commercialize the platform through payload integration, deployment support, and even RaaS packaging, but the public materials still stop short of the implementation detail a technical buyer would want before underwriting fleet rollout.[CE002, CE005, CE006, CE007, CE008, CE014]

5.3 Maturity and differentiation

Relative to many embodied-AI narratives, Deep Robotics looks more mature in industrial locomotion than in generalized humanoid claims. Third-party profiles frame the company’s installed base around X-series industrial inspection and patrol use cases, while external comparison copy distinguishes Deep’s rugged industrial positioning from the more cost-competitive consumer/research framing associated with Unitree. Independent coverage also suggests Deep is using its quadruped foundation to finance and justify a newer DR02 humanoid push rather than starting with humanoids as a zero-revenue bet. That is a real differentiation vector, but it also means the humanoid roadmap still deserves more proof of production deployments, reliability, and economics than the assigned pool provides.[CE009, CE010, CE011]

5.4 Trust, support, and diligence gaps

The evidence supports product seriousness more than it supports trust/compliance readiness. Official surfaces expose support-adjacent entry points such as downloads, FAQs, policy, service, and fault declaration, which is better than a brochure-only posture. But the same pool does not provide public APIs, payload SDK documentation for buyers, uptime reporting, warranty detail, cybersecurity architecture, privacy posture, or system-level certification evidence beyond IP66 environmental protection on LYNX M20. Developer-signal evidence is also thin: the available community proxy is a Reddit discussion reacting to a terrain demo, which shows attention but not a durable public ecosystem. The right technical read is therefore capable hardware with under-disclosed operating controls.[CE008, CE012, CE015, CE016]

5.5 Exhibits

6.1 Customer base is industrial-first, with utilities as the commercial anchor

Public evidence consistently places Deep Robotics with professional rather than consumer buyers. Official company materials describe deployments across energy, emergency response, industrial operations, and education, while independent profiles align that the X-series and related systems are used for industrial inspection, hazardous-site entry, and research programs rather than broad consumer adoption. That mix matters because it implies a buyer set dominated by utilities, industrial operators, public-safety agencies, and technical institutions with procurement and integration budgets, not impulse hardware demand. The strongest customer fit is power-grid inspection, with other sectors appearing as adjacent workloads or developer seeding channels rather than equal commercial pillars.[CU001, CU002, CU003, CU013]

6.2 Named customer proof exists, but it is concentrated and uneven in quality

The clearest production proof is in utility inspection. Independent and official sources both point to deployments across more than 100 substations with State Grid and China Southern Power Grid, and the same disclosures frame those relationships as long-term rather than one-off demos. Outside the Chinese utility base, the picture becomes thinner. Company and profile sources mention Singapore national-grid surveys, tunnel and corridor inspection, and deployments across dozens of countries, but they do not identify most counterparties, unit counts, contract terms, or what share of the footprint reflects paid recurring work versus pilots, showcases, or integrator-led projects. As a result, Deep Robotics has real adoption proof, but named proof is still narrower than the global footprint headline suggests.[CU003, CU004, CU005, CU006, CU007, CU010]

6.3 Durability and expansion remain under-disclosed, with humanoid traction still mostly roadmap-level

The company has enough external evidence to show real demand, but not enough to underwrite durability. Independent writeups still note that Deep Robotics has not shared customer counts, cohort behavior, or segment-level revenue split, which makes it difficult to judge concentration, renewal depth, or service attach. The disclosed financial and order figures show continued momentum in industrial quadrupeds, yet they do not separate new-logo demand from follow-on fleet expansion. The same caution applies to the humanoid story: public 2026 deployment language comes primarily from a paid company release, while the best independent evidence still describes the humanoid line as early and commercially negligible versus the core quadruped base. That leaves the customer thesis investable on utility inspection today, but speculative on broader land-and-expand and humanoid upsell.[CU008, CU009, CU011, CU012]

6.4 Exhibits

7.1 Capital-market and regulatory exposure

Deep Robotics' capital-market risk is not whether investor interest exists, but whether disclosure catches up before the IPO window closes. Tech in Asia says the company entered IPO guidance and expected preparatory work to finish by June 2026, while RobotChina says a STAR Market filing was accepted in May 2026 and targets roughly RMB 2.5 billion of fundraising at a reported RMB 13.9 billion valuation. Humanoids Daily separately says the company is using a revenue-generating quadruped base to fund a more capital-intensive humanoid push. That combination makes financing risk real even after the reported 500 million yuan round: investors still lack public cap-table terms, liquidation preferences, and accessible primary filing detail. Kill criteria are monitorable: if filing milestones slip, if valuation moves ahead of disclosed unit economics, or if humanoid-scale spending rises before the quadruped business shows durable cash conversion, the risk should remain a track-not-buy condition.[CR004, CR005, CR006, CR007, CR012]

7.2 Operational, quality, and serviceability exposure

The core operating risk is that Deep Robotics is marketing robots into environments where failure costs are unusually high. Company-issued materials place the systems in substations, tunnels, landslides, flood control, hazardous chemical operations, and disaster rescue, which means uptime, autonomy quality, and operator support matter more than a demo-grade success narrative. The same support terms that help document the product also show friction: there are no returns, free warranty coverage excludes a long list of misuse and harsh-environment scenarios, the customer may bear shipping and paid-repair costs, and the published after-sales terms are explicitly limited to mainland China. Public materials therefore prove real use, but not yet a robust global service stack. Investors should treat field-failure rates, warranty claims, overseas spare-parts coverage, and safety or cybersecurity certifications as explicit diligence gates rather than assuming them from the deployment logos.[CR001, CR002, CR003, CR008, CR009, CR013]

7.3 Concentration and dependency exposure

The partner and revenue base still looks narrower than the product narrative suggests. The U.S. about page highlights a short list of enterprise names-State Grid, China Southern Power Grid, Baosteel, and Fluke-which is strong validation but also evidence that a few industrial accounts and channels matter a lot. RobotChina additionally says quadruped and wheeled robots still drive more than 95% of revenue, so the humanoid line is strategically important without yet being proven as a second economic engine. That means diversification risk operates on both the customer axis and the platform axis. If a utility or heavy-industry spending cycle pauses, or if humanoid scale-up absorbs capital faster than it wins production demand, the company could face margin pressure even with solid top-line momentum. The most useful monitoring indicators are repeat orders outside energy inspection, disclosed revenue mix by product family, and conversion of reported orders on hand into cash collections.[CR010, CR011, CR012, CR013]

7.4 Exhibits

8.1 Recommendation and underwriting posture

The evidence supports continued engagement, but not an aggressive entry posture. On the positive side, Deep Robotics has fresh capital, multiple reported industrial deployments, and an expanding product set that now spans quadruped, wheel-legged, and humanoid platforms. Those points matter because they show the company is no longer only an R&D story. However, the public record is still missing the variables that determine whether a private round or secondary price is attractive: the current valuation, the preference stack, the dilution path, and a clean bridge from product traction to durable earnings power. That makes the correct recommendation track rather than buy. The company may be improving fundamentally, but without disclosed pricing terms and operating-quality metrics, investors would be paying for option value without being able to check whether the option is already fully priced.[CV001, CV002, CV005, CV006, CV008, CV012, CV014]

8.2 Financing context, price support, and entry discipline

The latest financing and IPO signals improve financing optionality, but they do not establish fair value. Yicai and Tech in Asia both report a roughly CNY500 million funding round in 2025, while Tech in Asia and Humanoids Daily point to A-share IPO guidance and a completed shareholding reform consistent with listing preparation. EquityZen adds a useful market signal because it shows that private secondary liquidity exists, yet the public-facing page keeps the cap table, funding history by share class, liquidity preferences, and financials behind accreditation and login gates. In practice, that means public investors can observe demand for access, but not the information needed to judge whether a quoted secondary mark is clean or burdened by preference overhang. Entry discipline therefore has to stay conservative: no stretching on a narrative premium until management provides current valuation and term transparency.[CV001, CV002, CV003, CV008, CV009, CV012, CV014]

8.3 Scenario framing and comparable lens

The bull case is that Deep Robotics turns today’s industrial foothold into a scaled embodied-AI platform before the public market window closes, using fresh capital to broaden the product line and convert more inspection, utility, and emergency-response deployments into repeatable programs. The base case is more restrained: quadruped demand continues, but the public record remains too thin on segment margins, repeat orders, and service mix to justify premium private-market pricing. The bear case is not that the company lacks technology; it is that humanoid expansion and IPO preparation absorb capital faster than disclosed commercial metrics mature. Public comparable evidence is also incomplete. A third-party industry analysis frames Deep Robotics against Boston Dynamics, ANYbotics, and Unitree, and Humanoids Daily explicitly places it behind Unitree in scale, but that still falls short of a true valuation comp set because Deep itself has not disclosed audited revenue, profit, or unit-shipment detail.[CV004, CV005, CV006, CV007, CV010, CV011, CV013]

8.4 Final diligence asks and thesis-break triggers

The remaining diligence agenda is straightforward and high leverage. First, request the latest priced round documents, share-class stack, liquidation preferences, anti-dilution mechanics, and any VAM or special rights tied to the IPO timeline. Second, request operating data that can separate real commercialization from headline momentum: revenue by product family, gross margin by hardware versus service, repeat-order cohorts, backlog, customer concentration, warranty costs, and field-service burden. Third, ask management to reconcile the commercialization story around DR02 and Lynx with actual shipped units, signed contracts, and realized pricing. The thesis breaks if IPO timing slips without a corresponding increase in disclosure, if humanoid investment outpaces monetization in the core quadruped business, or if customer traction proves too concentrated to support premium private-market pricing.[CV002, CV004, CV006, CV008, CV012, CV013, CV014]

8.5 Exhibits