Leju Robotics (乐聚机器人)

From university spinout to IPO candidate: whether China's third-ranked humanoid maker can convert research credibility and state-backed capital into durable commercial scale

---

How to Read This Report

This report applies a four-tier evidence taxonomy throughout. Every substantive claim is tagged or contextualised according to the following scheme:

A choreographed demonstration video is not treated as proof of autonomous real-world capability. A shipment figure is not treated as proof of productive deployment. A partnership announcement is not treated as proof of a paying customer relationship. Where the underlying dossier is thin, this report says so plainly rather than filling the gap with inference dressed as fact.

Inline citations use bracketed numerals keyed to the Sources list in §14. Only URLs present in the research dossier are cited.

---

01Executive Overview

Leju Robotics Technology Co., Ltd. (乐聚机器人) is a Shenzhen-based humanoid robot manufacturer that has, in the space of roughly nine years, travelled from a university laboratory at Harbin Institute of Technology to the threshold of a public listing on China's ChiNext exchange. The company's trajectory is, in several respects, a compressed version of the broader Chinese humanoid robotics story: academic origins, rapid product iteration, heavy state-backed capital inflows, and a race to establish commercial credibility before the market's current enthusiasm cools.

The headline numbers are striking. In 2025, Leju reported CNY 258 million in revenue, sold 577 units of its flagship KUAVO humanoid robot, and achieved a three-year compound annual revenue growth rate of 118.68% ⁶⁸. An independent data provider, IDC, ranked the company third globally in humanoid robot shipments and second in full-size bipedal humanoid shipments for the same year ⁶. In October 2025, the company closed a pre-IPO funding round of approximately 1.5 billion yuan (roughly $200 million USD), valuing the business at CNY 4.327 billion post-round ⁴¹⁴. The investor list spans state-linked funds, Tencent, and a roster of 25 named backers in total ¹⁴.

Against those positives sit a set of structural tensions that any serious analysis must confront. Leju posted a net loss of CNY 69.78 million in 2025 despite its revenue scale ⁶⁸ — a loss that is unremarkable by growth-stage hardware standards but that underscores the cost burden of building and scaling a humanoid platform. The company's primary customer base remains concentrated in research, education, and data-collection contexts ⁶⁸; its industrial deployments, most visibly a partnership with Schaeffler announced in March 2026 ¹¹, are early-stage and have not yet produced independently verified evidence of sustained productive operation. The KUAVO robot's autonomous capabilities are well-supported for locomotion tasks by published research ¹⁸²⁰ but are less clearly evidenced for the complex manipulation tasks that would justify industrial deployment at scale.

The competitive environment is severe. Unitree Robotics, Leju's most direct domestic rival, has demonstrated an ability to undercut on price dramatically — its G1 humanoid was announced at approximately $16,000, and its broader portfolio includes platforms at $5,500 ⁴³¹. Leju's stated target price of approximately $20,000 for KUAVO ³ positions it above Unitree's most aggressive offerings, which creates a differentiation imperative: the company must demonstrate capability and reliability advantages that justify the premium. Whether it can do so at the pace required by its IPO timeline is the central question this report examines.

The IPO itself is a significant data point. ChiNext guidance acceptance was confirmed, with the company targeting the issuance of no more than 20 million shares and gross proceeds of CNY 2.6 billion ⁶⁸⁹. The implied post-IPO valuation would represent a substantial step up from the pre-IPO round. That ambition is coherent only if Leju can demonstrate a credible path to profitability — which, in turn, requires converting its current research-and-education customer base into a broader industrial and service deployment base, and doing so before competitors with deeper pockets or lower cost structures crowd out the available market.

This report examines the evidence for and against that thesis across seven sections in Part I, covering the company's history, product portfolio, technology stack, research output, media evidence, and commercial reality to date.

Latest news

• Chinese factory assembling humanoid robot every 30 minutes while Tesla, Figure and Boston Dynamics struggle

  en.clickpetroleoegas.com.br·2026-06-03GENERAL
• Global Commercial Humanoid Robotics Market Research 2025-2030 with Focus on China and United States Featuring Figure AI, Agility, Tesla, UBTECH, Boston Dynamics, Apptronik, AgiBot, Unitree, and Leju

  GlobeNewswire·2026-05-19GENERAL

---

02The Leju Robotics Story

Origins at Harbin Institute of Technology

Leju Robotics was founded in 2016 by Ling Xiaokun and a group of graduates from Harbin Institute of Technology (HIT), one of China's premier engineering universities and a longstanding centre of robotics research ⁴⁷. The HIT connection is not incidental: the university has produced a disproportionate share of China's robotics talent and has institutional links to several of the country's most prominent robotics companies. Leju's founding team brought with them both technical expertise in bipedal locomotion and an academic network that would later manifest in research collaborations and published work.

The spinout model — university research group converts into commercial entity — is a well-worn path in Chinese deep technology, and Leju followed it with reasonable fidelity. The company's early years were spent developing the AELOS line of small educational robots, a product category with established demand in Chinese schools and STEM programmes. This was a pragmatic commercial decision: small educational robots have lower manufacturing complexity, shorter development cycles, and a customer base (schools, after-school programmes, competition organisers) that is less demanding about reliability than industrial or service deployments. AELOS gave Leju a revenue base and a manufacturing learning curve while the team worked on more ambitious platforms.

The AELOS Moment: 2018 Winter Olympics

The company's most prominent early validation came in February 2018, when AELOS robots performed at the closing ceremony of the PyeongChang Winter Olympics in South Korea ⁴⁷. This was a choreographed performance — a point this report flags explicitly, because a synchronised dance routine at a televised ceremony is not evidence of autonomous task capability in unstructured environments. What it does demonstrate is that Leju had, by 2018, achieved sufficient manufacturing quality and motion control reliability to execute a high-visibility, zero-fault-tolerance public performance. For a two-year-old company, that is a meaningful operational milestone, even if its technical significance is narrower than the marketing narrative suggests.

Scaling Up: ROBAN and the Path to KUAVO

Following the AELOS period, Leju developed ROBAN, a medium-sized platform targeting research and university laboratory use. ROBAN occupies a product tier between the small educational AELOS and the full-size humanoid that would become KUAVO. The ROBAN line appears to have served primarily as a research platform and a stepping stone in the company's hardware development capability, though detailed independent sales data for ROBAN is not publicly available [UNKNOWN].

The KUAVO programme represents Leju's most significant technical and commercial bet. The robot — approximately 1.4 metres tall, 45 kilograms, with 26 degrees of freedom per the official launch video ²² — was developed as a full-size bipedal humanoid targeting research, industrial, and service markets. The KUAVO launch was accompanied by the announcement of HarmonyOS integration, reflecting a strategic alignment with Huawei that would deepen into a formal innovation centre partnership ⁴⁷. The decision to build on HarmonyOS is notable: it ties Leju's software stack to a Chinese domestic operating system with its own ecosystem dynamics, which has both advantages (state support, domestic supply chain alignment) and risks (limited international developer community, dependency on Huawei's roadmap).

Funding History and Capital Structure

Leju's funding history reflects the broader surge of Chinese state and institutional capital into humanoid robotics from 2023 onwards. The company has attracted 25 investors in total per CBInsights ¹, with the most significant round being the October 2025 pre-IPO raise of approximately 1.5 billion yuan ⁴¹⁴. Named investors in that round include Shenzhen Investment Holdings Capital, Shenzhen Longhua Capital, Qianhai Basic Investment, Shijingshan Industry Fund, Orient Precision Engineering, Top Group, and CITIC Jinshi, alongside Tencent ⁴¹⁴¹⁵.

The ownership structure, disclosed in IPO filing documents, shows no single controlling shareholder. Ling Xiaokun, the founder and chief executive, holds 15.92% directly and controls 33.72% in total. Chang Lin holds 6.01% and An Ziwei holds 4.15% ¹². This diffuse structure is typical of Chinese technology companies that have gone through multiple funding rounds, and it has implications for governance: the founder retains effective control through a combination of direct and indirect holdings, but the presence of state-linked funds as significant investors introduces a layer of policy alignment that is both a commercial asset and a potential constraint on strategic independence.

The IPO Trajectory

The decision to pursue a ChiNext listing is consistent with the profile of Chinese deep-technology companies at Leju's stage. ChiNext, the Shenzhen Stock Exchange's growth enterprise board, has hosted several robotics and automation companies and is the natural venue for a company of Leju's size and sector ⁶⁸. The company completed IPO guidance acceptance — a formal regulatory step in China's listing process — and is targeting the issuance of no more than 20 million shares for gross proceeds of CNY 2.6 billion ⁶⁸⁹. The IPO date referenced in the dossier is 19 May 2026, though the status of the listing process as of the coverage date of this report is not independently confirmed [UNKNOWN].

The CNY 2.6 billion target implies a post-IPO valuation substantially above the CNY 4.327 billion post-money valuation from the October 2025 round ⁶¹⁴. Whether the public market will sustain that valuation depends heavily on how investors read the company's path to profitability, the credibility of its industrial deployment pipeline, and the competitive dynamics of the humanoid sector at the time of listing.

One conflict in the dossier is worth noting here. A video source cited in the research suggests a STAR Market (Shanghai) listing with a valuation target of approximately $7 billion ²³. The dossier's reconciliation analysis assigns higher credibility to the ChiNext/CNY 2.6 billion figure, which is supported by detailed IPO filing documentation ⁶⁸. The STAR Market figure may reflect confusion with another company or an earlier, superseded plan. This report treats ChiNext as the verified venue.

Leadership and Culture

Ling Xiaokun's background as a HIT-trained engineer and his continued role as the company's primary technical and strategic leader gives Leju a founder-led character that is common among Chinese robotics startups of this generation. The company's research output — discussed in §5 — reflects an academic orientation that distinguishes it from purely commercially-driven competitors. Whether that academic culture translates into durable product advantages or creates a tension with the execution discipline required to scale manufacturing and customer support is an open question [EDITORIAL INFERENCE].

---

03Product Portfolio: What Leju Robotics Actually Sells

Overview

Leju's product portfolio spans three size classes and multiple target markets, a breadth that reflects both the company's evolution over nine years and a deliberate strategy of addressing multiple customer segments simultaneously. The three primary lines are AELOS (small), ROBAN (medium), and KUAVO (large, flagship). The dossier also references KuaFu and LuBan series names, which appear to be alternate designations within the full-stack humanoid programme rather than entirely separate product lines [COMPANY CLAIM, confidence 0.95 per dossier].

AELOS: The Foundation Product

AELOS is Leju's original commercial product and the platform that established the company's manufacturing capability and brand recognition. It targets the Chinese educational robotics market — a segment with substantial institutional demand driven by national STEM curriculum initiatives and a well-developed competition circuit. The 2018 Winter Olympics performance ⁴⁷ remains the most widely cited public demonstration of AELOS capability, though as noted above, a choreographed ceremony performance is a limited proxy for general-purpose autonomous capability.

Detailed technical specifications for AELOS are not prominently featured in the available dossier sources, which reflects the product's positioning as an educational tool rather than a research or industrial platform. Independent sales data for AELOS is not publicly available [UNKNOWN]. The product's commercial significance to Leju's current revenue base is unclear, given that the 2025 revenue and sales figures cited in the dossier (CNY 258M, 577 units) appear to refer specifically to KUAVO ⁶⁸.

ROBAN: The Research Bridge

ROBAN occupies the middle tier of Leju's portfolio, targeting university research laboratories and advanced educational institutions. It appears to have served as a development platform that allowed Leju to accumulate experience with more complex bipedal systems before committing to the full-size KUAVO programme. Independent technical specifications and sales data for ROBAN are not publicly available in the dossier [UNKNOWN]. Its current commercial relevance relative to KUAVO is unclear.

KUAVO: The Flagship and the Commercial Bet

KUAVO is the product on which Leju's IPO case, its industrial partnerships, and its competitive positioning all rest. The robot was unveiled with an official launch video ²² that provides the most reliable publicly available technical specifications: approximately 1.4 metres tall, 45 kilograms, 26 degrees of freedom, a top walking speed of 4.6 kilometres per hour, and a vertical jump capability exceeding 20 centimetres. The platform runs HarmonyOS, Huawei's domestic operating system, and is described by the company as open-source in its software stack ²²⁴.

A conflict in the available data concerns the robot's height and weight. The official launch video and Robot Report cite 1.4 metres and 45 kilograms ²²⁴. A separate video source references a robot of 1.8 metres and 80 kilograms with a claimed lifting capacity of 100 kilograms ²³. The dossier's reconciliation analysis concludes that the 1.4 metre / 45 kilogram figures are more credible for the standard KUAVO, and that the larger figures may refer to a different or updated model — possibly a heavier-duty variant — or may reflect confusion with another company's product in the same video context. This report adopts the 1.4 metre / 45 kilogram specification as the verified baseline for the standard KUAVO, while flagging that a larger variant may exist [COMPANY CLAIM for the larger variant; UNKNOWN whether it is a distinct product].

Similarly, the degrees-of-freedom count is contested. The official launch video gives 26 DOF ²², while a Robot Report article citing Leju references more than 40 DOF ⁴⁷. The dossier reconciliation assigns higher confidence to the 26 DOF figure as the more precisely sourced datum. The 40+ DOF figure may include hand and finger articulation in a more capable variant, or may reflect a different model generation. This report uses 26 DOF as the verified baseline for the standard KUAVO.

KUAVO Specification Summary

KUAVO's Software and AI Integration

The KUAVO platform's software stack is a significant part of its value proposition. The company has integrated a unified adaptive locomotion policy — a single trained model deployed across varied terrain types — supported by published research ¹⁸²⁰. Natural language robot programming via a ROS and large language model integration is described in a peer-reviewed paper ²⁰, and a multimodal AI model (LingBot-VLA) is being developed in collaboration with Ant Group's Robbyant subsidiary ¹³²¹. The platform also incorporates multimodal training data contributions of approximately 10,000 to 20,000 hours ¹⁹.

These are substantive technical capabilities, and the research papers provide more independent validation than is typical for Chinese robotics startups at this stage. However, the distinction between research-stage capability and deployed product feature is important and is addressed in detail in §4 and §5.

KUAVO's HarmonyOS Integration and Huawei Partnership

The decision to build KUAVO on HarmonyOS and to establish a joint innovation centre with Huawei in Shenzhen ⁴⁷ is strategically significant in multiple dimensions. It aligns Leju with one of China's most powerful technology companies and with the domestic technology ecosystem that the Chinese government has been actively promoting. HarmonyOS provides a real-time operating system with IoT and edge computing capabilities that are relevant to robotics applications. The partnership also gives Leju access to Huawei's sensor, connectivity, and AI chip ecosystem.

The risks of this alignment are also real. HarmonyOS has a limited international developer community compared with Linux-based alternatives, which constrains the pool of external contributors to Leju's open-source software stack. Dependency on Huawei's roadmap introduces a single point of strategic risk. And the association with Huawei — a company subject to US export controls and international scrutiny — creates potential complications for any future international commercial expansion [EDITORIAL INFERENCE].

Pricing and Market Positioning

Leju's stated target price for KUAVO is approximately $20,000 ³. This figure comes from a single commerce/analysis source and has not been independently verified; it should be treated as a company roadmap claim rather than a confirmed market price. If accurate, it positions KUAVO above Unitree's most aggressively priced humanoid offerings (the G1 at approximately $16,000, with other platforms lower) ⁴³¹, which creates a differentiation requirement. The company's response to this competitive pressure — emphasising software capability, research-grade reliability, and ecosystem integration — is coherent in principle but has not yet been validated by market outcomes at scale [EDITORIAL INFERENCE].

Products & versions

KUAVO

Leju's flagship full-size humanoid robot (~1.4 m, ~45 kg, 26 DOF), featuring onboard AI locomotion, HarmonyOS integration, and targeting research, industrial, and commercial service applications.

ROBAN

Leju's medium-sized humanoid robot platform, positioned between the small AELOS and the flagship KUAVO, targeting education and research markets.

AELOS

Leju's small humanoid robot line aimed at the education market; notably performed at the 2018 Winter Olympics closing ceremony in South Korea.

---

04Technology Stack: Strengths and the Work That Remains

Framing the Assessment

Evaluating a robotics company's technology stack requires distinguishing between three categories that are frequently conflated in company communications: what the research papers demonstrate in controlled conditions, what the product documentation claims as deployed features, and what independent evidence confirms as reliable in real-world operation. For Leju, the first category is reasonably well-evidenced, the second is present but unverified, and the third is thin. This section works through each layer of the stack with that framework in mind.

Locomotion: The Strongest Technical Foundation

The most credible technical claim Leju can make is in bipedal locomotion. A 2025 paper on terrain-conditioned generative motion priors (T-GMP) ¹⁸, which lists Leju-affiliated authors, describes a unified adaptive locomotion policy trained in a single run and deployed across varied terrain types including stairs, slopes, and uneven ground. The approach uses generative motion priors conditioned on terrain geometry to produce natural, adaptive gait patterns without requiring separate policies for each terrain type.

This is a meaningful technical contribution. The unified policy approach addresses one of the persistent challenges in bipedal locomotion research: the brittleness of terrain-specific controllers when encountering out-of-distribution conditions. The paper's methodology is consistent with the state of the art in sim-to-real transfer for legged robots, and the results — demonstrated on the KUAVO platform — provide more independent validation of locomotion capability than a demonstration video alone would offer.

The official launch video ²² shows KUAVO walking on varied surfaces, navigating stairs, and executing the >20 centimetre vertical jump. These demonstrations are consistent with the research paper's claims. However, the video is a curated demonstration, not an independent operational review. The gap between a well-executed demonstration and reliable autonomous operation across the full range of real-world conditions encountered in industrial or service deployments is substantial and is not bridged by either the paper or the video [EDITORIAL INFERENCE].

Manipulation: Claimed but Underverified

Manipulation capability — the ability to grasp, move, and interact with objects in unstructured environments — is where the evidence base becomes thinner. The company claims real-time AI perception and adaptation for manipulation tasks, including sensor skin for tactile feedback [COMPANY CLAIM]. The LingBot-VLA model, developed in collaboration with Robbyant ¹³²¹, is described as a vision-language-action model for embodied AI tasks. A paper on pragmatic VLA foundation models ²¹ is associated with this work.

VLA models are an active and promising area of robotics AI research, and the collaboration with Ant Group's Robbyant brings additional AI research resources to the programme. However, VLA models for manipulation remain a research frontier: they perform well on trained task distributions and degrade on out-of-distribution scenarios. No independent third-party operational review of KUAVO's manipulation capability in real-world deployments has been identified in the dossier. The Schaeffler partnership ¹¹ is the most concrete evidence of industrial manipulation deployment, but it was announced in March 2026 and no operational results have been independently reported.

The community evidence is sobering in this context. A Reddit thread on factory automation ³¹ reflects a widely-held view among robotics practitioners that humanoid robots face fundamental reliability and cost-effectiveness challenges on factory floors — that legs are unnecessary on flat surfaces, that manipulation reliability at the required cycle times and error rates is not yet demonstrated, and that humanoids are unlikely to undercut minimum-wage labour in the near term. These are not Leju-specific criticisms; they apply to the humanoid category broadly. But they are relevant to assessing the credibility of Leju's industrial deployment claims.

Software Architecture: ROS, LLM Integration, and HarmonyOS

The KUAVO software stack integrates the Robot Operating System (ROS) as a middleware layer, with a large language model interface described in a published paper ²⁰. The ROS-LLM framework allows natural language task specification — a user describes a task in plain language, and the system generates a structured robot programme — with task feedback and structured reasoning to handle failures. This is a genuine capability that reduces the programming burden for research and education users, and the paper provides independent validation of the approach.

HarmonyOS provides the underlying operating system, with real-time control capabilities relevant to the low-latency requirements of bipedal locomotion. The open-source positioning of the software stack is a deliberate strategy to build a developer community and reduce the barrier to adoption in research settings. Whether this strategy has produced a meaningful external developer community is not publicly documented [UNKNOWN].

The multimodal AI capabilities — described as encompassing text, audio, images, video, 3D data, and embodied actions ¹⁹ — are associated with the RoboEgo system, which is described in a preprint ¹⁹ as an omnimodal model with native full duplexity. The paper is a system card rather than a peer-reviewed experimental study, which limits the strength of the evidence it provides. The claimed 10,000 to 20,000 hours of multimodal training data contribution is a company claim [COMPANY CLAIM] that has not been independently verified.

Hardware Design: Actuators, Sensors, and Manufacturing

The KUAVO's 26 DOF configuration is consistent with a full-size humanoid designed for bipedal locomotion and upper-body manipulation. The specific actuator technology — whether Leju uses proprietary actuators, licensed designs, or off-the-shelf components — is not detailed in the available dossier [UNKNOWN]. This is a significant gap: actuator quality, torque density, and thermal management are among the most consequential determinants of a humanoid robot's real-world reliability, and the choice of actuator supplier has implications for both performance and supply chain resilience.

The company's launch of automated production lines, referenced in the IPO guidance context ⁹, suggests investment in manufacturing scale. However, the details of production capacity, yield rates, and quality control processes are not publicly disclosed [UNKNOWN]. The 577 units sold in 2025 ⁶⁸ implies a production capacity of at least that volume, but whether the automated line represents a step change in capacity or an incremental improvement is unclear.

The Sensor Skin Claim

The company references sensor skin — a distributed tactile sensing layer — as a feature of KUAVO [COMPANY CLAIM]. Tactile sensing is an active research area with genuine relevance to manipulation tasks, particularly for grasping objects of variable compliance and geometry. However, the maturity of deployed tactile sensing in commercial humanoid robots is generally low across the industry, and no independent validation of Leju's sensor skin capability in operational conditions has been identified [UNKNOWN].

Summary Assessment

---

05Research, Papers, Authors and Labs

Research Output in Context

Leju Robotics occupies an unusual position among Chinese humanoid robotics companies in that it has a documented record of contributing to peer-reviewed and preprint research literature. This is a direct consequence of its HIT origins and the academic orientation of its founding team. The research output is not voluminous — the dossier identifies four relevant papers — but the quality and relevance of those papers to the company's core product claims is higher than is typical for a company at this commercial stage.

It is important to be precise about what research output does and does not demonstrate. A published paper showing that a locomotion policy works on a KUAVO robot in a research setting is meaningful evidence that the underlying technical approach is sound. It is not evidence that the same capability is reliably deployed in the 577 units sold to customers, or that it performs consistently across the range of real-world conditions those customers encounter. The gap between research demonstration and deployed product reliability is one of the defining challenges of the robotics industry, and Leju is not exempt from it.

T-GMP: Terrain-Conditioned Generative Motion Priors [18]

The most technically substantive paper in the dossier is the T-GMP paper ¹⁸, which describes a framework for generating natural and versatile humanoid locomotion across varied terrain types. The approach conditions a generative motion prior on terrain geometry, allowing a single trained policy to adapt its gait to stairs, slopes, uneven ground, and other challenging surfaces without requiring separate controllers for each terrain type.

The paper is significant for several reasons. First, it addresses a genuine technical challenge — terrain-adaptive locomotion — that is directly relevant to KUAVO's deployment in real-world environments. Second, it demonstrates results on the KUAVO platform, providing a direct link between the research and the product. Third, the unified policy approach, if it generalises as claimed, would represent a meaningful advantage over competitors relying on terrain-specific controllers.

The paper's limitations should also be noted. As an arXiv preprint ¹⁸, it has not necessarily completed peer review at the time of the coverage date. The experimental conditions, while described as varied, are controlled research settings rather than unstructured real-world deployments. The sim-to-real transfer gap — the difference between simulation-trained policies and real-world performance — is acknowledged in the locomotion research community as a persistent challenge, and the paper's results do not fully resolve it.

RoboEgo: Omnimodal Model with Native Full Duplexity [19]

The RoboEgo system card ¹⁹ describes a multimodal AI model designed for embodied robot interaction, incorporating text, audio, image, video, 3D, and action modalities with a claimed native full-duplex interaction capability — meaning the system can simultaneously process input and generate output without the turn-taking latency of conventional dialogue systems. The paper claims contributions of approximately 10,000 to 20,000 hours of multimodal training data.

As a system card rather than an experimental paper, RoboEgo ¹⁹ provides less rigorous independent validation than T-GMP ¹⁸. System cards describe design choices and capabilities without necessarily providing the controlled experimental comparisons that would allow an independent reader to assess performance relative to alternatives. The full-duplex claim is technically interesting and, if validated, would be relevant to natural human-robot interaction. However, the evidence base for this claim in the dossier is limited to the system card itself.

ROS-LLM: Embodied AI with Task Feedback [20]

The ROS-LLM paper ²⁰ is the most directly product-relevant of the four papers, describing a framework for natural language robot programming that integrates ROS with a large language model. The framework allows users to specify tasks in natural language, generates structured robot programmes, and incorporates task feedback to handle failures and ambiguities. The paper is published on arXiv and has Leju-affiliated authors.

This capability is directly relevant to KUAVO's positioning as a research and education platform, where reducing the programming burden for non-specialist users is a genuine value driver. The ROS-LLM approach is consistent with a broader trend in robotics towards natural language interfaces, and the paper's structured reasoning component addresses a known weakness of naive LLM-to-robot pipelines (hallucinated actions, failure to handle exceptions). The paper provides credible evidence that this capability exists as a research prototype; its deployment status in the commercial KUAVO product is not independently confirmed [UNKNOWN].

A Pragmatic VLA Foundation Model [21]

The VLA paper ²¹ is associated with the LingBot-VLA collaboration between Leju and Robbyant ¹³. Vision-language-action models represent the current frontier of robot learning for manipulation tasks, combining visual perception, language understanding, and action generation in a single model. The paper describes a pragmatic approach to VLA design that prioritises real-world deployability over benchmark performance.

The collaboration with Robbyant — Ant Group's embodied AI subsidiary — brings significant AI research resources to this programme ¹³. Ant Group's scale and AI expertise are genuine assets. However, the paper is a preprint and the LingBot-VLA model is described as under development rather than deployed. The gap between a promising VLA research prototype and a reliably deployed manipulation capability in commercial robots is substantial, and the dossier provides no independent evidence that this gap has been closed for KUAVO.

Author and Laboratory Affiliations

The research papers list authors affiliated with Leju Robotics and, in some cases, with academic collaborators. Specific author names and their institutional affiliations beyond Leju are not detailed in the dossier [UNKNOWN — the dossier does not provide author-level attribution beyond the company affiliation]. The HIT connection suggests ongoing academic collaboration, but the specific nature and scope of current HIT involvement in Leju's research programme is not publicly documented.

<!-- module: papers --> <!-- module: authors-labs --> <!-- module: repos --> <!-- module: datasets -->

---

06Media Evidence Library: What the Videos Prove

Methodology for Video Evidence Assessment

This section applies a deliberate standard: demonstration videos are assessed for what they actually show, not for what the accompanying narration or marketing copy claims. A robot walking up stairs in a controlled laboratory setting is evidence of stair-climbing capability in that setting. It is not evidence of reliable stair-climbing in unstructured environments, of the same capability being present in all shipped units, or of the capability being maintained over extended operational periods. This distinction matters because the gap between a curated demonstration and deployed reliability is where humanoid robotics companies most frequently overstate their position.

KUAVO Official Launch Video [22]

The official KUAVO launch video ²² is the primary video evidence source for the robot's hardware capabilities. It shows the robot performing bipedal walking on varied surfaces, navigating stairs, executing the claimed >20 centimetre vertical jump, and demonstrating upper-body motion. The video provides the most reliable publicly available technical specifications: 1.4 metres, 45 kilograms, 26 degrees of freedom, 4.6 kilometres per hour top speed.

What the video proves: the KUAVO robot, at the time of filming, was capable of controlled bipedal locomotion on the surfaces shown, could execute a vertical jump of the claimed height, and had the physical dimensions

08Markets and Use Cases

Leju Robotics has articulated a broad target market spanning education, research, industrial logistics, healthcare, elder care, home services, and commercial services ⁴⁶. The breadth of that list is itself a signal worth interrogating: companies that claim every vertical simultaneously are often still searching for the one or two use cases where their technology genuinely clears the economic bar. The evidence available allows a reasonably clear ranking of where Leju actually operates today versus where it aspires to operate.

Education and Research: The Proven Core

The most defensible reading of Leju's 577 KUAVO units sold in 2025 ⁶⁸ is that the overwhelming majority went to universities, research institutes, and technical education programmes. This is consistent with the company's HIT heritage, its open-source software posture, and the fact that research customers tolerate higher failure rates, accept human supervision during tasks, and value programmability over reliability. The AELOS small robot line has an established track record in educational robotics competitions and classroom settings, and the 2018 Winter Olympics closing ceremony performance in South Korea ⁴ — while a marketing event rather than a commercial deployment — demonstrated that the company could execute a choreographed multi-robot showcase at international scale.

Research customers purchasing KUAVO are primarily interested in locomotion research, manipulation policy development, and embodied AI data collection. The collaboration with Robbyant (Ant Group) explicitly targets the generation of real-world training data for embodied AI models ¹³, which positions Leju's robots as instruments of AI research as much as end-use products. This is a legitimate and potentially lucrative niche: the global shortage of high-quality embodied AI training data means that a robot capable of generating diverse, real-world interaction data has value to AI labs independent of whether it can perform economically useful work autonomously.

The ROBAN medium-sized platform occupies a similar research and education niche, offering a lower-cost entry point for institutions that cannot justify KUAVO's price. The existence of three distinct size classes is a sensible portfolio strategy for an education-and-research primary market, where budget constraints vary enormously between a secondary school robotics club and a well-funded university lab.

Industrial Logistics: Emerging but Unproven

The Schaeffler partnership, announced in March 2026 ¹¹, is the most concrete signal that Leju is attempting to move beyond research and into industrial deployment. Schaeffler is a serious industrial partner — a global manufacturer of precision components with genuine operational requirements — and the fact that it chose Leju for a humanoid robotics collaboration rather than a more established industrial automation supplier is noteworthy. However, the partnership announcement does not constitute evidence of paid deployment at scale. The terms, scope, and timeline of the Schaeffler collaboration are not publicly disclosed, and it is entirely consistent with the announcement being a pilot or proof-of-concept arrangement.

The use case logic for humanoid robots in industrial settings is contested. Community analysis raises the pointed observation that legs are unnecessary on flat factory floors, that wheeled or fixed-arm robots are cheaper and more reliable for repetitive tasks, and that humanoids are unlikely to undercut minimum wage labour costs in the near term ³¹. These are legitimate structural objections, not mere pessimism. The counterargument — that humanoid form factor allows deployment in facilities designed for humans without infrastructure modification — is theoretically sound but has not been demonstrated at commercial scale by any humanoid manufacturer globally, not just Leju.

The specific industrial use cases Leju targets include heavy load transport and logistics within facilities ⁶⁹. At 45 kg with 26 DOF ²², the standard KUAVO is not obviously suited to heavy industrial lifting; if the 1.8 m / 80 kg variant mentioned in one source ²³ refers to a heavier-duty model, that specification conflict matters commercially and should be resolved by prospective industrial customers before procurement.

Healthcare and Elder Care: Long-Horizon Aspiration

Healthcare logistics — moving supplies, medications, and equipment within hospital environments — and elder care assistance are both cited as target markets ⁴⁶. These are plausible long-term applications for humanoid robots with sufficient dexterity and reliability, but they are also among the most demanding in terms of regulatory approval, liability management, and operational reliability requirements. No evidence in the dossier confirms any healthcare or elder care deployment by Leju. These markets should be treated as aspirational for the foreseeable future.

Commercial Services and Data Collection

Commercial service deployments — reception, guidance, and customer interaction in retail or hospitality settings — represent a market where Chinese robotics companies have more demonstrated traction than in Western markets, partly because labour economics and cultural acceptance differ. Leju lists commercial services as a target ⁶, and the KUAVO's natural language interaction capabilities via LLM integration ²⁰ are relevant here. However, the competitive pressure from wheeled service robots (which are cheaper, more reliable, and already deployed at scale in Chinese hotels and malls) makes this a difficult market for a legged humanoid to enter on cost grounds alone.

Data collection is the most intellectually honest near-term commercial use case. Deploying KUAVO units to generate embodied AI training data — whether for Leju's own model development or for third-party AI labs — monetises the robot's capabilities at a stage where full task autonomy is not required. The Robbyant collaboration ¹³ and the cited figure of 10,000 to 20,000 hours of multimodal training data contributed ¹⁹ suggest this is already a real revenue stream, though the financial contribution relative to the CNY 258M total revenue is not disclosed.

Use Case Maturity Assessment

The honest summary is that Leju is a research and education robotics company with industrial ambitions. The 577 units sold and CNY 258M revenue are real achievements, but the customer base is almost certainly dominated by institutions that value the platform as a research instrument rather than a productive worker. The transition from research instrument to economic asset is the defining challenge of the next three to five years.

---

09Competitive Landscape

Leju operates in a Chinese humanoid robotics sector that has become extraordinarily crowded in a short period. IDC's ranking of Leju third globally in humanoid shipments in 2025 ⁶ is a meaningful data point, but it requires context: the global humanoid market is still small enough that third place does not imply scale, and the ranking reflects units shipped rather than units productively deployed.

The Unitree Problem

The most immediate competitive threat to Leju is Unitree Robotics, also Shenzhen-based, which has achieved international visibility with its H1 and G1 humanoid platforms and its four-legged robots. Unitree's G1 humanoid is priced at approximately $16,000 at launch, with the company signalling further price reductions ⁴. The dossier notes a $5,500 humanoid price point attributed to Unitree ³, which — if accurate for any shipping configuration — represents a severe pricing challenge for Leju's reported target of approximately $20,000 ³.

Unitree competes on price and on international distribution. Its robots appear frequently in Western research labs and maker communities, giving it a brand presence that Leju has not matched outside China. Leju's response — emphasising open-source HarmonyOS integration, the Huawei ecosystem, and a more comprehensive software stack — is a reasonable differentiation strategy, but it is more compelling to Chinese institutional buyers than to international research customers who may prefer ROS-native platforms.

Domestic Chinese Competitors

The Chinese humanoid landscape includes a number of well-capitalised competitors, several of which have stronger industrial pedigrees or deeper AI integration than Leju:

Fourier Intelligence has focused on rehabilitation robotics before pivoting toward general-purpose humanoids, giving it a head start in regulated healthcare applications. Its GR-1 and GR-2 platforms are direct KUAVO competitors in the research market.

UBTECH Robotics is a more established Chinese robotics company with consumer, education, and enterprise products. It has greater international distribution and brand recognition than Leju, though its humanoid ambitions are more recent.

AgiBot (AgileBot) is a Shanghai-based humanoid startup that has attracted significant funding and media attention. The dossier conflict around IPO venue ²³ may reflect confusion between Leju and AgiBot, suggesting the two companies are sometimes discussed in the same breath by analysts.

Zhiyuan Robotics (Galbot) and Astribot represent the newer wave of Chinese humanoid startups with strong AI-first positioning. Both have attracted venture capital on the strength of AI capabilities rather than hardware maturity.

DEEP Robotics and Anybotics (the latter Swiss-based) compete in the quadruped space that overlaps with Leju's ROBAN platform.

International Competitors

Internationally, the relevant comparators are:

Boston Dynamics Atlas — the technical benchmark for humanoid locomotion, but priced and positioned for enterprise and government customers rather than research volume sales. Not a direct commercial competitor at Leju's price point.

Figure AI and Apptronik — US-based humanoid startups with significant venture backing and partnerships with major manufacturers (BMW, Mercedes-Benz). Their robots are not yet in volume production, but their industrial partnerships are more advanced than Leju's Schaeffler pilot.

Agility Robotics (Digit) — the most commercially advanced humanoid in Western markets, with Amazon warehouse deployments. Digit is purpose-built for logistics rather than general-purpose research, which limits direct comparison but illustrates what a focused industrial humanoid deployment actually requires.

1X Technologies — Norwegian humanoid startup with a different form factor philosophy, targeting home and care applications.

Competitive Positioning Matrix

Leju's Defensible Advantages

Three genuine advantages are worth acknowledging. First, the HIT academic lineage provides access to research talent and credibility with Chinese university buyers that purely commercial startups lack. Second, the Huawei HarmonyOS integration is a meaningful differentiator within China's technology ecosystem, particularly as Chinese institutions face pressure to adopt domestic software stacks. Third, the open-source positioning of the software platform — if genuinely maintained — lowers the barrier for research customers to build on KUAVO, creating a community flywheel that proprietary platforms cannot easily replicate.

The vulnerability is that none of these advantages translates cleanly to industrial customers, who care about reliability, safety certification, total cost of ownership, and integration with existing manufacturing execution systems. Leju has not publicly demonstrated progress on any of these industrial requirements.

Competitive comparison

Robot	Maker	Autonomy	Conf.
iRobot Roomba Combo 10 Max	iRobot	Autonomous	0.90
Mobile ALOHA (Stanford)	Stanford University	Teleoperated	0.90
1X NEO	1X Technologies	Remote-Assisted	0.90

---

10Geopolitical Context and Constraints

Leju Robotics is not merely a robotics company; it is a Chinese technology company pursuing an IPO on a domestic exchange, integrating Huawei's operating system, backed by state-affiliated capital, and operating in a sector that multiple governments have identified as strategically significant. Each of these facts carries geopolitical weight.

The Chinese State as Structural Enabler

The composition of Leju's investor base is instructive. Shenzhen Investment Holdings Capital, Shenzhen Longhua Capital, Qianhai Basic Investment, and Shijingshan Industry Fund ¹¹⁰ are all state-affiliated or state-adjacent investment vehicles. This is not unusual for a Chinese deep-tech company at Leju's stage — state capital routinely co-invests with private capital in sectors the government has designated as strategic priorities. Humanoid robotics has been explicitly identified as a priority in China's 14th Five-Year Plan and subsequent policy documents, with the Ministry of Industry and Information Technology setting targets for humanoid robot commercialisation by 2025 and mass production by 2027.

The practical consequence of state backing is that Leju benefits from subsidised access to industrial parks, preferential procurement from state-owned enterprises, and a degree of protection from the kind of predatory pricing competition that might otherwise make its economics unviable. The Schaeffler partnership ¹¹ is notable precisely because it involves a German industrial company choosing to work with a Chinese humanoid manufacturer — a signal that at least some Western industrial players are willing to engage with Chinese humanoid suppliers despite the broader geopolitical climate.

The Huawei Dimension

The integration of HarmonyOS as the operating system for KUAVO ²² is commercially logical within China but geopolitically loaded outside it. Huawei is subject to US export controls that restrict its access to advanced semiconductors and that have prompted many non-Chinese companies to avoid deep technical dependencies on Huawei's ecosystem. For Leju, HarmonyOS integration means:

• Strong positioning within China's domestic market and with Chinese institutional buyers who are themselves under pressure to adopt domestic technology stacks.
• A meaningful barrier to international sales in markets where Huawei's involvement triggers procurement review or outright exclusion — particularly US federal agencies, European defence-adjacent institutions, and any customer subject to ITAR or similar export control regimes.
• Dependency on Huawei's continued development of HarmonyOS for robotics, which is not guaranteed and introduces a single-point-of-failure risk in the software stack.

The Huawei innovation centre co-located in Shenzhen ⁴ deepens this dependency. Whether it constitutes a strategic asset or a strategic liability depends almost entirely on the customer's geography and regulatory environment.

Export Control and Technology Transfer Risks

Leju's robots incorporate advanced actuators, sensors, and AI inference hardware. The specific semiconductor content of KUAVO is not publicly disclosed, which means it is not possible to assess from public information whether the robot's bill of materials includes components subject to US export controls. This matters in two directions: Leju may face restrictions on importing certain advanced chips for its own production, and its robots may face export restrictions when sold to certain international customers.

The broader US-China technology decoupling creates a structural ceiling on Leju's international ambitions. Research universities and commercial customers in the United States, and increasingly in Europe, face institutional pressure — and in some cases legal requirements — to scrutinise procurement from Chinese technology companies. This does not make international sales impossible, but it raises the cost and complexity of international market entry significantly.

The IPO as a Geopolitical Signal

The decision to list on ChiNext rather than a US exchange ⁶⁸ reflects both the post-2021 regulatory environment for Chinese technology companies seeking US listings and a deliberate strategic choice to anchor Leju's capital structure within China's domestic financial system. A ChiNext listing provides access to Chinese retail and institutional capital, aligns with government preferences for domestic listings of strategic technology companies, and avoids the disclosure requirements and political exposure of a US listing.

The CNY 2.6 billion fundraising target ⁶ is ambitious relative to the company's CNY 258M revenue and CNY 69.78M net loss ⁶⁸. The implied revenue multiple of approximately ten times trailing revenue is consistent with growth-stage technology valuations in China's domestic market but would face scrutiny in more sceptical capital markets. The IPO's success will depend significantly on the continued enthusiasm of Chinese retail investors for humanoid robotics as a thematic investment, which is itself sensitive to policy signals and media narratives that the government can influence.

Taiwan Strait and Supply Chain Concentration

Leju's Shenzhen manufacturing base ⁹ places it within the Pearl River Delta electronics manufacturing ecosystem, which provides unmatched access to component suppliers, contract manufacturers, and engineering talent. It also concentrates supply chain risk in a geography that is subject to tail-risk scenarios related to Taiwan Strait tensions. This risk is not unique to Leju — virtually every major electronics manufacturer has similar exposure — but it is worth noting for international customers evaluating long-term supply chain resilience.

---

11The Hype, the Real and the Ugly

The humanoid robotics sector is structurally prone to hype. The combination of anthropomorphic form, AI narrative, and Chinese technology nationalism creates conditions in which claims are amplified without scrutiny and scepticism is dismissed as failure of imagination. Leju is neither the worst offender in this regard nor entirely innocent of it. A disciplined separation of what is real, what is plausible but unverified, and what is misleading is essential for any serious evaluation.

What Is Genuinely Real

The revenue and sales figures are credible. CNY 258M in revenue and 577 KUAVO units sold in 2025 ⁶⁸ are reported consistently across multiple independent sources and are the kind of specific, auditable figures that would appear in IPO filings subject to regulatory scrutiny. These are real numbers. They are not large numbers by the standards of mature industrial automation, but they represent genuine commercial traction.

The academic and research credentials are genuine. The HIT spinout origin ⁴, the published research papers on locomotion policy ¹⁸ and multimodal AI ¹⁹²⁰²¹, and the Robbyant collaboration ¹³ reflect real technical work by real researchers. The unified adaptive locomotion policy described in the T-GMP paper ¹⁸ is a substantive research contribution, not a marketing document.

The funding is real. The Pre-IPO round of approximately CNY 1.5 billion ¹⁰¹⁴ with 25 named investors ¹ is confirmed by Bloomberg ¹⁴ and multiple independent sources. The capital exists.

The Schaeffler and Robbyant partnerships are real announcements. Whether they represent paid commercial contracts or exploratory collaborations is not disclosed, but the partnerships themselves are confirmed by multiple sources ¹¹¹³.

The AELOS Olympic performance happened. The 2018 Winter Olympics closing ceremony deployment ⁴ is a verified historical fact, not a claim.

What Is Plausible but Unverified

The $20,000 target price point ³ is a vendor roadmap claim from a single source. It is plausible as an aspiration — the economics of Chinese manufacturing at scale could support it — but it has not been independently verified and may not reflect current transaction prices.

The claim of real-time AI perception and adaptation for manipulation tasks is plausible given the research papers cited, but no independent teardown, operational review, or third-party assessment of KUAVO's manipulation reliability in real deployments has been published. The locomotion capabilities are better evidenced than the manipulation capabilities.

The 3rd-place global ranking in humanoid shipments ⁶ cites IDC data, which is a credible source, but the methodology underlying that ranking — how IDC defines a humanoid robot, whether it includes all size classes, and what counts as a shipment versus a deployment — is not publicly available. The ranking is plausible but should not be treated as a precise competitive fact.

The LingBot-VLA model ²¹ is described as a collaboration between Leju and Robbyant. The paper exists and describes a vision-language-action model. Whether this model is deployed in production KUAVO units or remains a research artefact is not confirmed.

What Is Misleading or Requires Significant Qualification

Choreographed demo videos are not evidence of autonomous task capability. The KUAVO launch video ²² shows impressive locomotion — walking, jumping, navigating terrain — but these demonstrations are prepared environments with known conditions. The gap between a prepared demo and reliable autonomous operation in an unstructured real-world environment is substantial and well-documented across the humanoid robotics industry. Leju's marketing materials do not always make this distinction clear.

Partnership announcements are not customer confirmations. The announcements of collaborations with Huawei, Schaeffler, Robbyant, Alibaba, and Haier ⁴¹¹¹³ are real events, but they do not confirm that any of these organisations has paid for KUAVO robots in volume, deployed them in production environments, or achieved the operational outcomes that the partnerships nominally target. The distinction between a memorandum of understanding, a pilot agreement, and a production contract matters enormously and is not made clear in available public information.

The IPO valuation implies a growth trajectory that is not yet evidenced. A CNY 4.327 billion post-round valuation ¹⁰ on CNY 258M revenue and a CNY 69.78M net loss ⁶⁸ implies that investors expect very rapid revenue growth. The 118.68% three-year CAGR ⁶ is impressive, but sustaining that rate from a larger base requires either significant industrial deployment wins or a dramatic expansion of the research and education market. Neither is guaranteed.

The DOF conflict is unresolved and commercially significant. The discrepancy between 26 DOF ²² and "more than 40 DOF" ⁴ for KUAVO is not a minor rounding difference. Degrees of freedom directly affect manipulation capability, and a customer purchasing KUAVO for dexterous manipulation tasks needs accurate specifications. The conflict has not been publicly resolved by Leju.

The height and weight conflict matters for industrial applications. The difference between a 1.4 m / 45 kg robot and a 1.8 m / 80 kg robot ²²²³ is not trivial for industrial deployment planning, safety assessments, or facility design. If Leju has multiple KUAVO variants with significantly different physical specifications, this should be clearly communicated; if the larger figure refers to a different product entirely, that should also be clarified.

The Structural Ugly

The deepest problem for Leju — and for the humanoid robotics sector broadly — is that the economic case for humanoid robots in industrial settings remains unproven at any company's scale. The community critique ³¹ that humanoids are a non-starter for factory automation because legs are unnecessary, reliability is insufficient, and costs exceed minimum wage labour is not irrational pessimism. It reflects the actual experience of industrial automation engineers who have evaluated humanoid robots against the alternatives.

Leju's response to this challenge — emphasising AI capability, open-source software, and ecosystem partnerships — is appropriate for a research and education market but does not directly address the reliability, safety certification, and total cost of ownership questions that industrial customers ask. Until a humanoid manufacturer — Leju or any other — can demonstrate sustained, reliable, economically positive deployment in a real industrial environment with independently verified metrics, the industrial use case remains a hypothesis.

Claim tracker

KUAVO operates autonomously for locomotion and manipulation tasks using onboard AI, including walking complex terrain, jumping, and real-time AI perception and adaptation — without remote operators performing tasks.Unknown

Research papers (arXiv [18][21]) confirm a unified adaptive locomotion policy for humanoid robots, and no evidence of remote operators was found; however, no independent third-party operational review verifies real-world manipulation task reliability at scale, and community sources [31] raise legitimate reliability concerns about humanoids in deployed settings.

Leju Robotics sold 577 KUAVO robots and achieved CNY 258M (~$35M USD) in revenue in 2025, ranking 3rd globally in humanoid robot shipments per IDC.Supported

Two independent news sources [6][12] citing IPO filing documents report 577 units sold and CNY 258M revenue; IDC ranking is cited by an independent news source [6], though the IDC methodology and exact shipment count breakdown are not publicly detailed.

KUAVO is a general-purpose humanoid robot capable of serving industrial, healthcare, elder care, and home service markets — not just research and education.Not supported

The dossier confirms 577 units sold primarily to research, education, and data-collection customers [6][12], with industrial deployment limited to a single announced Schaeffler partnership [11]; no independent evidence of operational deployments at scale in healthcare, elder care, or home service contexts was found.

Leju Robotics has deployed KUAVO in real industrial settings through its partnership with Schaeffler (announced March 2026).Not supported

The Schaeffler partnership announcement [11] confirms a collaboration agreement was signed, but no independent source verifies that KUAVO robots are operationally deployed on Schaeffler production floors at any scale beyond a pilot or proof-of-concept.

KUAVO's hardware specifications are: ~45 kg, 26 DOF, 1.4 m tall, 4.6 km/h top speed, capable of jumping over 20 cm.Unknown

The official KUAVO launch video [22] and Robot Report [4][7] provide these specs, but all sources are vendor-adjacent; a conflicting video source suggests 1.8 m / 80 kg for what may be a different or updated model, and no independent teardown or third-party hardware verification has been published.

Leju's LingBot-VLA model (developed with Ant Group's Robbyant) enables multimodal embodied AI — processing text, audio, images, video, 3D data, and physical actions — for real-world robot use.Unknown

The Robbyant partnership and VLA collaboration are confirmed by an independent news source [13], and a related arXiv paper [21] describes a pragmatic VLA foundation model; however, the specific LingBot-VLA system's real-world performance on KUAVO has not been independently benchmarked or verified outside vendor and partner announcements.

Leju Robotics targets a ~$20,000 price point for KUAVO by late 2025, positioning it as cost-competitive against other full-size humanoid robots.Not supported

The $20,000 target price comes from a single commerce/analysis source [3] with low confidence (0.72) and is a vendor roadmap claim; no independent source confirms this price was achieved, and the dossier notes Unitree's competing humanoid at $5,500 creates severe downward pricing pressure that makes the $20,000 figure commercially questionable.

AELOS performed at the 2018 Winter Olympics closing ceremony in South Korea, demonstrating Leju's early real-world deployment capability.Supported

Robot Report [4][7] independently confirms the AELOS performance at the 2018 Winter Olympics closing ceremony; this is a verified historical deployment, though it reflects the small AELOS platform's capabilities rather than the flagship KUAVO humanoid.

---

12Future Scenarios

The following scenarios are editorial inferences from the available evidence. They are not predictions; they are structured possibilities intended to help readers calibrate their assessments of Leju's trajectory.

Scenario A: Research-to-Industrial Bridge (Base Case, Moderate Probability)

In this scenario, Leju successfully completes its ChiNext IPO, raises approximately CNY 2.6 billion ⁶, and uses the capital to scale manufacturing, deepen the software stack, and convert the Schaeffler pilot ¹¹ and similar industrial partnerships into recurring revenue contracts. The research and education market continues to grow as Chinese universities expand robotics programmes and as embodied AI data collection becomes a recognised commercial activity. Industrial revenue begins to contribute meaningfully to the top line by 2027-2028, though it does not yet dominate.

The conditions required for this scenario: the IPO closes successfully; at least one industrial partnership converts to a production contract with publicly verifiable metrics; the KUAVO platform achieves sufficient reliability for supervised industrial tasks; and the Chinese government continues to support humanoid robotics through procurement and policy.

The risks: IPO market conditions deteriorate; industrial reliability targets are not met; Unitree or another competitor captures the research market on price; international sales remain blocked by geopolitical friction.

Scenario B: Research Niche Consolidation (Moderate Probability)

In this scenario, Leju does not successfully bridge to industrial deployment within the next three to five years. The company remains primarily a research and education robotics supplier, growing steadily but not at the rate implied by its current valuation. The IPO proceeds but at a lower valuation than targeted, or is delayed. Revenue grows to CNY 500-700M by 2028 driven by research volume, but the net loss persists as R&D spending remains high.

This is not a failure scenario — a profitable, growing research robotics company with a strong academic ecosystem is a viable business. But it would imply that the current valuation is stretched and that investors expecting industrial-scale returns will be disappointed.

Scenario C: Ecosystem Play Succeeds (Lower Probability, High Upside)

In this scenario, the Huawei HarmonyOS integration and the Robbyant embodied AI collaboration create a genuine platform effect. KUAVO becomes the reference hardware for a Chinese embodied AI ecosystem, with Leju earning revenue not just from robot sales but from data licensing, software subscriptions, and ecosystem participation fees. The LingBot-VLA model ²¹ or its successors become commercially significant AI products in their own right.

This scenario requires that the embodied AI data economy develops faster than currently evidenced, that Leju successfully positions itself as the hardware layer of that ecosystem rather than being commoditised by it, and that the Huawei ecosystem retains its domestic dominance. It is the highest-upside scenario but also the one most dependent on factors outside Leju's direct control.

Scenario D: Competitive Squeeze (Lower Probability, Significant Downside)

In this scenario, Unitree's aggressive pricing ⁴ and international distribution erode Leju's research market share faster than Leju can compensate with industrial wins. The IPO is delayed or undersubscribed. The Schaeffler partnership does not convert to production contracts. Chinese government procurement shifts toward other suppliers. Leju's net losses deepen as revenue growth slows.

The conditions that would signal this scenario: Unitree announces significant price reductions below $10,000 for a comparable platform; Leju's IPO is withdrawn or significantly downsized; no industrial deployment announcements emerge in 2026-2027; key technical staff depart.

Scenario E: Geopolitical Disruption (Low Probability, Severe Impact)

A significant escalation in US-China technology tensions — including expanded export controls on semiconductor components used in KUAVO, or restrictions on Chinese robotics companies operating in key international markets — could materially impair Leju's international growth ambitions and potentially its domestic supply chain. This scenario is not specific to Leju but would affect the entire Chinese robotics sector.

Scenario Probability Summary

These probabilities are editorial inferences, not quantitative forecasts. The base case (Scenario B) reflects the observation that most robotics companies that begin in research and education take longer than their investors expect to achieve industrial-scale deployment.

---

13What to Watch: A Live Monitoring Checklist

The following indicators are the most diagnostically useful signals for tracking Leju's actual progress against its stated ambitions. Readers following this company should monitor these specifically rather than relying on press releases or funding announcements alone.

IPO and Financial Signals

• ChiNext IPO completion and final valuation: Whether the IPO closes at, above, or below the CNY 2.6 billion target ⁶ will signal market confidence in Leju's industrial transition story. A significant discount or withdrawal would be a negative indicator.
• Revenue mix disclosure: If IPO filings break down revenue by customer segment (research vs. industrial vs. education), this will be the most important single data point for assessing whether the industrial transition is real.
• Net loss trajectory: The CNY 69.78M net loss ⁶⁸ on CNY 258M revenue is manageable but needs to narrow as revenue scales. Widening losses at higher revenue would suggest the unit economics are not improving.
• Gross margin by product line: Not currently disclosed. If disclosed in IPO filings, KUAVO gross margin will reveal whether the $20,000 target price ³ is economically viable at scale.

Technical and Product Signals

• DOF specification clarification: Leju should resolve the 26 DOF vs. 40+ DOF conflict ²²⁴ publicly. Resolution in favour of the higher figure would be a positive capability signal; confirmation of 26 DOF would require reassessment of manipulation claims.
• KUAVO variant clarification: The height and weight discrepancy ²²²³ needs resolution. If Leju has launched a larger, heavier-duty KUAVO variant, that product's specifications and target market should be formally announced.
• Independent reliability data: Any third-party assessment — academic paper, industrial case study, or teardown — of KUAVO's operational reliability in a real deployment would be highly informative. The absence of such data after 577 units shipped is itself a signal.
• LingBot-VLA deployment status: Whether the vision-language-action model described in ²¹ is integrated into shipping KUAVO units or remains a research prototype is a key indicator of AI capability maturity.

Partnership and Commercial Signals

• Schaeffler deployment metrics: Any public statement from Schaeffler about the scope, timeline, or outcomes of its Leju collaboration ¹¹ would be significant. A Schaeffler press release citing specific operational metrics would be strong positive evidence; continued silence would suggest the pilot remains exploratory.
• Named customer announcements beyond research: Any announcement of a KUAVO deployment by a named industrial, healthcare, or commercial services customer — with specific task descriptions and volume — would represent a genuine commercial milestone.
• Robbyant data licensing revenue: Whether the Robbyant collaboration ¹³ generates disclosed revenue for Leju, and at what scale, would clarify the viability of the data economy business model.
• Alibaba and Haier engagement depth: Both are listed as partners ⁴ but the nature of these relationships is not publicly detailed. Any announcement of specific product integrations or deployments would be informative.

Competitive and Market Signals

• Unitree pricing moves: Any further reduction in Unitree's humanoid pricing below $10,000 would intensify pressure on Leju's research market and accelerate the need for industrial differentiation.
• IDC humanoid shipment rankings for 2026: Whether Leju maintains its third-place global ranking ⁶ in 2026 will indicate whether it is holding market share as the competitive field expands.
• Chinese government procurement: Any announcement of KUAVO procurement by Chinese state-owned enterprises or government agencies would signal policy-backed demand that could significantly accelerate revenue growth.

Geopolitical and Regulatory Signals

• Export control developments: Any expansion of US export controls to cover components used in KUAVO's bill of materials would affect both Leju's production costs and its international sales prospects.
• HarmonyOS robotics roadmap: Huawei's continued investment in HarmonyOS for robotics applications is a dependency for Leju's software strategy. Any signals of reduced Huawei commitment to this area would be a risk indicator.
• ChiNext regulatory environment: Changes to ChiNext listing requirements or Chinese securities regulation that affect technology company valuations would directly impact Leju's IPO prospects.

---

14Sources and Methodology

Methodology

This report was produced using a structured intelligence analysis methodology applied to a research dossier compiled on 25 June 2026. The dossier contained 32 sources across six categories: official (0), commerce (5), research (4), news (13), video (6), and community (4). The absence of official sources — meaning regulatory filings, product documentation, or company-published technical specifications directly accessible to the analyst — is noted as a limitation; all official claims were accessed through secondary reporting.

Evidence was classified into four categories throughout the report:

Conflicts between vendor claims and independent sources were adjudicated on the basis of source specificity, independence, and corroboration. Where conflicts were unresolved, both positions are reported with confidence assessments. Choreographed demonstration videos were not treated as evidence of autonomous operational capability. Partnership announcements were not treated as evidence of paid commercial contracts. Shipment figures were not treated as evidence of productive deployment.

The overall dossier confidence score of 0.82 reflects good corroboration of financial and corporate facts but thinner independent evidence on technical capabilities and commercial deployment outcomes.

Sources

¹ LEJU ROBOT Stock Price, Funding, Valuation, Revenue & Financial Statements — https://www.cbinsights.com/company/leju-robotics/financials

² Leju Robotics funding, news & analysis | Sacra — https://sacra.com/c/leju-robotics

³ How to Invest in Leju Robotics Pre-IPO Shares: Fees, Minimums, and Risks | Jarsy — https://www.jarsy