Zhijian Power / 至简动力 / Simplexity Robotics

1.1 Identity, product and operating footprint

Public reporting places Simplexity Robotics in Hangzhou and dates its establishment to July 2025. The founding group is identified as Jia Peng, Wang Kai and Wang Jiajia, all former Li Auto executives, with roles reported as CEO, chairman and COO respectively. The company describes a full-stack embodied-intelligence approach spanning models, control software and robotic hardware, with the LaST₀ foundation model, ManualVLA long-horizon task model and TwinRL real-world reinforcement-learning framework as named technology components. Its stated initial deployment path is deliberately controlled: factories, supermarkets and logistics operations precede more open environments. This is a coherent product thesis, but the public record remains mostly company- or media-reported rather than customer-verified.[CO001, CO002, CO003, CO004, CO005]

1.2 Founders, leadership and key-person dependence

The leadership team combines intelligent-driving research, technical leadership and mass-production experience at Li Auto, according to multiple reports. That background is relevant to a robotics company trying to move from model research to manufacturable systems, and the reported division of responsibilities covers technology, capital and operations. However, the retained evidence does not identify an independent board, non-founder executives, safety leadership, or governance controls. The company therefore appears founder-concentrated in the public record: the team is a strength for execution speed, while the lack of disclosed organizational depth is a diligence gap rather than evidence of a governance failure.[CO006, CO007, CO008]

1.3 Financing, valuation and execution milestones

Four independent or near-independent media accounts in the assigned pool describe five financing rounds completed in less than six months and approximately RMB 2 billion raised. One account reports a post-money valuation above US$1 billion, while the available evidence does not provide term sheets, ownership, liquidation preferences, debt, cash balance or audited revenue. The operational milestones are notable but early: the first self-developed prototype reportedly followed within 45 days of the first employee joining, two generations of embodiments were completed, and the first generation entered small-batch production and PoC validation. These milestones support an execution narrative, not commercial scale.[CO009, CO010, CO011, CO012]

1.4 Milestones and evidence boundaries

The timeline below distinguishes reported company milestones from external industry context. Simplexity's public milestones run from July 2025 formation through the 2026 funding, research and hardware-development disclosures. The wider sector is moving toward formal standards and commercial deployment, but those developments do not prove Simplexity compliance, customer adoption or production economics. Independent reporting also cautions that humanoid demonstrations may remain technology demonstrations and that reliable, safe and maintainable deployment is still a systems-engineering challenge. This is the central overview judgment: strong early sponsorship and technical ambition are visible; repeatable deployment, paid demand, safety performance and financial quality remain unverified.[CO013, CO014, CO015, CO016]

1.5 Exhibits

2.1 Market boundary and substitutes

The investable boundary should begin with closed or semi-structured commercial environments, not the entire embodied-AI or consumer-robotics category. Independent reporting says Simplexity prioritizes factories, retail stores, and logistics, while its hardware has reached small-batch production and proof-of-concept rather than disclosed scaled deployment. The practical included pool is therefore spending on mobile manipulation or humanoid hardware, perception and action software, systems integration, deployment, and ongoing support for those workflows. Fixed industrial automation, conventional AMRs that do not perform the same manipulation job, and unstructured household robotics are adjacencies or substitutes rather than automatic components of Simplexity's serviceable market. Buyers can also retain labor, redesign the workflow, or deploy task-specific machines; each substitute raises the ROI bar for a general-purpose system.[CM001, CM002, CM003, CM013]

2.2 Evidence-constrained sizing

Public estimates establish category momentum but not a clean TAM/SAM/SOM stack. IDC figures relayed by DirectIndustry put 2025 global humanoid volume at about 18,000 units and hardware revenue at approximately US$440 million. Broader estimates cited by the same publication include software and services and range from nearly US$3 billion currently to US$15 billion by 2030, while a separate forecast reaches US$35.4 billion by 2033. A China-specific forecast exceeds RMB20 billion in 2026, and TrendForce expects Chinese output to rise 94% in 2026. These values differ by geography, category boundary, denominator, and forecast year. They are useful sensitivity markers, but none isolates the workflows Simplexity can serve, the portion buyers will procure, or the share the company can win.[CM004, CM005, CM006, CM007, CM008, CM009]

2.3 Buyers, budget ownership, and adoption path

Industrial and logistics deployments provide the clearest adoption path because operations leaders can compare a robot with labor, fixed automation, and existing mobile robots on throughput, uptime, integration cost, and safety. Peer evidence suggests real demand exists: a Geekplus-issued release reports roughly 950 end customers and a 78% large-customer repurchase rate, while an independent Chinese market roundup describes industrial use as the principal deployment front in 2026. That evidence validates buyer interest in warehouse and factory automation, not Simplexity's own commercial traction. For Simplexity, the likely sequence is a technical proof-of-concept, site integration and safety acceptance, a paid pilot, measured operating performance, and only then repeat-site or fleet rollout. Public evidence does not identify its payer, budget owner, pilot pricing, or conversion rate.[CM001, CM002, CM015, CM016]

2.4 Drivers, constraints, and valuation relevance

Supply-side momentum is substantial: TrendForce expects 94% Chinese output growth in 2026 and projects Unitree and AgiBot to capture nearly 80% of shipments. That concentration is simultaneously a demand signal and a warning that scale benefits may accrue to a small number of manufacturers. Adoption still depends on moving beyond demonstrations. Independent experts cited by DirectIndustry describe many humanoids as mid-readiness systems requiring material work on reliability, safety, maintainability, and integration, and note that task-specific robots can remain more practical. China's 2026 standards framework adds lifecycle, application, data, safety, and ethics requirements across six areas. These constraints make paid deployments, uptime, intervention rates, service cost, certification, and repeat orders more valuation-relevant than top-down category forecasts or financing volume.[CM009, CM010, CM011, CM012, CM013, CM014]

2.5 Exhibits

3.1 Competitive landscape and buyer alternatives

The relevant competitive set is broader than humanoid start-ups alone. AgiBot and Unitree are visible scale and brand leaders in China, while UBTech emphasizes industrial humanoids and coordinated factory work. Galaxy General, 智平方, and Zhijian Power represent full-stack, model-led approaches with industrial ambitions. Leju addresses education and research as well as full-height platforms, and XPeng brings an automotive manufacturing and distribution option. Customers can also buy application-specific robots, retain conventional automation, or build integrations internally. That creates a substitution boundary: a general-purpose robot must justify flexibility against the reliability, throughput, and task economics of a dedicated machine.[CP002, CP003]

3.2 Peer profiles and differentiation

Publicly reported peer evidence shows several distinct routes to differentiation. AgiBot's reported 2025 shipment volume and deployments in factories and service locations indicate a scale-and-scenario route. Unitree combines consumer visibility, international distribution, and reported profitability, although shipment estimates differ between independent reporting and the CEO's claim. UBTech's Walker line is positioned around automotive and logistics workflows, battery swapping, and multi-robot coordination. Galaxy General emphasizes a large industrial payload and retail footprint; 智平方 emphasizes a self-developed model and a large order; Leju remains more education and research oriented. These are materially different wedges rather than a single homogeneous peer group.[CP005, CP006, CP007, CP008, CP009, CP010]

3.3 Buying criteria, pricing, and switching

The available pool does not provide comparable realized prices or contract terms for Zhijian Power and its peers. One public reference places Unitree's G1 starting price at $13,500, but that is a list price for a specific model and is not evidence of deployed-system economics. The better-supported buying criteria are task fit, reliability, integration with factory systems, operator dependence, supply access, and serviceability. A buyer may prefer application-specific equipment when a general-purpose platform cannot match its efficiency. Switching costs should rise with accumulated task data, workflow integration, custom end-effectors, and validated safety procedures, but no public source in this pool quantifies Zhijian Power's lock-in or renewal rates.[CP011, CP012, CP013, CP014]

3.4 Moat durability and displacement risk

The industry's strongest apparent moats are manufacturing scale, access to real operating data, component and supply-chain control, and trusted deployment channels. They remain contestable. TrendForce describes a shift toward tangible user value and reports rapid scaling by leading vendors, while independent commentary warns that many demonstrations remain technology showcases rather than reliable industrial systems. Public reporting also identifies overcapacity and bubble risk as production expands faster than orders. For Zhijian Power, former Li Auto leadership and a world-model-plus-hardware thesis may aid recruiting and technical execution, but the pool does not prove that those assets translate into lower cost, safer operation, or durable customer preference. Incumbent response and dedicated-robot substitution therefore remain material risks.[CP015, CP016, CP017, CP018]

3.5 Exhibits

4.1 Revenue model and monetization visibility

Simplexity is described as building embodied-intelligence products for factories, retail, logistics, and other commercial environments, with two generations of hardware, small-batch production, and proof-of-concept validation reported by Gasgoo and 36Kr. That evidence supports a hardware-plus-software commercialization hypothesis, not recognized revenue. The reviewed sources do not disclose a price list, contract structure, implementation fee, recurring software charge, usage fee, shipment revenue, backlog, or revenue-recognition policy. A company statement that funding will support model training, hardware R&D, data acquisition, and algorithm development indicates a development-heavy cost base, but it does not establish customer-funded economics or gross-profit contribution. List pricing, realized pricing, and any pilot-to-production conversion remain unverified.[CI001, CI002, CI003, CI004]

4.2 Cost structure and unit-economics proxies

The disclosed operating model is capital intensive by design. 36Kr reports that the company chose on-device computing and a general-purpose body to collect, train, test, validate, and deploy data in the user's setting, while also noting that the choice carries a higher per-machine cost. The same source reports that the company develops and manufactures its stack in-house; Gasgoo says the funding is intended for model training, hardware R&D, data acquisition, and algorithm development. These are useful cost-driver signals, but neither source supplies bill of materials, labor content, cloud or edge-compute cost, warranty reserve, service burden, utilization, gross margin, or contribution margin. Workplace robotics also carries safety-engineering and maintenance exposure: OSHA notes that many accidents occur during programming, maintenance, testing, setup, or adjustment. The proper underwriting posture is to model these as diligence inputs, not to convert public descriptions into invented unit economics.[CI005, CI006, CI007, CI008]

4.3 Capital adequacy and financial verdict

The strongest financial signal is financing velocity. Gasgoo reports five rounds in six months and RMB 2 billion of cumulative funding; 36Kr reports that the final round's post-money valuation exceeded US$1 billion, while 10jqka likewise reports five rounds and RMB 2 billion. These figures are media-reported and are not accompanied in the reviewed pool by term sheets, a cap table, cash-on-hand disclosure, debt schedule, burn rate, or runway. The financing can fund a long product-development cycle, but funding raised is not cash available after fees, prior spending, restrictions, or investor-specific terms. The adverse case is also material: Business Times reports an NDRC warning about a bubble risk, more than 150 humanoid-robot makers in China, and limited widespread household or factory adoption at the time of its report. SESEC documents an emerging standards and safety framework, while OSHA identifies workplace hazards and the absence of specific OSHA robotics standards. The verdict is research-more: capital access is a strength, while revenue quality, margin path, capital intensity, and next-round dependence remain unresolved.[CI009, CI010, CI011, CI012, CI013, CI014]

4.4 Exhibits

5.1 Product definition and customer workflow

The public product description places Zhijian Power at the intersection of a robotics foundation model and robot bodies intended for industrial and service settings. A March 2026 Sina Finance report says the company planned to work on base-model training, robot-body development and iteration, data collection, and core algorithms, with expansion from factories into manufacturing and service scenarios. The same report describes LaST0 as combining a world model's physical prediction with a VLA fast/slow system. This gives a coherent workflow hypothesis: observe a physical scene, infer a temporally extended state, generate an action, collect embodied data, and improve the model. It does not establish a sellable SKU, a named customer workflow, a paid deployment, or a measured production benefit. The LaST0 project page lists demonstrations such as placing a dish on a rack, wiping a whiteboard, scooping popcorn into a bowl, and opening a pot to pick corn; these are useful task examples, not evidence that Zhijian Power has commercialized them.[CE001, CE002, CE003, CE004]

5.2 Architecture, operating model, and maturity

The LaST0 paper provides the clearest public architecture description associated with the company's reported technical direction. It describes a latent spatio-temporal chain of thought that models future visual dynamics, 3D structure, and robot proprioception, with a low-frequency reasoning expert and a high-frequency acting expert implemented through a Mixture-of-Transformers design. The paper reports evaluation across 10 real-world tasks and states that mean success rates improved by 13%, 14%, and 14% over prior methods across tabletop, mobile, and dexterous manipulation categories. Those are paper-reported results, not a company deployment guarantee. The paper's project page lists demonstrations but no public API, integration specification, uptime record, manufacturing bill of materials, or service-level commitment. A March 2026 company profile additionally describes an intended data loop using on-device collection and human demonstrations, but it does not provide dataset size, labeling protocol, sim-to-real validation, or reproducibility artifacts. The resulting maturity assessment is research-prototype to early productization, not proven production infrastructure.[CE005, CE006, CE007, CE008, CE009]

5.3 Trust, safety, security, and implementation risk

The available evidence supports a real technical risk surface but not a completed control framework. China’s 2026 standards system for humanoid robots and embodied intelligence covers data lifecycle, model training, inference, deployment, applications, safety, and ethics, while its published work includes embodied-model system requirements, data-quality specifications, trustworthiness requirements, evaluation guidelines, and full-lifecycle management. Separately, the translated Personal Information Protection Law requires security measures such as internal operating procedures, categorical management, encryption or de-identification, access control, training, incident plans, and periodic compliance audits for personal-information handling. These external requirements are not evidence that Zhijian Power complies with them. They define diligence gates for any robot system collecting images, voice, human demonstrations, or workplace telemetry. Public materials also do not verify collision testing, fail-safe behavior, cybersecurity architecture, incident reporting, component traceability, maintenance procedures, or deployment acceptance criteria. These gaps are especially important because the company narrative emphasizes real-world data collection and expansion from closed to more open environments.[CE010, CE011, CE012, CE013]

5.4 Exhibits

6.1 客户与使用场景：方向明确，买方尚未可核验

公开报道把至简动力的目标场景描述为工厂和物流，但没有把场景描述连接到具体买方、地点、机器人型号、部署数量或合同状态。对工业机器人而言，使用者、付款者和决策者可能分别是产线运营、工厂信息化团队、系统集成商与集团采购；本章没有证据证明这些角色已经在至简动力的实际项目中被拆分或签约。因而，现阶段可以把工厂和物流视为公司声称的客户工作流方向，而不能把它们计为已验证客户细分。行业材料显示，具身智能公司正在向工业制造、物流等场景扩展，但同业案例不构成至简动力的客户证明。投资判断应要求逐项目提供客户身份、部署地点、生产/试点状态、验收标准与付款凭证。 The available evidence identifies a target workflow, not a verified account. Buyer identity, user identity, payer identity, site identity, and procurement authority remain separate unknowns. A diligence-ready segment needs a named organization, a documented use case, an operational location, a contract or purchase order, and a status such as pilot or production. None of those fields is supplied for Simplexity in this pool. The sector comparison therefore provides context only; it cannot be used as customer attribution. This distinction prevents logos, partnership language, or broad application claims from being counted as recurring commercial adoption.[CU001, CU002, CU003, CU004]

6.2 采用轨迹与留存：融资进展不等于商业采用

本次分配的直接公司证据仅支持一项融资与定位叙述：报道指称公司在成立不到一年内连续完成五轮融资、累计约20亿元，并主攻工厂和物流。该信息不能替代收入、客户数、活跃设备、部署地点、利用率、复购、续约或净收入留存。技术论文列出与Simplexity Robotics关联的研究作者和现实机器人任务结果，但论文实验结果属于技术信号，不是客户部署或商业留存证据。行业报道中其他公司的订单、客户数和交付量也不能外推至至简动力。故本章将采用轨迹标为未证实，并把客户合同、验收和重复购买数据列为材料级缺口。 Funding momentum is a financing signal, not a customer funnel. A credible adoption trajectory would show leads, pilots, conversions, active sites, utilization, paid volume, service burden, renewals, and expansion. The cited robotics paper improves the technical signal because it links the company to a research project and reports real-world tasks, but laboratory or benchmark success does not establish a paying deployment. The report therefore separates capital access, engineering evidence, and customer evidence. Missing denominators are material: a count of projects without paid status or time period cannot reveal growth, retention, or revenue quality. The current record supports interest in the company, but not durable commercial usage.[CU005, CU006, CU007, CU008]

6.3 命名客户、扩张与集中度：尚无公开证明

在本次成功预取且分配给本章的来源中，没有一份由客户、采购方、监管机构或独立评测者直接确认至简动力生产部署或采购结果的材料。至简动力的公开报道没有披露客户名称、试点到量产的转化率、单一客户收入占比、渠道伙伴、售后服务水平或停机/安全事件。与此同时，国家标准框架已经把应用开发、运行维护、安全与伦理纳入具身智能的全生命周期要求；这提高了未来客户验收和合规的证据门槛，而不是证明公司已经满足这些要求。对于工厂和物流客户，集中度、采购周期、系统集成依赖和现场安全责任都应在数据室中逐项核实。 Named customer proof is the missing bridge between positioning and underwriting. The absence of a public customer name is not proof that no customer exists; it is an evidence limitation. The same limitation applies to concentration: without a revenue denominator, top-customer share and channel dependence cannot be calculated. Industrial buyers also require safety, maintenance, cybersecurity, and acceptance documentation before production use. Public standards describe those diligence surfaces, but they do not certify Simplexity. A buyer-quality data room should include references that can be independently contacted, signed commercial terms, acceptance records, incident logs, service-level commitments, and a cohort view of renewals. Until then, the expansion thesis remains an open question rather than a measured outcome.[CU009, CU010, CU011, CU012]

6.4 Exhibits

7.1 Regulatory and compliance exposure

The regulatory risk is forward-looking rather than tied to a disclosed enforcement event. China’s 2026 standards framework spans the full lifecycle of humanoid robots and embodied intelligence, including model training, inference, deployment, applications, safety, and ethics. That breadth increases the surface area for testing, data-governance, identity, and operational-compliance work as products move from demonstrations into factories. The available official-adjacent standards account says more than 120 institutions participated and that an initial list of 52 standards was published, while also listing several items as drafting, pre-research, or under approval. This indicates an evolving rulebook rather than a settled compliance baseline. The company-specific diligence question is whether its data-collection, teleoperation, model deployment, and industrial-safety controls are already mapped to applicable requirements. No assigned source confirms a Zhijian Power license, certification, breach, recall, or enforcement action; that absence is an evidence gap, not proof of clean status.[CR001, CR002, CR003, CR004, CR011]

7.2 Commercialization and operating execution

The most immediate operating risk is the gap between demonstrated capability and repeatable, safe production economics. A February 2026 local report describes a first industrial dual-arm robot with adaptive gripper and teleoperation data gloves as having entered sample testing and productization, and says the company is working to solve scale-up in real physical settings. That is useful evidence of a development milestone, but it is not evidence of paid production deployment, uptime, service cost, acceptance testing, or repeat orders. The same report frames the Suzhou innovation center and the Green Harmonic partnership as a route to connect research, manufacturing, and application; it does not disclose binding volumes or service-level obligations. Competitor reporting shows how quickly the category is moving toward large financings and orders, which raises execution and timing pressure without proving that Zhijian Power has comparable traction. Underwriting should therefore treat reliability, safety, installation time, maintenance burden, and customer acceptance as unresolved operating variables.[CR005, CR006, CR007, CR008]

7.3 Partner, capital, and thesis-break risks

The company’s disclosed route to commercialization appears dependent on an ecosystem of precision-component suppliers, local industrial infrastructure, and continued private financing. The reported Green Harmonic relationship is strategically logical because it combines general-model and application capability with harmonic-reducer manufacturing, but the public account does not establish exclusivity, pricing, capacity reservations, quality guarantees, or a second source. Category financing reports also describe a market where capital is concentrating in a small number of leaders and where peers are announcing very large rounds, deployments, and orders. That creates a financing and competitive risk: a young company may need to fund hardware iteration, data collection, field support, and compliance before revenue quality is visible. Public sources do not disclose Zhijian Power’s revenue, gross margin, burn, runway, debt, cap-table preferences, or customer concentration. A thesis-break should be tied to measurable failure to secure qualified production deployments, a material supplier interruption without a qualified substitute, or financing terms that materially dilute control before commercial proof.[CR009, CR010, CR011, CR012]

7.4 Exhibits

8.1 Evidence and investment call

Simplexity Robotics sits in a market where commercialization expectations have moved materially upward, but the available evidence is market-level rather than company-level. CNBC reports that Morgan Stanley expects 50,000 humanoid robots to ship in China in 2026, a $2 billion market this year, and a $15 billion market by 2030; the forecast excludes prototypes, pre-order trials, and internal-use units. That distinction matters: it separates external sales from demonstrations, but it still does not establish Simplexity's paid deployments, revenue, margin, or repeat orders. Public sector coverage also shows that comparable companies are raising large rounds, reporting revenue, or approaching public markets, yet those are comparables, not substitutes for a company-specific cap table. The investment call is therefore research-more, with high risk and an indeterminate valuation stance until the entry price and financing terms are verified. A buy decision would require named customers, audited or management revenue, financing documents, preference terms, and evidence that deployments convert into recurring commercial revenue.[CV001, CV002, CV003, CV004, CV005]

8.2 Thesis, anti-thesis, and price discipline

The positive thesis is that sector adoption is moving from demonstrations toward factory, logistics, and retail use cases, while standards activity is becoming more formal. CNBC cites factory, logistics, unmanned-retail, and interactive-service rollouts in Morgan Stanley's field research; SESEC describes a 2026 Chinese standards system spanning foundational standards, intelligent computing, complete systems and integration, applications, and safety and ethics. The anti-thesis is that market forecasts can capitalize anticipated adoption before unit economics are proven, and safety standards do not prove a vendor's compliance or commercial success. ISO 10218-1 covers the robot as a machine and explicitly points to ISO 10218-2 for integration; it also excludes several use cases, including mobile platforms and public-access service robots. For Simplexity, price discipline should therefore be milestone-based: pay only against verified paid deployments, gross-margin evidence, safety/compliance ownership, and financing terms that do not create hidden preference or dilution overhang.[CV006, CV007, CV008, CV009]

8.3 Scenarios, exit readiness, and diligence

A scenario framework is more defensible than a point valuation because the assigned evidence does not disclose Simplexity's revenue base, cash burn, valuation, or financing terms. The bull case assumes the company converts technical capability into paid industrial deployments and benefits from the expanding external-sales market; the base case assumes pilots and partnerships remain difficult to translate into repeatable revenue; the bear case assumes commercialization lags the market narrative, while safety, integration, or funding requirements consume capital. Public comparable coverage is directionally useful: a sector article reports Galaxy General's claimed $151 million round and more than $330 million of total investment, while another report describes other Chinese robotics companies' disclosed revenue, margins, orders, and funding. These figures are not directly comparable to Simplexity because issuer, accounting basis, product mix, and stage differ. Exit readiness is not demonstrated. The next diligence gate is a data room tying price to a dated cap table, preference waterfall, customer contracts, deployment economics, safety evidence, and a cash runway model.[CV010, CV011, CV012, CV013]

8.4 Exhibits

The exhibits preserve the distinction between market evidence, peer disclosures, and unresolved company-specific valuation inputs. No precise target return or intrinsic-value point estimate is presented because the retained sources do not provide a defensible Simplexity revenue, margin, cash, or cap-table denominator. This restraint is intentional: a peer financing headline can inform diligence questions, but it cannot establish a comparable multiple without matching stage, product mix, accounting basis, ownership, and preference terms. The reader should treat every scenario as a gate for evidence collection, not as a forecast.[CV004, CV010]

8.5 Exhibits