Galaxy Bot

1.1 Identity and operating model

Galaxy Bot was founded in 2023 and is publicly branded as Galbot. The retained evidence supports an embodied-AI strategy spanning foundation models and robot hardware for retail and industrial tasks, rather than a pure software business. Galbot's own site emphasizes an integrated model spanning high-level reasoning, lower-level control, and neural control, while independent MERICS analysis describes its GraspVLA, TrackVLA, and GroceryVLA models as targeted systems with narrow remits. Multiples classifies the company as privately held and gives Beijing as its headquarters, but its own introductory narrative calls the company Shenzhen-headquartered. Beijing should therefore be treated as the working location, not a fully reconciled legal headquarters fact. The company announced bookings for a new native-agent robot in September 2026, demonstrating continued product activity without establishing shipment volume, pricing, or revenue.[CO001, CO002, CO003, CO004, CO005]

1.2 Leadership and governance

Independent coverage consistently identifies He Wang as founder and ties him to Peking University; Robotics & Automation News additionally describes him as Stanford-trained. That combination offers credible founder-market fit in robot learning and embodied AI, but the assigned evidence does not identify a broader current executive team, board composition, independent directors, voting structure, or succession arrangements. The resulting diligence posture is one of apparent key-person concentration: the public narrative is anchored on one technical founder while functional leadership coverage is not visible. This is not evidence that other leaders or governance mechanisms do not exist. It is an evidence boundary that requires a current organization chart, board register, cap table, and key-person continuity plan before underwriting governance quality.[CO001, CO006]

1.3 Capital formation and stakeholder concentration

The financing record indicates rapid capital accumulation but contains internal and cross-source inconsistencies that prevent a defensible total-raised figure. The June 2025 financing is reported as $151 million by Robotics & Automation News and $153 million by Multiples, a difference consistent with conversion or rounding but retained rather than normalized. Multiples reports a $300 million December 2025 round at a $3 billion valuation, and CNBC later described Galbot as carrying a $3 billion-plus valuation. A March 2026 round was reported at RMB2.5 billion, with dollar conversions ranging from roughly $350 million to $368 million. More importantly, Multiples states both $619 million across five rounds and, elsewhere on the same page, $800 million by December 2025 before the March financing. Funding raised is therefore not presented as cash available, and ownership or control cannot be inferred from investor participation.[CO007, CO008, CO009, CO010, CO011]

1.4 Milestones and commercialization evidence

The strongest company-specific deployment evidence is CATL's reported placement of the Galbot S1 into module and battery-pack production for long-running autonomous material handling and picking. Earlier reporting also described ten unmanned convenience-store pilots in Beijing and a plan to reach one hundred stores, but that forward plan is not treated as an achieved deployment. These signals are meaningful, yet sector evidence argues against extrapolating from pilots to broad commercial maturity. MERICS says Chinese humanoids remain constrained by precision, dexterity, narrow task scope, and cost, while Associated Press reporting found that demand may lag manufacturing capacity. As a public-company benchmark rather than a Galaxy Bot metric, UBTECH reported 1,079 full-size embodied-intelligence robot units and RMB820.6 million of associated 2025 revenue. Galaxy Bot discloses no comparable audited unit, revenue, margin, incident, or retention series in the assigned evidence.[CO012, CO013, CO014, CO015, CO016]

1.5 Exhibits

2.1 Market boundary and substitutes

The investable boundary is not the entire robotics or artificial-intelligence market. Galaxy Bot's relevant near-term market comprises embodied-AI humanoid and mobile-manipulation systems sold with deployment and support for bounded industrial, logistics, retail, healthcare, and research workflows. Conventional fixed industrial robots, robotic arms, automated guided vehicles, and human labor remain the principal substitutes. This distinction matters because China produced far more conventional industrial robots than humanoids in 2025, and many factories already automate repetitive work without needing a human form. Consumer household care belongs in an adjacency, not the current serviceable market, because available evidence still describes expensive, fragile systems that depend on structured environments.[CM001, CM002, CM004, CM010, CM014]

2.2 Evidence-constrained sizing

The retained evidence does not support a clean TAM/SAM/SOM calculation. It mixes production, shipments, company sales, revenue, and long-range forecasts from different publishers and time horizons. The most useful current anchor is therefore adoption velocity: MERICS reported 12,800 Chinese humanoids produced in 2025, while Associated Press reporting cited more than 13,000 global shipments and an approximately 85% Chinese share. Morgan Stanley's roughly 28,000-unit China forecast for 2026 suggests rapid growth, but its cited $5 trillion market estimate lacks a horizon in the fetched copy. UBTECH's audited 1,079-unit, RMB820.6 million 2025 result proves that meaningful commercial revenue is possible without establishing Galaxy Bot's obtainable share.[CM003, CM004, CM005, CM006, CM007, CM008, CM009, CM015]

2.3 Buyers, users, and adoption paths

Industrial buyers offer the clearest path from pilot to recurring deployment. Plant automation and operations leaders can fund systems for material handling, picking, sorting, inspection, and other repetitive or hazardous tasks, while line workers and maintenance teams are the daily users. CATL's reported use of Galbot S1 for production-line handling and picking is a concrete workflow signal, and TechNode reported Galbot's claim of several thousand cumulative industrial orders. Logistics, research, retail, and healthcare broaden the opportunity, but their budget owners and acceptance criteria differ. Every sale therefore requires site integration, safety validation, service coverage, and an ROI case against labor or established automation rather than a generic appetite for humanoids.[CM010, CM011, CM012, CM014]

2.4 Drivers, constraints, and valuation relevance

China's dense robotics and electric-vehicle supply chains, policy support, and large factory base should accelerate component localization, training-data collection, and deployment learning. MERICS estimated that Chinese companies represented 63% of key firms in the global humanoid-component supply chain and supplied 57% of China's industrial-robot market in 2024. The counterweight is economic and technical: MERICS cited typical Chinese humanoid prices of CNY300,000-500,000 versus a CNY160,000 commercial-viability threshold, while independent reporting described limited use cases, fragile operation, and demand trailing manufacturing capacity. Dependence on Nvidia's training and edge-compute ecosystem adds geopolitical and supply risk. Valuation should therefore reward verified deployments and task-level economics, not broad market forecasts alone.[CM013, CM014, CM015, CM016]

2.5 Exhibits

3.1 A crowded field with several forms of competition

The buyer's choice set is broader than humanoid startups alone. Direct Chinese peers include UBTECH, Unitree, AgiBot, AI2 Robotics, Matrix Robotics, and EngineAI; US entrants such as Figure and Apptronik compete for capital, software talent, and multinational factory programs. Incumbent fixed arms, conveyors, and purpose-built automation remain the strongest status-quo substitutes because many plants already use them for repetitive tasks. Internal integration around third-party robot bodies and software is another viable substitute as hardware becomes cheaper and modular. The field is fragmented: Associated Press reported more than 140 Chinese humanoid manufacturers and more than 330 models in 2025. Shipment leadership is concentrated elsewhere, with AgiBot and Unitree each reported above 5,000 units in 2025 while Figure and Tesla shipped only a few hundred or fewer. Galbot's reported $1.1 billion of disclosed funding and $3 billion-plus valuation provide staying power, but neither metric proves comparable commercial scale.[CP001, CP002, CP003, CP004, CP016]

3.2 Software ambition differentiates Galbot, while peers lead visible scale

Galbot's clearest differentiation is an intelligence-first architecture for autonomous mobile manipulation. MERICS identifies separate GraspVLA, TrackVLA, and GroceryVLA models and notes Galbot's preference for synthetic training data. That specialization is meaningful in retail picking and navigation, but it also shows how narrow today's autonomy remains. UBTECH offers the strongest disclosed proof of industrial commercialization: its filing reports Walker S2 mass production, more than 6,000 units of annualized full-size production capacity at year-end 2025, and humanoid revenue of RMB820.6 million. Unitree and AgiBot lead reported shipment volume, while Figure retains a software-platform narrative at a much higher valuation. The map therefore separates visible deployment scale from autonomy orientation; it is an evidence-backed ordinal comparison, not a performance benchmark.[CP003, CP005, CP010, CP011, CP012]

3.3 Price pressure is visible, but Galbot's commercial terms remain opaque

Public evidence does not expose Galbot's list price, realized price, contract length, service fees, or discounting. The available market anchors nevertheless point to intensifying price pressure: MERICS places China's average industrial humanoid price at RMB300,000-RMB500,000 and cites a calculated commercial-viability threshold of RMB160,000, while Unitree's G1 is materially below those averages. Galbot may offset price pressure through task performance and distribution. Its CATL relationship puts the S1 on a battery production line and proposes access to CATL's aftermarket service network, a potentially important deployment and support advantage. Trust requirements are also becoming structural rather than optional: China's new robot identification regime applies a no-code, no-market-access rule and lifecycle traceability. That raises compliance costs but may favor vendors with mature manufacturing and service partners.[CP008, CP009, CP013, CP014]

3.4 The moat depends on deployment data and economics that are not yet public

Galbot can plausibly build a learning loop from retail and industrial deployments, but public evidence does not quantify fleet size, task success, uptime, intervention rates, or customer retention. Chinese suppliers already control much of the component ecosystem and domestic robot market, reducing hardware scarcity as a durable moat. At the same time, reliance on Nvidia's robotics stack and imported high-end components limits full-stack control. Independent reporting is adverse on category economics: MERICS says current humanoids remain site-specific and too expensive for widespread use, while Associated Press reports that buyers can retain fixed automation and that many orders have come from state-owned enterprises. Galbot's defensibility therefore rests on proving superior task-level ROI and accumulating proprietary deployment data faster than scaled peers, not on funding, valuation, or a single anchor partnership.[CP006, CP007, CP015, CP016]

3.5 Exhibits

4.1 Commercial proof is ahead of financial disclosure

The strongest public commercial signal is operational rather than financial: CnEVPost reports that the Galbot S1 entered CATL's production line for material-handling and picking work. The same account says the partners intend to extend CATL's after-sales system to embodied-intelligence robots, implying that deployment, maintenance, and lifecycle support may all matter to the economic model. Neither item establishes whether Galbot sells hardware, leases capacity, charges for integration, earns recurring service fees, or recognizes revenue only after acceptance. The assigned evidence also provides no contract value, realized unit price, customer concentration, revenue, ARR, or gross-margin figure. Accordingly, CATL is evidence of product use, not proof of material revenue or attractive unit economics. Underwriting should separate a robot placed in a factory from a paid, accepted, collectible deployment with a defined support burden.[CI001, CI002]

4.2 Sector price points imply a demanding ROI threshold

Public price evidence is available for the category but not for Galbot. Associated Press reporting cites roughly USD 99,000 for Matrix Robotics' MATRIX-3 and CNY 180,000 (USD 26,600) for a basic EngineAI humanoid. MERICS places China's average humanoid price at CNY 300,000-500,000 and cites a CNY 160,000 commercial-viability threshold based on CNY 80,000 of annual labor cost and a two-year ROI period. Those figures are different products, configurations, and evidence vintages, so they define a range rather than a Galbot price. Both MERICS and Morgan Stanley expect major price compression before broad adoption. That creates a margin challenge: Galbot must reduce bill-of-materials, installation, reliability, and support costs at least as quickly as market prices decline. A software-AI usage model is expressly not a valid proxy for this hardware-and-service cost structure.[CI007, CI008, CI009, CI010, CI011]

4.3 Capital raised does not resolve the cash and burn questions

CnEVPost reports a March 2026 financing of CNY 2.5 billion (USD 368 million), while CNBC reports a valuation above USD 3 billion. These are financing and valuation observations, not measures of cash currently available, operating cash consumption, debt obligations, or runway. The cost base is likely to combine robotics hardware, model training, field integration, maintenance, and working capital, but Galbot does not disclose their amounts in the assigned evidence. UBTECH offers a useful public-company boundary: it reported strong humanoid revenue growth and higher gross margin in 2025 while remaining loss-making and using cash in operations. UBTECH also reports that product revenue is recognized at a point in time and that undelivered or unaccepted projects deferred revenue, highlighting delivery acceptance and collection timing as central diligence issues. These peer data do not estimate Galbot's burn or runway.[CI003, CI004, CI012, CI013, CI014, CI015, CI016]

4.4 Financial verdict and underwriting blockers

Galbot has credible deployment and financing signals, but the present evidence cannot support a revenue-quality, margin, or capital-adequacy conclusion. Sector evidence is adverse: buyers remain limited, much reported Chinese demand has come from state-owned enterprises, and independent analysts describe current robots as too costly for widespread deployment. The investment case therefore depends on converting reference deployments into repeatable, paid units while containing installation and after-sales labor as prices fall. Before underwriting, investors need customer-level billings and collections, contract and acceptance terms, hardware and service gross margins, warranty and field support costs, inventory and receivables aging, monthly operating cash use, unrestricted cash, debt, committed capex, and the trigger for the next financing. Until those records are supplied, the financial profile should be treated as high-uncertainty and financing-dependent rather than as demonstrated scale economics.[CI001, CI002, CI003, CI005, CI006, CI008, CI009, CI010, CI012, CI013, CI014, CI015, CI016]

4.5 Exhibits

5.1 Product surface and customer workflow

Galbot's public product surface is best understood as two related operating envelopes rather than one proven general-purpose robot. The G1 is described by MERICS through three narrow vision-language-action models: GraspVLA for hands, TrackVLA for navigation, and GroceryVLA for retail. That decomposition supports an intelligence-first thesis but also shows that autonomy is still organized around bounded jobs. The S1 supplies a heavier industrial surface: CnEVPost reports a 50-kilogram dual-arm payload and operation on CATL's battery production line, including material handling and picking. These sources establish named workflows and deployed hardware, but they do not publish task-success rates, intervention frequency, cycle-time distributions, or customer ROI. Product diligence should therefore treat current proof as task-specific deployment evidence, not evidence that one model can transfer reliably across retail, manufacturing, and healthcare.[CE001, CE002, CE006, CE009]

5.2 Architecture, data, and differentiation

The visible architecture combines task-level VLA policies, synthetic grasp data, Nvidia simulation and edge-compute tooling, and a mobile manipulation platform. MERICS reports that Galbot's DexGraspNet is a simulated grasp dataset built with Nvidia Isaac Sim and that Galbot was among early recipients of Jetson Thor modules. This gives the company a plausible data and iteration advantage for dexterous manipulation, while creating an external dependency at both simulation and compute layers. The architecture remains only partially observable: the retained sources do not disclose model evaluation protocols, data-rights controls, on-device versus cloud boundaries, fallback behavior, fleet orchestration, or release management. The architecture map is therefore an evidence-backed diligence model, not a claim that the listed layers reproduce Galbot's internal implementation.[CE003, CE006, CE007, CE008]

5.3 Deployment maturity, reliability, and support

CATL is the strongest named deployment signal in the pool. CnEVPost reports long-running autonomous work by S1 in module and battery-pack production, an eight-hour battery-life claim attributed to CATL, and a plan to extend CATL's Ning Service network from battery inspection and maintenance to embodied-intelligence robots. This indicates that deployment design includes energy and field-service considerations, not only robot capability. It does not establish uptime, mean time between failures, remote-assistance burden, spare-parts availability, or whether service economics are repeatable beyond CATL. UBTECH's 2025 filing provides a useful competitor benchmark: it links mass production to preventive design, manufacturing-process controls, module testing, and complete-robot testing. Galbot has not disclosed comparable controls in the retained evidence, so technical maturity should be underwritten below demonstrated deployment visibility.[CE002, CE004, CE005, CE014, CE015, CE016]

5.4 Trust, safety, standards, and diligence gaps

Safety and quality assurance remain the least evidenced layer. ISO explains that standards set consistent benchmarks for reliability and trust, but the retained source set contains no Galbot-specific certificate, conformity assessment, formal safety test, status history, or incident register. That is an evidence gap, not evidence that incidents or controls do not exist. MERICS also cautions that Chinese humanoids generally remain limited in precision and dexterity and are commonly deployed in constrained, site-specific settings; the warning is a sector benchmark rather than a direct failure finding about Galbot. No assigned source supplies repository activity, releases, package downloads, API documentation, or a credible practitioner-community proxy, so developer adoption and software operating cadence cannot be scored. Diligence should request certification scope, hazard analyses, fail-safe behavior, penetration-test summaries, privacy and training-data governance, field incidents, uptime, intervention logs, and a versioned roadmap before assigning production-grade maturity.[CE009, CE010, CE011, CE012, CE013]

5.5 Exhibits

6.1 Customer base: industrial proof, broader positioning

Galbot's strongest identifiable customer segment is large-scale industrial manufacturing. CATL has placed the S1 on a smart production line for module and battery-pack handling and picking, moving the relationship beyond a logo or demonstration. Retail remains a product-positioning segment rather than a verified customer cohort in the reviewed material: MERICS describes GroceryVLA as a model for retail scenarios, but names no Galbot retailer. Industry evidence also points to research laboratories, state-owned enterprises, logistics operators, factories, and malls as active humanoid buyers or deployment sites, yet those observations cannot be attributed to Galbot's own customer base. The resulting segmentation is therefore evidence-tiered rather than a claimed account census.[CU001, CU002, CU007, CU010, CU014]

6.2 Adoption proof and trajectory

CATL is the clearest marker of production adoption: CnEVPost reports that the S1 performs long-running autonomous work rather than a one-off pilot. The wider demand environment is credible but uneven. China had 470 operational industrial robots per 10,000 manufacturing employees in 2023, indicating an automation-literate buyer base, while 2026 reporting says Chinese humanoid vendors are already shipping into factories and malls. Peer disclosures show that commercialization can become financially material—UBTECH reported 1,079 full-size robots and RMB820.6 million of related 2025 revenue—but those peer figures are not evidence of Galbot volume, revenue, or repeat usage. No Galbot account count or installed-base series is available in the assigned evidence.[CU002, CU009, CU010, CU011, CU012, CU013]

6.3 Durability and expansion

The CATL relationship contains one credible expansion mechanism: the parties intend to fold embodied-intelligence robots into CATL's established Ning Service inspection, maintenance, and recycling network. That could lower field-support friction and enable expansion across production sites, but the evidence does not state a rollout schedule, committed unit count, contract term, renewal trigger, or economics. CATL is also described as an investor in Galbot, so deployment evidence and strategic backing are intertwined rather than fully independent customer validation. With no disclosed NRR, GRR, churn, renewal, utilization, satisfaction, or cohort data, customer durability cannot be underwritten from public evidence.[CU003, CU004, CU015]

6.4 Customer risk and underwriting implications

Customer risk is dominated by proof concentration and category-level commercialization uncertainty. The reviewed Galbot-specific evidence names CATL as the sole production reference, while independent industry analysis says humanoids remain expensive, constrained to limited tasks, and often deployed in site-specific trials. Associated Press reporting adds that 2025 Chinese humanoid orders were heavily supported by state-owned enterprises and research demand, and that buyers may lag manufacturing capacity. Galbot may still build a durable industrial wedge, but current evidence cannot distinguish paid production units from strategic pilots, quantify revenue concentration, or demonstrate renewal economics. An investment case should therefore require contract-level proof from multiple unrelated customers.[CU005, CU006, CU007, CU008, CU015]

6.5 Exhibits

7.1 Ranked risk posture

Commercialization and execution are the highest residual risks. Independent reporting says humanoid supply capacity may be running ahead of buyer demand and that robots remain expensive, fragile, and dependent on highly structured environments. MERICS separately finds that Chinese humanoids still lack precision and dexterity and are generally confined to limited tasks or site-specific trials. Those findings matter directly to Galbot because a successful CATL deployment proves access to a demanding industrial setting, not repeatable economics across customers or unstructured workflows. The investment case should therefore be conditioned on measured task success, intervention frequency, uptime, deployment cost, and renewal or expansion evidence rather than order announcements or technical demonstrations. Peer results reinforce the financial risk: UBTECH grew humanoid revenue sharply in 2025 yet remained loss-making, so sector commercialization can scale before operating self-sufficiency is achieved.[CR001, CR002, CR003, CR007, CR012, CR013, CR015, CR016]

7.2 Regulatory, legal, and geopolitical exposure

The assigned evidence does not establish Galbot's product-safety certifications, privacy controls, incident record, insurance coverage, export-control classification, or litigation posture. That is a material diligence gap, not evidence that no incidents or proceedings exist. The clearest externally evidenced regulatory exposure is prospective: MERICS identifies dependence on Nvidia's robotics ecosystem and notes that increasingly capable edge-computing chips could become export-control targets. CNBC also reports that US-China tensions and national-security scrutiny have chilled cross-border investment. These conditions can constrain compute access, financing sources, overseas customers, and exit optionality even without a company-specific enforcement action. Investment documentation should require a jurisdiction- by-jurisdiction compliance matrix, safety test files, incident logs, data-governance architecture, and supplier substitution plan before treating this risk as mitigated.[CR004, CR005, CR009, CR010]

7.3 Operational and partner dependencies

Galbot's CATL relationship is both evidence of deployment and a concentration risk. Reporting says the S1 entered CATL's smart production line for material handling and picking, while CATL's Ning Service network is expected to extend into embodied-robot after-sales support. This can reduce field-service friction, but it also leaves deployment credibility and support capacity partly dependent on one strategic investor-customer ecosystem. Technical concentration is broader: MERICS reports dependence on Nvidia software and chips and on German and Japanese precision-component suppliers. Monitoring should therefore separate Galbot-controlled performance from partner-provided batteries, service, compute, and components. Kill criteria include failure to qualify alternative compute or critical components, persistent human intervention above contracted thresholds, or an inability to reproduce CATL-level performance at two unrelated customers.[CR004, CR005, CR007, CR008, CR016]

7.4 Mitigation priorities and thesis-break triggers

The strongest visible mitigation is a structured industrial deployment with a large manufacturer and an announced route into that partner's service network. It does not resolve demand breadth, safety governance, reliability, or economics. Management should provide monthly fleet uptime, autonomous task-completion rate, human interventions per operating hour, mean time between failures, field-service cost, gross margin by deployment, receivable aging, and customer concentration. A thesis break should be triggered by stalled production deployments, repeated safety-critical incidents, restricted access to essential compute without a qualified substitute, loss of CATL service support, or unit economics that do not improve with deployment scale. Competitive concentration also matters: TrendForce projects Unitree and AgiBot to capture nearly 80% of 2026 shipments, increasing the risk that Galbot's funding and valuation outrun its commercial share.[CR001, CR006, CR007, CR008, CR011, CR014, CR016]

7.5 Exhibits

8.1 Recommendation and price discipline

The recommendation is research-more rather than buy or pass. CNBC reported an April 2026 private valuation above $3 billion for Galbot, while CNEVPost reported a RMB2.5 billion ($368 million) financing announced in March. Those marks show access to capital and investor demand, but neither establishes enterprise value on fully diluted terms, the preferences attached to the latest securities, or a current operating multiple. The reviewed public record contains no Galbot revenue, gross margin, cash-burn, backlog, customer-concentration, or unit-economics disclosure. Without those denominators, a fair-value range or target return would be false precision. Entry should therefore remain conditional on a primary data room, contract-level deployment proof, and a financing-terms review.[CV001, CV004, CV016]

8.2 Thesis and anti-thesis

The positive case is not purely promotional: CNEVPost reported that Galbot S1 entered CATL's smart production line for long-running autonomous material-handling and picking work. That is a strategically relevant customer setting and a better proof point than a staged demonstration. The anti-thesis is that sector supply is advancing faster than verified demand. Associated Press reporting described limited use cases, more than 140 Chinese humanoid manufacturers and more than 330 models in 2025, while MERICS characterized most deployments as limited tasks and site-specific trials. A Robotics & Automation News homepage excerpt also framed the move from demonstrations and pilots to factories as a manufacturing-scale challenge. Galbot therefore has evidence of technical and partner progress, but not yet public proof of repeatable commercial economics.[CV005, CV006, CV008, CV015]

8.3 Comparables and scenario logic

Private peer marks are useful only as sentiment references. CNBC placed Figure at at least $39 billion, Apptronik at $5 billion, Galbot above $3 billion and AI2 Robotics at roughly $2.93 billion, while also reporting that US humanoid startups are perceived as broad AI platforms and Chinese companies more as industrial hardware plays. Those prices do not normalize security rights or operating maturity. UBTECH offers a more concrete operating reference because its 2025 filing disclosed RMB2.001 billion of revenue, RMB820.6 million from full-size embodied humanoid products and services, a 37.7% gross margin, and a RMB789.8 million accounting loss. It is still a competitor filing, not an independent valuation benchmark, and its diversified product mix makes direct multiple transfer inappropriate.[CV002, CV003, CV010, CV011, CV012, CV013, CV014]

8.4 Exit readiness, diligence asks, and thesis breaks

Exit readiness cannot be established from the assigned public evidence. A financing mark is not liquidity, and funding raised is not cash available; no reviewed source supplies a Galbot IPO timetable, secondary-market terms, audited financial history, or acquisition precedent. The key diligence asks are a fully diluted cap table with liquidation preferences, audited revenue and cash-flow statements, signed customer contracts, fleet uptime and intervention logs, warranty and service costs, and a bridge from deployments to gross margin. Thesis-break triggers are measurable: failure to convert CATL-related work into paid repeat deployments, persistent economics above viable labor substitution, dependence on subsidized or state-linked demand, or a financing round below the current mark. Until those points are tested, downside cannot be bounded and upside cannot be converted into a target return.[CV005, CV007, CV009, CV013, CV016]

8.5 Exhibits