Sudu Technology

1.1 Identity, scope, and operating stage

Current reporting identifies Sudu Technology as a Chinese general-purpose embodied-AI company, while an S&P Capital IQ item names the legal entity as Shanghai Sudu Technology Co., Ltd. The October financing report says proceeds are intended for full-stack technology development and large-scale validation at industrial production lines and workstations. That evidence supports describing Sudu as moving from laboratory prototypes toward industrial validation, not as a proven commercial-scale operator. The assigned sources do not establish a founding date, headquarters address, product pricing, recurring business model, or production deployment count. Those fields remain explicit gaps rather than facts inferred from the company's financing profile.[CO001, CO002, CO004, CO009]

1.2 Financing record and stakeholder map

RobotHot reports an October 2026 round of more than 1 billion yuan, translated by that source as $140 million. A separate S&P Capital IQ item is titled as a $500 million financing and records convertible preferred shares plus participation by strategic and financial investors. Its fetched body contains malformed dollar formatting, so the title amount should be confirmed from primary closing documents. The two records must not be added together: the evidence does not reconcile whether they are separate rounds, overlapping closes, or differently described transactions. It also does not disclose ownership percentages, board rights, debt, secondary proceeds, or cash available after either event.[CO003, CO005, CO006, CO007]

1.3 Leadership and governance visibility

Neither retained Sudu-specific financing text identifies founders, executives, directors, or the board structure. This is a disclosure limitation, not evidence that the roles are absent or that leadership has changed. It prevents assessment of founder-market fit, technical key-person dependence, capital-allocation authority, and succession depth at a stage when the company says it is expanding real-world industrial validation. Named investors provide a starting point for governance diligence, but investment participation does not prove current ownership or control. Management should provide a signed leadership roster, biographies, the current capitalization table, board composition, reserved matters, and retention arrangements before governance quality is scored.[CO006, CO008]

1.4 Milestones, capital intensity, and validation risk

The retained evidence supports only two dated Sudu financing milestones, so the timeline also labels sector events as context rather than company achievements. Category financing is substantial: an independent tracker reports $13.8 billion of global robotics startup funding in 2025, while Tether announced up to $1.4 billion for NEURA Robotics and TechNode reported Unitree's planned 4.202 billion yuan IPO raise. Capital availability does not validate Sudu's economics or technical performance. IFR-linked reporting cautions that real-world humanoid production capabilities remain largely demonstrators or pilots, and independent Unitree analysis shows how public filings can overturn earlier private-company narratives. Sudu therefore needs primary operating evidence, not valuation-by-comparable.[CO010, CO011, CO012, CO013, CO014, CO015, CO016]

1.5 Exhibits

2.1 Market boundary and adoption context

The relevant market is the software, robot systems, and deployment services required to make physical work adaptable across tasks and environments. It includes embodied-intelligence models, robot bodies, sensing and control systems, simulation and training infrastructure, integration, and recurring support for industrial or service deployments. It excludes conventional fixed industrial automation that performs a single repetitive task without a general physical-intelligence layer, as well as speculative household demand that has not reached a repeatable purchasing workflow. Sudu's available descriptions position it as a general-purpose embodied intelligent robot-platform developer, while its October 2026 financing report says its proceeds will support scaled verification in real industrial scenarios. That wording makes industrial deployment evidence more decision-relevant than a broad humanoid-robot headline market.[CM001, CM002, CM003, CM004]

2.2 Multiple sizing lenses and comparability limits

The evidence supports several non-equivalent lenses rather than one company-specific TAM. A sector tracker reports $13.8 billion of global robotics-startup funding in 2025 and cites a $38 billion 2035 humanoid robotics projection from Goldman Sachs, but those figures represent capital formation and an external category forecast, not Sudu revenue or an addressable serviceable market. RoboticsTomorrow, summarizing IFR material, reports that China accounted for 54% of annual industrial-robot installations and says near- and medium-term universal humanoid adoption is unlikely; industrial automation therefore remains a substitute and timing constraint. Unitree's IPO disclosure provides an operating-market reference point, with 2025 revenue of 1.699 billion yuan and more than 5,500 humanoid robots shipped. A separate Unitree analysis says 73.6% of humanoid revenue came from research and education and roughly 9% from industrial deployment, illustrating that shipped units do not yet prove factory-scale adoption. Unitree is a competitor, so its figures cannot be transferred to Sudu. A fetched S&P Capital IQ item also renders a reported Sudu funding amount with malformed comma placement, making it unusable as a sizing input. The available sources do not isolate Sudu's SAM, SOM, pricing, deployment count, or customer budget, so those values remain diligence gaps.[CM005, CM006, CM007, CM008, CM009, CM015, CM016]

2.3 Buyer segmentation, drivers, and constraints

The most supportable initial buyer path is enterprise and industrial: factories, logistics operators, and other sites where a general physical intelligence layer can be verified against a defined workflow. The buyer, user, and payer may differ: operations or plant leadership owns the productivity problem, engineering and safety teams validate the system, and a corporate capital or automation budget funds deployment. The October 2026 financing account explicitly names real industrial scenarios and scaled verification, while the China policy coverage describes demonstrators and pilot projects as the current state of many humanoid deployments. This implies a staged adoption path from pilot to repeatable multi-site deployment, with ROI, safety, integration, data control, and reliability as gating criteria. Switching from established industrial robots is constrained by precision and speed requirements; switching from manual work is constrained by trust, training, and operational risk. Sudu's public evidence does not identify a named paying customer, contract value, conversion rate, or budget owner, so the segment map is a diligence hypothesis anchored to the reported industrial verification use case.[CM010, CM011, CM012, CM013, CM014]

2.4 Exhibits

3.1 Competitive landscape and strategic positions

The relevant landscape is broader than humanoid robot manufacturers. Sudu's stated proposition is a general-purpose robot brain that generalizes perception and manipulation skills through standardized model capabilities and APIs, so the closest direct peers include software-layer foundation-model companies as well as integrated robot vendors. ChinaBulls describes Sudu as pursuing “a ‘robot brain’ with strong generalization capabilities” and moving from locomotion intelligence toward perceptual and interactive intelligence. Against that position, Skild AI and Physical Intelligence emphasize hardware-agnostic intelligence, while NEURA, Apptronik and Unitree pair software with robot bodies or deployment programs. Agility and internal automation builds are substitutes for particular warehouse or factory jobs, and conventional fixed automation remains the status quo where tasks are structured. Funding and shipment figures indicate that alternatives are attracting substantial capital, but do not by themselves establish equivalent technical capability or customer value.[CP001, CP002, CP003, CP004]

3.2 Buyer-facing capability, distribution, and pricing comparison

Buyer choice depends on whether the requirement is a robot body, a reusable intelligence layer, or a production-ready workflow. Unitree's reported 2025 revenue of 1.699 billion yuan, shipment of more than 5,500 humanoid robots, and planned 4.202 billion yuan IPO provide stronger public scale signals than are available for Sudu, although the metrics are not directly comparable to a software platform. Apptronik publicly identifies manufacturing and logistics as initial use cases and names Mercedes-Benz, GXO Logistics, Jabil, and Google DeepMind partnerships; NEURA's Tether announcement describes a multi-form-factor portfolio, enterprise-client pipeline, and the Neuraverse software ecosystem. Physical Intelligence has raised more than $1 billion but, according to TechCrunch, had no commercialization timeline at the cited date. Public list pricing is sparse: the assigned sources provide no directly comparable Sudu, Skild, Physical Intelligence, NEURA, Apptronik, or Unitree price per task, API call, robot, or contract. Unknown pricing should therefore be treated as a diligence gap rather than a competitive advantage.[CP005, CP006, CP007, CP008, CP009]

3.3 Switching costs, moat durability, and displacement risk

The most durable potential moat is not funding alone but a combination of reusable skills, deployment data, integration interfaces, partner access, and evidence that simulation-trained policies transfer reliably to unfamiliar physical environments. Sudu's public positioning emphasizes standardized model capabilities and APIs, which could reduce customer switching friction if interfaces are portable, but could also make the intelligence layer easier to substitute if competing models achieve similar performance. Integrated vendors can create lock-in through hardware, service, and workflow integration: NEURA describes a platform connecting robots, components, models, data, compute, and services, while Apptronik is using strategic partnerships to scale Apollo deployments. Unitree's public revenue, profit, and shipment disclosure illustrates a different advantage: operating scale and a pathway to public-market financing. Adverse evidence is concentrated in execution uncertainty: Physical Intelligence had no commercialization timeline in the cited TechCrunch report, and public sources do not establish Sudu's production-scale sim-to-real performance, customer retention, or realized pricing. These gaps keep moat durability unproven.[CP010, CP011, CP012, CP013, CP014]

3.4 Competitive implications for underwriting

Sudu appears to compete for the intelligence layer of embodied robotics while also facing full-stack vendors that can bundle hardware, deployment, financing, and support. The company may differentiate if its simulation-first approach produces reliable generalization, low data-collection cost, and APIs that work across robot bodies; the assigned evidence does not independently verify those outcomes. The strongest observable competitive signals are therefore proxies: competitor capital, disclosed deployments, shipment and financial scale, ecosystem partnerships, and commercialization timing. A buyer or investor should test whether Sudu can win against a software-only alternative on task success and integration time, and against a vertically integrated vendor on total delivered cost and accountability. The current record supports tracking the competitive field, not assuming a durable moat or treating competitor funding as proof that Sudu's valuation is justified.[CP003, CP006, CP010, CP013, CP014]

3.5 Exhibits

4.1 Revenue model and observable traction

The retained evidence does not disclose Sudu Technology revenue, ARR, GMV, recognized units, utilization, realized pricing, gross margin, or customer concentration. The May 2026 S&P Capital IQ record identifies a financing event and says the company issued convertible preferred shares, but its amount is rendered as "$500,0000,000" and cannot be normalized without another source. The Chinese announcement page in the pool is not usable for a transaction or operating claim because its fetched body contains only the text “看横店” and “不喜欢,” rather than an auditable disclosure. Accordingly, the public record cannot classify Sudu as pre-revenue, early-commercial, or scaled; absence of a disclosed metric is not evidence that its value is zero. The commercial mechanism is also not public enough to classify confidently as SaaS, robotics-as-a-service, hardware sales, licensing, or a blended model. For a simulation- trained robot-brain company, revenue quality will depend on whether customers pay for robot deployments, software access, integration, data, or services, and whether those receipts recur. The current record supports a diligence request for contracts, invoices, backlog, deferred revenue, shipment records, and a revenue-recognition policy before any run-rate or growth multiple is used.[CI001, CI002, CI003, CI004, CI015]

4.2 Cost structure and unit economics

No retained source provides Sudu Technology's bill of materials, compute spend, simulation cost per useful training hour, data-collection expense, manufacturing commitments, support burden, gross margin, CAC, sales cycle, payback, or working-capital profile. The financing structure therefore matters: a convertible preferred round can fund engineering and manufacturing scale, but even a verified amount raised would not equal cash currently available after fees, prior obligations, or planned deployment. Comparable sector evidence shows why operating proof is material. An industry funding overview describes Figure's use of capital for AI training, robot manufacturing, engineering headcount, and commercial deployment; BMW's report on Figure 02 documents a production trial but does not establish Sudu's economics. Unitree's later public financial disclosure provides a benchmark for what useful hardware underwriting can eventually look like—revenue, profit, mix, gross margin, and customer composition—but those secondary figures belong to Unitree and are peer context rather than an independent benchmark for Sudu. Sudu should therefore be underwritten on measured unit economics only after management supplies company-specific denominators.[CI005, CI006, CI007, CI008, CI016]

4.3 Capital adequacy and underwriting verdict

The strongest current financial fact is the reported May 2026 financing event, with GL Ventures, BlueRun, Tencent's investment arm, Ant Group, CATL, Alibaba, and IDG Capital listed as participants. The source says the securities were convertible preferred shares, but its amount field is malformed and is not treated here as a verified $500 million figure. The record does not disclose post-money valuation, cash balance, burn rate, runway, debt, liquidation preferences, use of proceeds, or the next-round trigger. Financing dependency therefore cannot be quantified: without a cash-flow statement and deployment plan, the round cannot be translated into months of runway. Sector capital is abundant but selective. The retained 2026 overview reports $13.8 billion raised by robotics startups globally in 2025 and describes capital concentrating in companies with deployments, corporate partnerships, and defensible software. Those are market-context observations, not Sudu metrics. The financial verdict is research-more: the financing event may provide capacity to build, but the unverified amount and absence of revenue, margin, cash, and contract evidence prevent a credible return or solvency assessment. Priority diligence is a reconciled cash bridge from the financing close to current cash, monthly burn by function, committed capex, debt and preferred-share terms, and signed commercial obligations.[CI009, CI010, CI011, CI012, CI013, CI014]

4.4 Evidence gaps that block underwriting

The open questions are not cosmetic. Revenue recognition cannot be assessed without contracts and delivery evidence; unit economics cannot be assessed without the unit, customer, and period denominators; and capital adequacy cannot be assessed without current cash, burn, liabilities, and financing terms. A financing headline can coexist with rapid cash consumption during model training, robot prototyping, facility buildout, or customer integration. Until those records are produced, a valuation comparison to better-disclosed robotics peers would conflate capital raised with enterprise value and conflate market enthusiasm with operating performance.[CI004, CI006, CI010, CI014]

4.5 Exhibits

5.1 Product definition and operating workflow

Sudu Technology describes Sudo R1 as a general-purpose robot brain rather than a narrow application: the product is intended to connect perception, reasoning, and physical action for manipulation in human-built environments. Humanoid.Guide independently describes the model as linking vision, language, and action and as using simulation to reduce real-world training requirements, while an Origin of Bots account describes a workflow in which human motion or task intent is processed with visual and spatial inputs before joint actuation and balance correction. A Chinese industry article reports the company is targeting industrial manufacturing and logistics, including joint development with CATL, but that account is company-linked reporting rather than independent deployment evidence. The practical buyer workflow is therefore best treated as an assisted-manipulation or automation platform whose value depends on repeatable task transfer, not as proof of broad autonomy.[CE001, CE002, CE003, CE004]

5.2 Architecture, training, and maturity

The disclosed technical thesis is a simulation-first stack combining a 3D world model, reinforcement learning, a manipulation policy, and closed-loop control. The Chinese report attributes near-100% zero-shot success on unspecified key tasks to a no-real-robot-data setup and separately describes a 60-minute unedited test involving unfamiliar object shapes, lighting, and backgrounds. Those are reported company or company-adjacent claims, not a reproducible benchmark: the public record does not state the test population, denominator, failure taxonomy, robot hardware, evaluator protocol, or sim-to-real holdout design. Origin of Bots reports a 15–25 Hz closed-loop policy but also says that robot size, weight, and hardware specifications were not published. The architecture is consequently legible at the conceptual layer while its engineering maturity and generalization claims remain unverified.[CE005, CE006, CE007, CE008]

5.3 Differentiation and deployment path

Sudu's proposed differentiation is the ability to scale policy learning through simulation rather than relying primarily on expensive teleoperation and real-robot data. That could lower marginal data-collection cost and make a common policy more portable across workcells, but the public evidence does not establish that the claimed advantage survives changes in robot embodiment, sensors, latency, contact dynamics, or factory safety constraints. The company-linked account says the system is being developed for multiple manufacturing and logistics workstations and that a developer center and toolchain are being built. These statements show a roadmap and intended ecosystem, not measured production uptime, integration time, customer conversion, or support burden. MarketResearchFuture provides sector context that AI, computer vision, machine learning, sensors, and actuators are competitive layers in China's humanoid market; it does not validate Sudu's implementation.[CE009, CE010, CE011, CE012]

5.4 Trust, reliability, and diligence blockers

The available sources do not document a security model, privacy impact assessment, safety case, certification, incident history, uptime commitment, software update process, or formal quality system for Sudo R1. The simulation-first approach may reduce the need to collect sensitive customer demonstrations, as claimed in the Chinese report, but it does not by itself demonstrate that customer data, teleoperation channels, model artifacts, or robot control interfaces are protected. Nor does a successful manipulation demo establish safe behavior under sensor degradation, novel contact forces, network interruption, or adversarial inputs. IDG Capital's own site warns about impersonation and says official information is released only through its official website; that notice is not evidence about Sudu's product, but it reinforces the need to verify partner and investor claims through primary records. The diligence priority is a reproducible benchmark and deployment evidence with safety, security, and reliability controls.[CE013, CE014, CE015, CE016]

5.5 Exhibits

6.1 Customer segments and use cases

The strongest customer-facing evidence is a company-linked report describing Sudo R1 for industrial sorting and warehouse depalletizing. It says the system can adapt across transparent, flexible, reflective, and other unknown objects and that deployment does not require collecting sensitive production data. Those claims point to industrial operators whose buying criteria include rapid integration, data governance, and tolerance for unstructured objects, rather than to a general consumer product. The same report names sorting and palletizing as commercial scenarios, but it does not identify a paying customer, contract, site, purchase order, or production acceptance test. Broader sector evidence places initial humanoid adoption in factories, logistics centers, and other semi-structured environments; this is a market context, not proof of Sudu adoption.[CU001, CU002, CU003, CU004, CU013]

6.2 Adoption proof and trajectory

Sudu's public adoption proof is a product demonstration rather than a customer deployment record. The reported test ran continuously for 60 minutes and involved grasping multiple classes of unknown objects, with a claimed near-100% zero-shot success rate. That is relevant to technical feasibility and a prospective buyer's evaluation, but it lacks the denominator definition, task-by-task log, independent witness, site name, paid-versus-pilot status, utilization history, and post-deployment outcome needed to infer active accounts or repeat usage. The adjacent robotics market has clearer public examples of factory and warehouse deployments, which raises the evidentiary bar for Sudu: a demo should not be presented as production traction. No source in this assigned pool discloses Sudu customer count, revenue by customer, renewal, churn, NRR, GRR, or contract duration.[CU005, CU006, CU007, CU008, CU014]

6.3 Expansion, procurement, and concentration

The proposed expansion path is from a first industrial manipulation workflow into adjacent sorting, warehousing, and palletizing tasks. A simulation-first approach could reduce the need for sensitive customer data during onboarding, but the evidence does not establish implementation time, integration cost, safety acceptance, service levels, or whether one buyer can expand across sites. Funding and investor disclosures show capital participation, not customer concentration or commercial durability. The practical diligence priority is therefore a customer-level evidence pack: signed referenceable accounts, production-versus-pilot status, paid deployment dates, task-level uptime and success logs, expansion economics, and the share of revenue attributable to the largest account or channel.[CU009, CU010, CU011, CU012, CU014, CU015]

6.4 Exhibits

7.1 Regulatory and data exposure

The clearest externally verifiable risk is the compliance perimeter around deploying embodied-AI systems in China. China's Data Security Law applies to data processing in China and defines processing broadly to include collection, storage, use, processing, transmission, provision, and disclosure. It also requires lifecycle security controls, risk monitoring, incident response, and risk assessments for important data. For a robot brain trained on simulation and potentially connected to real-world deployments, the relevant diligence question is not whether the company has suffered a public breach; no such conclusion is supported here. The question is whether its data classification, access controls, incident response, and cross-border transfer procedures are documented and tested. The national industrial-robot safety standard GB/T 20867.1-2024 was published on 2024-08-23 and took effect on 2025-03-01, replacing the 2007 standard. That creates a concrete standards-refresh obligation for any industrial deployment, while the assigned sources do not establish Sudu's certification or conformity status.[CR001, CR002, CR003, CR004, CR005]

7.2 Execution, reliability, and commercialization

Sector evidence shows that Chinese policy is pushing humanoid and embodied AI companies from demonstrations toward verified workplace deployment. The eWeek report describes a program targeting more than 100 application scenarios and more than 10,000 humanoid robots by the end of 2026, while also noting that real-world deployment is intended to improve operational data and hardware and software performance. This creates both an adoption opportunity and an execution hazard: a company can face compressed validation timelines, safety incidents, integration failures, and warranty or service burdens before its system is proven at production scale. A Unitree comparison reports that research and education generated 73.6% of its humanoid revenue while industrial deployment generated about 9%; this is not a Sudu benchmark, but it illustrates how sector shipment narratives can run ahead of factory adoption. The fetched ManiSkill page exposed only title metadata, so it could not supply a usable technical benchmark. Sudu-specific production uptime, incident rates, manufacturing readiness, customer acceptance, and repeat deployment data are not disclosed in the assigned pool. Sim-to-real transfer and field reliability therefore remain material unresolved exposures, not evidence of failure.[CR006, CR007, CR008, CR009, CR015, CR016]

7.3 Dependency, financing, and thesis-break triggers

The sector's capital intensity and concentration make financing and dependency risks central to the thesis. An S&P Capital IQ item reports a $500 million Sudu financing announced on 2026-05-08 and describes the issuance as convertible preferred shares with participation from strategic and financial investors. That report does not disclose terms, runway, liquidation preferences, covenants, or cash availability, so the financing should not be treated as equivalent to unrestricted cash or operating runway. Sector funding evidence also indicates that large investors and corporate partners are concentrating around companies with deployments and manufacturing capacity, while the KraneShares page shows that the investable ecosystem spans brain, body, sensing, materials, and integrator companies. Sudu's supplier, cloud, capital-provider, and customer dependencies remain unverified; the practical mitigation is to obtain a dependency map, concentration schedule, financing documents, and production acceptance metrics before underwriting scale.[CR010, CR011, CR012, CR013, CR014]

7.4 Exhibits

8.1 Recommendation and entry discipline

The appropriate public-evidence stance is track/research-more, not buy. S&P Capital IQ identifies a May 2026 Sudu financing whose title labels the amount $500 million and whose body says convertible preferred shares were issued; the fetched body renders the amount inconsistently, so the exact consideration still needs transaction-document verification. The item does not state a post-money valuation, dilution, liquidation preference, conversion terms, or cash runway. A current valuation is therefore not observable from this pool. The recommendation is price-sensitive: funding raised is evidence of investor participation, not proof of cash available or an attractive entry price. The company-level record also does not establish revenue, ARR, customer count, gross margin, or production volume. Those omissions prevent a defensible mark-to-revenue or unit-economics conclusion. The disciplined next step is to obtain the cap table, signed term sheet, audited or management financials, and customer contracts before underwriting an entry price.[CV001, CV002, CV003, CV004]

8.2 Investment thesis and anti-thesis

The thesis has two evidence-supported elements. First, the reported financing names GL Ventures, BlueRun, Tencent, Ant Group, CATL, Alibaba, and IDG Capital as participants, which indicates that sophisticated financial and strategic investors assigned value to the opportunity, although it does not reveal their price or diligence findings. Second, the sector has substantial policy and capital tailwinds: The Diplomat describes embodied intelligence as a dedicated strategic-emerging-industry track in China's draft 15th Five-Year Plan. Category trackers put Figure AI near $1.9 billion of total funding and a $39 billion valuation and put Apptronik near a $5.5 billion-or-higher valuation, while Tech Market Briefs calculates a 23–29x trailing-revenue range for a reported Unitree IPO valuation scenario. These are directional references, not normalized Sudu comparables: the instruments, stages, revenue profiles, and evidence quality differ. The anti-thesis is stronger for underwriting today. Voxos reports that controlled-lab success can fall materially in deployment and that humanoid programs remain early pilots, while Morningstar's neutral rating on a broad physical-AI ETF illustrates that thematic exposure is not the same as expected outperformance. Sudu-specific commercialization, repeatability, and price support remain unverified.[CV005, CV006, CV007, CV008, CV009, CV013, CV014]

8.3 Bull, base, and bear cases

Scenario analysis should be treated as a decision framework, not a forecast. The bull case requires Sudu to convert technical differentiation and financing access into repeatable paid deployments, disclose enough operating data to support a model, and raise future capital without punitive dilution. The base case assumes continued technical progress but delayed commercialization, leaving the next round dependent on strategic investors and narrative support rather than measurable recurring revenue. The bear case combines a down-round or stalled financing with a persistent sim-to-real gap, high hardware costs, and limited customer willingness to pay. Voxos describes 90–120 minute typical humanoid battery life against 8–20 hour industrial requirements and a 60% real-world success figure after a 95% controlled result; these are sector-level signals, not Sudu measurements, but they justify a material execution discount. A Sudu valuation range should remain unstated until the company provides denominators for growth, margin, deployment, and ownership.[CV010, CV011, CV012, CV016]

8.4 Diligence asks and exit readiness

Exit readiness is currently low from the available evidence. A credible underwriting package needs the exact May 2026 financing instrument, fully diluted ownership, option pool, preference stack, debt and project-finance obligations, and any investor side letters. It also needs a monthly revenue bridge, backlog and signed customer pipeline, gross-margin assumptions by hardware and software component, deployment uptime, safety incidents, and sim-to-real performance measured outside curated demonstrations. Comparable data must be normalized by business model: Humanoid Index mixes private funding, public companies, acquisitions, and subsidiaries, while Morningstar's ETF holdings are public component and industrial companies rather than direct Sudu peers. ManiSkill documents simulation throughput and sim-to-real tooling, but that capability is not evidence of Sudu production performance. Thesis-break triggers are a financing at materially lower terms, inability to demonstrate paid repeat deployments, or failure to close the evidence gaps above.[CV003, CV009, CV015, CV016]

8.5 Exhibits