Humanoid Robot Vendor Risk 2026: The Procurement Intelligence Buyers Need Before Signing a Multi-Year Contract
Physical AI Journal Research Team — Independent analysis by Sekason Research Limited, United Kingdom
Executive Intelligence Synthesis
Humanoid robot vendor risk in 2026 is primarily a supplier-continuity question, not a technology question. Interact Analysis estimates that global humanoid production exceeded 20,000 units in 2025, yet only approximately 10% were deployed in commercially viable real-world operations. Capital positions range from Apptronik's verified $935 million Series A to vendors with no disclosed public financials. The procurement challenge is not whether robots can perform tasks — it is whether the supplier can sustain a five-year commercial relationship. Evidence quality, financial durability, and contractual resilience are the variables that determine that answer.
Humanoid robot vendor risk is not a technology problem. It is a supplier-continuity problem: which vendors have the financial resilience, commercial evidence, and service infrastructure to support a five-year industrial contract — and what contractual protections to require before committing capital. This report answers both questions using publicly verified financial disclosures, SEC filings, and customer-reported deployment outcomes.
Signal 1: The production-deployment gap is the market's defining structural condition.
Interact Analysis estimates that global humanoid robot production exceeded 20,000 units in 2025 — approximately ten times 2024 production levels (verified / analyst estimate — Interact Analysis, June 2026). Of those units, only approximately 10% were deployed in commercially viable real-world operations (verified / analyst estimate). A vendor that demonstrates manufacturing throughput at scale does not automatically demonstrate the ability to support a multi-year commercial fleet. Procurement buyers who assess vendors primarily on production volume are measuring the wrong variable.
Signal 2: Capital divergence is extreme, but capital alone does not establish five-year supplier continuity.
The verified financing range across commercially active humanoid vendors runs from Apptronik's confirmed $935 million Series A total to vendors with no disclosed public financial position. Yet Unitree's August 2026 post-listing experience demonstrates the limits of capital-market signals as a durability proxy: Reuters verified an approximately 45% share-price decline from the company's post-IPO peak, with implied valuation falling by approximately $30 billion in a short period. Investor enthusiasm and supplier financial durability are not the same quantity.
Signal 3: Product-generation cycles are accelerating faster than typical procurement contract windows.
Figure AI moved from Figure 02 to Figure 03 while an active BMW Group deployment was underway. Apptronik unveiled Apollo 2 in June 2026 while the original Apollo commercial programme was ongoing. Agility Robotics is transitioning from Digit v4 to Digit v5 as its commercial order structure scales. A five-year procurement contract signed in 2026 around a specific hardware platform may outlive that platform's active support cycle. What a vendor guarantees when the contracted hardware is superseded is a procurement question, not a technology question.
Signal 4: Robots-as-a-Service structures reduce upfront exposure but concentrate five-year vendor dependency.
Agility's investor materials illustrate an $8,500/month RaaS subscription scenario (company-claimed / management model — not verified market pricing). That structure shifts the financial risk profile from capex to opex — but it simultaneously concentrates the buyer's five-year operational exposure on the vendor's software platform, cloud services, and service infrastructure. If the vendor is acquired, restructured, or exits the market, the buyer's dependency is concentrated rather than distributed across assets they own.
Signal 5: Evidence quality is unevenly distributed across the vendor landscape.
BMW Group — the customer, not the vendor — disclosed operational outcomes from its Figure 02 deployment at Spartanburg, South Carolina: more than 30,000 vehicles supported, more than 90,000 components moved, approximately 1,250 operating hours over ten months (verified — BMW customer disclosure, 27 February 2026). That is the only publicly available industrial deployment where a named enterprise customer, rather than the robot manufacturer, reported sustained operational results. Most other commercial humanoid deployments remain at announcement or pilot stage on the PAJ Evidence Pipeline.
The Humanoid Vendor Landscape in 2026: Who Is Selling, Who Is Deploying
Six vendors — Agility Robotics, Figure AI, Apptronik, Unitree, UBTECH, and Tesla with Optimus — represent the primary active commercial landscape for procurement buyers evaluating humanoid robots in 2026. The International Federation of Robotics noted in August 2025 that many companies remained at demonstrator or first-trial stages while serial production was being prepared. Chinese manufacturers account for more than 90% of current global humanoid production volume. Commercial maturity varies significantly: buyers should assess vendors by deployment evidence and contract structure, not by production capacity alone.
The International Federation of Robotics framed the 2025 market precisely in its August 2025 position paper. Serial production was being prepared, the IFR noted, while many new companies remained at demonstrator, prototype, or first-trial stages. Humanoid application fields still needed to be proven in practice. IFR President Takayuki Ito stated directly: "If and when a mass adoption of humanoids will take place remains uncertain." (IFR, 14 August 2025.)
Interact Analysis estimated that Chinese manufacturers accounted for more than 90% of global humanoid production in 2025, with approximately 75% of all units delivered within China (verified / analyst estimate — Interact Analysis, June 2026). BofA Global Research, cited by Reuters on 27 August 2026, placed China's share of global humanoid shipments at approximately 95% for 2025 (verified for the reported estimate). China's Ministry of Industry and Information Technology set a target of more than 100,000 humanoids to be built domestically in 2026 (government-claimed).
For procurement buyers outside China, this concentration has two immediate implications. First, the vendors with the highest production volumes are primarily serving domestic demand under conditions that differ from Western industrial procurement. Second, the vendors with the strongest Western commercial deployment evidence — Agility, Figure, and Apptronik — have meaningfully smaller production footprints but a stronger track record of named customer relationships in North American and European industrial environments.
The component-supply ecosystem is expanding in ways that signal broader industrial commitment to the sector.
Schaeffler CEO Klaus Rosenfeld, speaking on 5 May 2026, noted:
"We have been investing significantly in the humanoid robotics area and at the moment we are collaborating with around 45 humanoid robotics players globally."
Rosenfeld added that Schaeffler had five customer contracts in the segment and expected its humanoid robotics order book to build into the "hundreds of millions of euros by 2030" — subject to demand materialising (verified executive statement — Reuters, 5 May 2026). That signal from an industrial-component supplier indicates that the procurement infrastructure supporting humanoid deployment is developing in parallel with the vendor landscape. It does not establish a timeline for end-user adoption at scale.
PAJ Analytical Classification — not a ranking of technical capability or platform quality.
Vendor Maturity Tier | Example Vendors | Buyer Interpretation |
Commercial Deployment Evidence | Agility Robotics, Figure AI, Apptronik | Eligible for structured procurement diligence |
Large-Scale Production / Industrial Push | Unitree, UBTECH, AgiBot | Strong supply-side potential; deployment quality and geopolitical exposure require explicit assessment |
Strategic Industrial Development | Tesla Optimus, Boston Dynamics Atlas | Major corporate backing; external commercial availability not yet established at procurement scale |
Prototype / Early Deployment | Numerous smaller vendors | Pilot-only; insufficient commercial evidence for multi-year commitment |
The existence of a functioning humanoid robot does not establish a commercially durable humanoid supplier. Buyers should treat this classification as the starting point for diligence, not the conclusion.
Deployment Evidence: What Has Actually Been Confirmed
A deployment announcement confirms that a vendor and a customer are working together. It does not confirm what the deployment achieved, how the robot performed relative to operational requirements, or whether the engagement has advanced beyond a structured trial. As of mid-2026, the strongest publicly available industrial evidence comes from BMW Group's customer-disclosed operational outcome at Spartanburg, South Carolina with Figure AI. Most other commercial humanoid deployments remain at the Announcement or Pilot stage of the PAJ Evidence Pipeline. Evidence classification determines the weight a procurement buyer should assign to any given deployment reference. The Deployment Outcome Tracker at physicalaijournal.org/deployment-tracker maintains current status across named deployments.
The PAJ Evidence Pipeline stages deployment claims across four levels: Announcement — vendor and customer confirm engagement; Pilot — structured trial with defined scope but no confirmed commercial outcome; Commercial Deployment — operating contract with confirmed sustained activity; Production — repeatable commercial operation at scale with customer-reported performance data. Each stage carries different evidential weight and warrants different procurement treatment.
The distinction is not academic. A vendor that has completed 100 product demonstrations has not established the same commercial credential as a vendor with a single contractually confirmed deployment reported by the customer. Procurement buyers who treat announcement volume as a proxy for commercial maturity will systematically overestimate supplier continuity.
Deployment | Vendor | Customer | Outcome Metric | Evidence Classification | PAJ Evidence Stage |
BMW X3 production and logistics sequencing | Figure AI | BMW Group | 30,000+ vehicles; 90,000+ components; ~1,250 hrs; ~1.2M steps | Verified — BMW customer disclosure | Production |
Tote movement — Flowery Branch, GA | Agility Robotics | GXO Logistics | 100,000+ totes moved | Company-claimed | Commercial Deployment |
RaaS agreement post-pilot — Cambridge, Ontario | Agility Robotics | Toyota Motor Manufacturing Canada | Pilot stated successful; quantitative outcome not disclosed | Verified announcement; outcome not disclosed | Commercial Deployment (contract) |
Production testing | Apptronik | Mercedes-Benz | Deployment confirmed; operating metrics not disclosed | Verified deployment confirmation | Pilot |
Customer ROI assertion | Figure AI | Unnamed customers | ROI achieved (executive assertion) | Company-claimed — unverified customer ROI | Announcement |
Fleet operating hours | Agility Robotics | Multiple customers | 65,000 operating hours | Company-reported / SEC-filed | Commercial Deployment |
BMW Group's customer disclosure is the market's strongest publicly available evidence point. BMW reported that Figure 02 supported production of more than 30,000 BMW X3 vehicles across approximately ten months of operation, with robots working five days per week in 10-hour shifts, moving more than 90,000 components, logging approximately 1,250 operating hours, and recording approximately 1.2 million steps (verified — BMW Group press release, 27 February 2026). BMW subsequently advanced to Figure 03 for logistics sequencing work at the same facility (verified — BMW Group, 25 June 2026). The evidential significance is not simply that the deployment occurred — it is that the customer reported the operational outcome. That is a materially different standard than a vendor-reported milestone.
Figure's CEO Brett Adcock's assertion that early commercial customers are already achieving ROI requires explicit classification. The statement, reported via a trade publication, originates from the vendor's chief executive, not from a named customer's verified account (company-claimed — unverified customer ROI). Independent verification would require: named customer identity, deployment duration, baseline labour cost, robot utilisation rate, intervention frequency, maintenance costs incurred, and a defined payback calculation. Without those inputs, the claim is a commercial assertion, not a procurement data point.
Agility's 65,000 operating hours, disclosed in SEC-filed investor materials (company-reported / SEC-filed), represents an unusually detailed operational disclosure for the sector. It describes sustained cumulative activity across multiple customer deployments — a more informative benchmark than a demonstration count. It is not independently audited operational data. Procurement buyers should treat it as a credible operational indicator that warrants customer reference verification during diligence.
Economics and the 5-Year Exposure Model
Independently verified humanoid robot lease rates, maintenance costs, insurance, spare-parts pricing, and field-service fees remain largely undisclosed across all commercially active platforms as of mid-2026. The only publicly available reference structure is Agility Robotics' investor-disclosed illustrative RaaS scenario: an $8,500/month subscription (company-claimed / management model — not verified market pricing). Five-year humanoid contract economics must therefore be assessed structurally — by identifying what financial information a vendor must disclose before a buyer can responsibly model total cost of ownership.
See also the Physical AI Journal report RaaS vs. Leasing vs. Buying Humanoid Robots: The Enterprise Acquisition Decision Framework.
The absence of publicly available TCO data is itself a procurement signal. A buyer committing to a multi-year humanoid contract in 2026 is making a five-year financial exposure decision without a verified pricing benchmark anywhere in the public market. This is not a reason to defer all procurement — it is a reason to make vendor financial disclosure a contractual pre-condition rather than a negotiating aspiration.
Agility Robotics' June 2026 SEC-filed investor presentation provides the sector's most detailed public reference for humanoid commercial economics. These figures are illustrative management estimates and must not be treated as established market prices or contractual terms (company-claimed / management model — not verified market pricing):
1. $8,500/month illustrative RaaS subscription (company-claimed / management model — not verified market pricing)
2. Approximately $100,000 annual illustrative RaaS cost per robot (company-claimed / management model — not verified market pricing)
3. Approximately $500,000 five-year illustrative RaaS revenue to Agility per robot (company-claimed / management model — not verified market pricing)
4. Approximately $25,000 illustrative RaaS deployment fee (company-claimed / management model — not verified market pricing)
5. Approximately $400,000 five-year illustrative ownership revenue per robot (company-claimed / management model — not verified market pricing)
No other commercially active humanoid vendor has publicly disclosed comparable contract-level pricing as of mid-2026. For cross-platform pricing context, see the Physical AI Journal report Humanoid Robot Pricing 2026: Enterprise Decision Intelligence on Purchase Costs, Lease Economics & TCO.
Agility's $300 million+ in committed Digit v5 orders, disclosed in SEC materials, represents a structure of 1,000 robots under a 3-year RaaS contract (company-claimed / SEC-filed — not recognised revenue). The contractual structure matters: it establishes that Agility has multi-year, contractually committed customer relationships — a stronger commercial signal than a pipeline of prospect conversations.
Only Agility's illustrative Year 1 RaaS figure is sourced from publicly available company disclosure (company-claimed / management model). All other cells represent cost categories with no current public pricing benchmark. Buyers should require vendors to populate these categories as a condition of contract negotiation.
Cost Category | Year 1 | Year 2 | Year 3 | Year 4 | Year 5 |
Acquisition / RaaS subscription | Agility illustrative: ~$100K/robot/year (company-claimed / management model); all other vendors: not publicly disclosed | Ongoing per contract | Ongoing per contract | Ongoing per contract | Ongoing per contract |
Software / AI platform / cloud services | Not publicly disclosed across any commercial humanoid platform as of mid-2026 | Not disclosed | Not disclosed | Not disclosed | Not disclosed |
Maintenance, spare parts, field service | Not publicly disclosed | Not disclosed | Not disclosed | Not disclosed | Not disclosed |
Integration, upgrade, hardware refresh | Not publicly disclosed; product-generation cycle risk elevates this category | Not disclosed | Not disclosed | Not disclosed | Not disclosed |
Vendor-failure contingency / transition costs | Not typically modelled; must be addressed contractually before commitment | — | — | — | — |
Vendor Failure Scenarios: What the Buyer Inherits
Five structural scenarios determine the financial consequences of vendor-continuity failure across a multi-year contract.
Scenario A — Vendor survives and operates normally.
TCO follows the contracted structure. Software, service, and hardware commitments are met on the timeline agreed. This is the base case; the four scenarios below represent risks that must be contractually mitigated before signing.
Scenario B — Vendor is acquired or undergoes a change of control.
New ownership may alter pricing structures, geographic service coverage, software development priorities, or support commitments. Without explicit change-of-control provisions, the buyer may be bound to a contract with an entity that did not sign it and does not share the original commercial intent.
Scenario C — Robot model discontinued mid-contract.
Hardware support, spare parts, and software updates for a discontinued platform are at the vendor's discretion unless explicitly contracted. The accelerating platform cycle — Figure 02 to Figure 03, Apollo to Apollo 2, Digit v4 to Digit v5 — makes this a plausible scenario within any five-year contract window.
Scenario D — Vendor enters restructuring or insolvency.
Cloud services, software update infrastructure, and field-service capability are all at risk. Without source-code escrow and data-portability provisions established at contract signature, the buyer's operational investment may be stranded with no contractual remedy.
Scenario E — Vendor withdraws from the buyer's geography.
Field-service coverage, spare-parts supply, and local support infrastructure may be unavailable if the vendor's commercial priorities shift. This risk is particularly material for non-domestic vendor sourcing and for Chinese vendors serving buyers in Western markets.
Named Deployment Case Studies
Four named humanoid robot deployments represent the strongest currently available public evidence of commercial operation in 2026: BMW Group / Figure AI at Spartanburg, South Carolina (verified customer disclosure); GXO Logistics / Agility Robotics at Flowery Branch, Georgia (company-claimed operational milestone); Toyota Motor Manufacturing Canada / Agility Robotics at Cambridge, Ontario (verified commercial RaaS agreement); and Mercedes-Benz / Apptronik (deployment confirmed; outcome not publicly quantified). Evidence quality varies significantly across these four cases — that variation is itself a procurement data point about each vendor's commercial track record and reference-disclosure practice.
Evidence quality across the four case studies below ranges from customer-reported production-scale outcomes to vendor-claimed operational milestones to confirmed deployments with no disclosed performance metrics — a distribution that directly illustrates the reference-diligence challenge facing procurement buyers. These deployments are not exhaustive of all humanoid deployments; they represent the strongest publicly available evidence base on which vendor-continuity assessments can currently be grounded.
Case Study 1 — BMW Group + Figure AI, Spartanburg, South Carolina
Evidence classification: Verified — BMW customer disclosure (highest available tier)
BMW Group disclosed in February 2026 that Figure 02 had supported production of more than 30,000 BMW X3 vehicles at its Spartanburg, South Carolina facility across approximately ten months of operation. Robots operated five days per week in 10-hour shifts, moved more than 90,000 components, logged approximately 1,250 operating hours, and recorded approximately 1.2 million steps (verified — BMW Group press release, 27 February 2026).
BMW subsequently confirmed that Figure 03 had begun logistics sequencing work at the same facility, indicating a progression in the commercial relationship beyond an initial deployment trial (verified — BMW Group, 25 June 2026). Figure AI's manufacturing update of April 2026 reported that its BotQ facility had delivered more than 350 Figure 03 robots and reached a production rate of 1 robot per hour (company-claimed), consistent with the demands of an expanding BMW deployment programme.
Procurement intelligence: This is the only publicly available industrial humanoid deployment where the customer — not the robot manufacturer — disclosed a sustained operational outcome. The BMW disclosure establishes that Figure AI has a verifiable commercial relationship with a named global automotive manufacturer producing measurable operational results at production scale. It does not establish fleet-wide reliability metrics, intervention rates, software-update frequency, or service performance across a multi-year period — those remain undisclosed. Buyers should use this case as a starting reference and require service-level, hardware-refresh, and contract-continuity disclosures to complete the diligence picture.
Case Study 2 — GXO Logistics + Agility Robotics, Flowery Branch, Georgia
Evidence classification: Company-claimed operational milestone
Agility Robotics reported in November 2025 that its Digit platform had moved more than 100,000 totes in live commercial deployment at GXO Logistics' Flowery Branch, Georgia facility (company-claimed — Agility Robotics, 20 November 2025). This milestone describes sustained operational activity rather than a demonstration or structured trial — an unusually useful commercial benchmark in a sector where most evidence is announcement-stage.
The operational depth implied by 100,000+ totes of sustained activity is directionally consistent with Agility's SEC-filed cumulative operating-hours figure of 65,000 hours (company-reported / SEC-filed). However, the tote-movement figure originates from Agility, not from GXO Logistics. No independent corroboration from the customer has appeared in any publicly available source reviewed.
Procurement intelligence: The milestone supports Agility's deployment-depth claim and provides a practical operational-scale reference point. Buyers shortlisting Agility should request direct customer reference confirmation from GXO as a standard due-diligence condition. The absence of independent corroboration does not invalidate the claim — it identifies a verification step that must be completed before treating this figure as equivalent to a customer-reported outcome.
Case Study 3 — Toyota Motor Manufacturing Canada + Agility Robotics, Cambridge, Ontario
Evidence classification: Verified announcement — quantitative outcome not publicly disclosed
Agility Robotics announced in February 2026 that Toyota Motor Manufacturing Canada had signed a Robots-as-a-Service agreement following a stated successful Digit pilot at its Cambridge, Ontario manufacturing facility (verified announcement — Agility Robotics, 19 February 2026). The commercial agreement covers manufacturing, supply-chain, and logistics operations.
A multi-year RaaS agreement signed by a global automotive manufacturer, following a completed pilot, establishes that at least one major enterprise customer has made a post-pilot, recurring-revenue commitment to a humanoid platform. No quantitative performance metrics from either the pilot or the commercial deployment have been publicly disclosed.
Procurement intelligence: The absence of disclosed metrics is itself a diligence data point. Buyers evaluating Agility should inquire directly whether a non-disclosure agreement with Toyota restricts customer reference sharing, and should factor that question into their reference-check process. The RaaS structure means Toyota is carrying multi-year vendor-continuity exposure on Agility's platform. Prospective buyers may treat Toyota's willingness to sign as a secondary endorsement, while recognising it cannot substitute for disclosed performance data.
Case Study 4 — Mercedes-Benz + Apptronik, Germany
Evidence classification: Deployment and test confirmed — commercial outcome not publicly established
Mercedes-Benz Group confirmed that it was testing Apptronik's Apollo humanoid robot in production environments, including delivery of assembly kits and transport of totes within manufacturing settings (verified deployment confirmation — Mercedes-Benz Group). No quantified operational metrics — operating hours, task-completion rates, or throughput figures — have been publicly disclosed from this programme.
Mercedes-Benz is simultaneously a customer and a named strategic investor in Apptronik, forming part of the capital base that contributed to the verified $935 million Series A total (verified — Reuters, 11 February 2026). That dual customer-investor relationship means the deployment cannot be assessed as a fully arm's-length commercial evidence point.
Procurement intelligence: Enterprise customer validation from a named global manufacturer carries evidential weight regardless of the investor relationship. However, prospective buyers must note the dual role and factor it into the independence weighting they assign this reference. Buyers should request quantified outcome data and a clear explanation of how the investor relationship affects commercial terms, service commitments, and pricing when conducting diligence.
Friction, Risk, and Unresolved Issues
Eight structural risks face any automation buyer committing to a multi-year humanoid robot contract in 2026. The three highest-priority are: the production-deployment gap (approximately 10% of 2025 production in commercially viable real-world operation — Interact Analysis); the AI capability bottleneck acknowledged by the CEO of the sector's highest-volume producer; and product-generation acceleration creating hardware obsolescence risk within a typical five-year contract window. These risks are procurement-planning variables with identifiable contractual mitigations — the extent to which a vendor enables that mitigation is itself a continuity indicator.
Risk 1: The production-deployment gap creates a misleading market signal.
Of the estimated 20,000+ humanoid units produced globally in 2025 (verified / analyst estimate — Interact Analysis, June 2026), only approximately 10% were deployed in commercially viable real-world operations (verified / analyst estimate). A vendor's manufacturing throughput cannot be used as a proxy for its commercial deployment capability. Buyer risk: vendors assessed primarily on production volume will appear more commercially mature than the deployment evidence supports.
Risk 2: AI capability remains a bottleneck that even leading producers acknowledge.
Wang Xingxing, Founder and CEO of Unitree, told Reuters in August 2026 that humanoids were not yet capable enough for mass deployment and identified AI model limitations and decision-making performance as major bottlenecks constraining the sector (verified executive statement — Reuters, 20 August 2026). That acknowledgement establishes that software capability gaps are sector-wide, not company-specific. Buyer risk: hardware procurement commitments may carry an unstated dependency on AI-model improvements that are not contractually committed by any vendor.
Risk 3: Chinese vendor production ambitions may outpace sustainable commercial economics.
Reuters' August 2026 investigation found that Chinese-manufactured humanoids remained limited by dexterity, intelligence, and adaptability in factory environments despite very high stated production targets. The investigation identified concerns about overcapacity and subsidy dependence among domestic vendors. Buyer risk: low unit pricing from a Chinese vendor may not reflect sustainable vendor economics; buyers should evaluate pricing durability alongside supply-chain jurisdiction, software-update infrastructure, and service-capability evidence.
Risk 4: Financial-market valuations in the humanoid sector can be extremely volatile.
Unitree's post-IPO experience provides the sector's most current data point. Reuters verified that the company's implied valuation briefly reached approximately $66 billion post-listing before an approximately 45% share-price decline from that peak (Reuters, 25 August 2026). This does not establish that Unitree is financially distressed — the company had recently completed a capital raise. It establishes that humanoid robotics equity valuations carry volatility that buyers should not conflate with underlying operational durability.
Risk 5: Production scaling can reveal fleet-level operational challenges invisible at pilot scale.
Figure's April 2026 production update (company-claimed) acknowledged that scaling to larger fleet operations had exposed failures invisible at smaller deployment size, including what the company described as a "long tail" of edge-case failures. This is a production-management disclosure. Its procurement relevance is direct: service-level agreements and maintenance commitments must account for fleet-scale challenges that do not appear in a ten-unit pilot programme.
Risk 6: Product-generation acceleration creates hardware obsolescence within contract windows.
Figure 02 to Figure 03, Apollo to Apollo 2, and Digit v4 to Digit v5 transitions all occurred within a twelve-month period in 2025–2026. Buyers signing five-year contracts cannot assume that the contracted hardware generation will remain actively supported throughout the contract term. Hardware refresh rights, upgrade pricing, and backward-compatibility commitments are non-optional contractual requirements in this environment.
Risk 7: Service capability disclosure is not consistent across the vendor landscape.
Agility is, as of mid-2026, the only commercially active humanoid vendor to have publicly disclosed detailed field-service, software-update, maintenance, and fleet-management commitments — in its SEC-filed investor materials. Comparable disclosure from Figure, Apptronik, Unitree, and others is either absent or at a significantly lower level of specificity. Buyer risk: buyers cannot assume equivalent service infrastructure simply because a vendor asserts capability. Require written SLA commitments with financial remedies before any contract signature.
Risk 8: RaaS structures amplify vendor dependency over the contract term.
RaaS reduces upfront capital exposure and transfers technology-obsolescence risk to the vendor. However, it simultaneously concentrates the buyer's operational exposure on the vendor's software, cloud services, and service infrastructure for the full contract period. In the event of vendor acquisition, restructuring, or geographic withdrawal, a RaaS buyer carries higher operational disruption exposure than an outright purchaser who retains hardware ownership. The contractual protections required for a RaaS commitment are, accordingly, more extensive than those required for an asset-purchase transaction.
Physical AI Journal Humanoid Vendor Continuity Score™: The Procurement Assessment
The Physical AI Journal Humanoid Vendor Continuity Score™ assesses six humanoid robot vendors across nine dimensions: financial resilience, commercial traction, deployment depth, production readiness, strategic backing, service infrastructure, product-generation continuity, customer concentration risk, and contract resilience transparency.
Scores are expressed as L (Low risk) / M (Moderate) / E (Elevated) / H (High) / U (Unresolved — insufficient public evidence).
Each vendor receives a PAJ Procurement Classification: Proceed to Procurement / Pilot with Conditions / Contract Protection Required / Wait. This is a market-intelligence framework based on publicly available evidence — not a financial rating, engineering evaluation, or investment recommendation.
The Physical AI Journal Humanoid Vendor Continuity Score is a market-intelligence framework based on publicly available evidence. It is not a financial rating, a credit assessment, a bankruptcy prediction, an engineering evaluation, or an investment recommendation. Scores reflect the strength of observable evidence for each dimension, not a definitive assessment of each vendor's business prospects. Buyers should use this framework as a structured starting point for procurement diligence, not as a substitute for it.
Dimension | Weight | What Is Assessed |
Financial resilience | 20% | Disclosed capital, financing events, public market access |
Commercial traction | 15% | Disclosed paying customers, contract stage beyond pilot |
Deployment depth | 15% | Verified operating hours, task cycles, customer-reported outcomes |
Production readiness | 15% | Manufacturing capacity vs disclosed contracted demand |
Strategic backing | 10% | Named strategic investors and partners with disclosed financial stakes |
Service infrastructure | 10% | Disclosed field service, maintenance, spare-parts, and software support commitments |
Product-generation continuity | 5% | Platform evolution pace and disclosed hardware refresh provisions |
Customer concentration risk | 5% | Disclosed customer count and any disclosed revenue concentration data |
Contract resilience transparency | 5% | Public disclosure of exit, escrow, SLA, and continuity provisions |
Agility Robotics
Agility holds the strongest publicly evidenced combination of commercial metrics across all nine dimensions of any vendor in this assessment. Its $390 million+ in equity raised as of May 2026 (company-claimed / SEC-filed) and the proposed $2.5 billion pre-money SPAC transaction with $620 million+ expected gross proceeds (verified / SEC-filed — June 2026) provide meaningful financial resilience, tempered by the fact that the SPAC transaction had not yet closed at the time of research. Commercial traction is the strongest of any Western vendor in this assessment: 9 committed customer facility deployments (company-claimed / SEC-filed), the confirmed Toyota Motor Manufacturing Canada RaaS agreement (verified announcement), and the GXO Logistics commercial deployment (company-claimed). Deployment depth — 65,000 cumulative operating hours (company-reported / SEC-filed) — is the most detailed operating metric publicly available in the sector. Service infrastructure disclosure, via SEC-filed investor materials, is the most comprehensive of any vendor in this report. The company's manufacturing capacity of 10,000 robots per year at the RoboFab facility (company-claimed) compares favourably to its disclosed 1,000-robot order commitment (company-claimed / SEC-filed — not recognised revenue).
PAJ Procurement Classification: Proceed to Procurement. Standard contractual protections apply. Monitor SPAC transaction execution — the proposed public-market listing would provide enhanced ongoing financial transparency.
Figure AI
Figure's financial position is among the strongest in the sector: a verified $1 billion+ Series C commitment at a $39 billion post-money valuation (verified — Reuters, September 2025), with named strategic investors from the technology and consumer-electronics sectors (company-claimed — Figure investor disclosures). Commercial traction is anchored by the BMW Group deployment at Spartanburg — the strongest customer-reported industrial evidence in the market. Production readiness is scaling rapidly: more than 350 Figure 03 deliveries (company-claimed) by April 2026 and a production rate of 1 robot per hour at the BotQ facility (company-claimed). Service infrastructure has been established at a system level — field-service and fleet-management capability are stated (company-claimed) — but public disclosure is less detailed than Agility's SEC-filed framework. Product-generation continuity risk is Elevated: the Figure 02 to Figure 03 transition while an active BMW deployment was underway raises the question of what contractual commitments govern hardware-generation changes during a live deployment. Contract transparency is Unresolved: no public commercial contract terms or SLA frameworks have been disclosed.
PAJ Procurement Classification: Proceed to Procurement. Require explicit service-level agreements, hardware-refresh provisions, and contract-continuity terms before signing. The BMW customer relationship provides strong commercial evidence; the contract-transparency gap is the primary diligence item to resolve.
Apptronik
Apptronik's verified $520 million Series A extension (Reuters, 11 February 2026), bringing the total Series A to $935 million at an approximately $5 billion valuation (verified), establishes one of the strongest financial positions of any non-public humanoid vendor. The named strategic investors — Google and Mercedes-Benz — provide both capital durability and industrial-sector credibility. Commercial traction is Moderate relative to Agility and Figure: the Mercedes-Benz deployment is the primary publicly evidenced customer relationship, and the dual customer-investor relationship limits the independence of that reference. Deployment depth is Moderate: the deployment is confirmed, but operating hours and task-performance metrics are not publicly disclosed. The Apollo to Apollo 2 transition announced in June 2026 creates Elevated product-generation continuity risk within an already early commercial ramp period. Contract resilience transparency is Unresolved: no public commercial terms have been disclosed.
PAJ Procurement Classification: Pilot with Conditions. Capital position is strong; commercial deployment evidence at production scale is not yet established in public records. Require disclosed reference customers beyond the investor-customer relationship with Mercedes-Benz, and a field-service SLA framework, before committing to production-scale deployment.
Unitree
Unitree completed its Shanghai IPO in August 2026, priced at approximately 61 billion yuan / $9.04 billion (verified — Reuters, 6 August 2026). The subsequent approximately 45% share-price decline from a peak implied valuation of approximately $66 billion (verified — Reuters, 25 August 2026) introduces financial-market uncertainty that procurement buyers must factor into a five-year continuity assessment. Production readiness is the strongest of any vendor in this assessment: Interact Analysis estimates that Unitree and AgiBot each shipped more than 5,000 units in 2025 (verified / analyst estimate). However, commercial industrial deployment depth for Western buyers is Unresolved: production volume is concentrated in China, and the G1 platform at a starting price of $13,500 (company-claimed) is positioned primarily for research and early commercial applications rather than Western industrial fleet deployment. CEO Wang Xingxing's acknowledged software capability bottleneck and the absence of disclosed Western service infrastructure place international-buyer service performance at High risk. Geopolitical exposure requires explicit legal assessment that falls outside the scope of this framework.
PAJ Procurement Classification: Contract Protection Required for Western industrial buyers. Geopolitical exposure, acknowledged software maturity gap, post-IPO valuation volatility, and absence of documented Western service infrastructure require explicit contractual mitigation and legal-jurisdiction assessment before any multi-year commitment.
UBTECH
UBTECH's public listing on the Hong Kong Stock Exchange and its reported RMB 1.4 billion+ in 2025 humanoid orders (company-claimed) establish financial visibility and growing commercial traction in its primary market. The Airbus industrial humanoid deal (verified — Reuters / UBTECH, January 2026) is the company's strongest Western-market commercial reference, representing a named international aerospace manufacturer. Production capacity above 10,000 units in 2026 (company-claimed) signals meaningful supply-side capability. For Western industrial procurement buyers, international deployment depth, service infrastructure, and contract continuity terms are all Unresolved in public evidence. The supply-chain jurisdiction and geopolitical considerations applicable to Unitree apply equally to UBTECH: international procurement requires a legal-jurisdiction assessment that this framework flags but cannot resolve.
PAJ Procurement Classification: Contract Protection Required for Western industrial buyers. The Airbus relationship and the Hong Kong listing provide a credible international foundation. Western multi-year procurement requires geopolitical risk assessment, supply-chain jurisdiction analysis, and full contractual continuity provisions before commitment.
Tesla Optimus
Tesla's corporate financial position as the parent entity removes meaningful financial survival risk from the vendor-continuity assessment for Optimus. Strategic backing from Tesla's industrial and technology ecosystem is unmatched among all vendors assessed here. The procurement-availability assessment is, however, materially different from the financial-survival assessment. No publicly disclosed external commercial procurement agreements for Optimus exist as of mid-2026. No established commercial pricing has been confirmed: Tesla's historically discussed $20,000–$30,000 target (company-claimed target — not a current procurement price; a future production-cost aspiration reported by Reuters, not an enterprise sales structure) predates any confirmed external commercial programme. Tesla may carry the lowest corporate survival risk of any vendor evaluated here, while simultaneously carrying the highest procurement-availability risk. An automation buyer cannot make a commercial commitment to a platform that is not yet commercially available.
PAJ Procurement Classification: Wait. Reassess when external commercial availability, procurement pricing, and service terms are publicly established.
L = Low risk · M = Moderate · E = Elevated · H = High · U = Unresolved (insufficient public evidence)
PAJ market-intelligence framework. Not a financial rating, engineering evaluation, or investment recommendation.
Vendor | Fin. Resil. | Comm. Traction | Deploy. Depth | Prod. Readiness | Strategic Backing | Service Infra. | Product Continu. | Customer Concentr. | Contract Transpar. | PAJ Classification |
Agility Robotics | M | L | L | L | L | L | M | U | L | Proceed to Procurement |
Figure AI | L | L–M | L | L | L | M | E | U | H | Proceed to Procurement |
Apptronik | L | M | M | M | L | M | E | U | U | Pilot with Conditions |
Unitree | E | M | U | L | M | H | E | U | U | Contract Protection Required |
UBTECH | M | M | U | M | M | H | M | U | U | Contract Protection Required |
Tesla Optimus | L | H | H | H | L | U | M | U | H | Wait |
Strategic Recommendations for Procurement Buyers
Automation and procurement buyers evaluating humanoid robot vendors in 2026 should treat vendor continuity as a procurement variable of equal weight to robot specifications. The six priority actions are: establishing verified operating hours as a pre-commitment reference condition; demanding full five-year TCO disclosure with stated assumptions; separating vendor financial survival risk from procurement availability risk; building explicit contractual protections covering acquisition, model discontinuation, and cloud-service dependency; using the PAJ Procurement Classification to stage capital commitment; and engaging jurisdiction-specialist legal counsel for any non-domestic vendor contract. The Physical AI Journal Platform Index at physicalaijournal.org/platform-index is updated quarterly as commercial evidence evolves.
Recommendation 1:
Treat vendor continuity as a procurement variable of equal weight to robot specifications. Every buyer evaluation process that scores platforms on payload, runtime, and unit cost without scoring the vendor on financial resilience, commercial traction, and service infrastructure is missing the variable with the highest long-term consequence. The PAJ Humanoid Vendor Continuity Score provides a structured starting framework. Buyers should build an equivalent assessment into their own procurement processes, using publicly available disclosures as a starting point and direct diligence to verify.
Recommendation 2:
Require verified operating-hours data as a pre-commitment reference condition. Agility's 65,000 operating hours (company-reported / SEC-filed) represents the current sector disclosure standard. Require comparable operating-hours disclosure from every vendor shortlisted — with a clear statement of whether the figure is independently audited, company-reported, or a management estimate. For a complete pre-commitment diligence framework, see the Physical AI Journal report Humanoid Robot Vendor Due Diligence 2026: 20 Questions Before You Sign.
Recommendation 3:
Demand full five-year TCO modelling with stated assumptions before any contract is signed. Agility's illustrative model (company-claimed / management model) is the only publicly available reference structure. Use it as a disclosure template — the cost categories it identifies, not the specific figures — and require every shortlisted vendor to provide an equivalent five-year model with stated assumptions as a pre-commitment condition.
Recommendation 4:
Separate vendor financial survival risk from procurement availability risk. Tesla Optimus is the clearest illustration of why these are distinct assessments. Tesla's corporate position removes financial survival risk; procurement-availability risk for Optimus remains elevated because external commercial availability has not been established. Apply this distinction across every vendor evaluated: financial durability does not guarantee product availability, service infrastructure, or contractual terms accessible to buyers.
Recommendation 5:
Build contractual protections for every identified vendor-failure scenario.
Buyer Protection Checklist — require these provisions in every multi-year humanoid contract:
Source-code and software escrow provisions, triggered automatically by acquisition, insolvency, or product discontinuation
Explicit hardware refresh rights and upgrade pricing — stating what the vendor provides when the contracted platform is superseded
Spare-parts availability commitment for a minimum of five years post-model discontinuation date
Data portability guarantees — all operational data, robot logs, and API access remain buyer-controlled at all times, including on contract termination
Defined service-level agreements with measurable uptime thresholds and financial remedies for underperformance
Change-of-control provisions — contract continuity or buyer exit rights triggered by vendor acquisition or restructuring
Third-party servicing permissions if the vendor is unable to fulfil service obligations
Minimum financial-health disclosure obligations during the contract term (annual summary or equivalent)
Termination-for-convenience rights with defined notice periods and exit obligations
Explicit geographic coverage guarantees for service, field support, and spare-parts supply in the buyer's operating region
Recommendation 6:
For non-domestic vendor commitments, engage jurisdiction-specialist legal counsel. Geopolitical exposure, data-sovereignty requirements, and service-infrastructure availability are risk variables this framework flags but cannot resolve through evidence assessment alone. Buyers evaluating Unitree, UBTECH, or any vendor with primary operations in a jurisdiction subject to technology controls or trade restrictions should complete a jurisdiction risk assessment as a pre-commitment condition — not a post-signature review.
Executive FAQ
Which humanoid robot vendors are financially strong enough to support a five-year industrial contract in 2026?
Apptronik ($935 million Series A — verified), Figure AI ($1 billion+ Series C at a $39 billion post-money valuation — verified), and Agility Robotics (proposed $2.5 billion SPAC transaction with $620 million+ expected gross proceeds — verified / SEC-filed) hold the strongest disclosed financial positions among Western vendors as of mid-2026. Capital raised is a necessary but not sufficient condition for five-year continuity — it must be assessed alongside commercial traction, deployment depth, service infrastructure, and contractual resilience. Unitree is publicly listed at approximately $9 billion valuation (verified), but post-IPO share-price volatility of approximately 45% from peak (verified — Reuters) introduces financial-market uncertainty that Western buyers should factor explicitly into a continuity assessment.
How should a company assess humanoid robot vendor risk before signing a multi-year contract?
The PAJ Humanoid Vendor Continuity Score provides a structured assessment across nine dimensions: financial resilience, commercial traction, deployment depth, production readiness, strategic backing, service infrastructure, product-generation continuity, customer concentration risk, and contract resilience transparency. The three dimensions with the highest long-term consequence are financial resilience, deployment depth, and service infrastructure — because these determine whether the vendor can sustain the operational relationship across the full contract term. Start with publicly available disclosures and SEC filings; advance to direct reference calls and commercial diligence for vendors classified as Proceed or Pilot.
Which humanoid robot companies have the strongest evidence of real industrial deployments?
BMW Group's customer-disclosed outcome from the Figure AI deployment at Spartanburg, South Carolina — reporting more than 30,000 vehicles supported and approximately 1,250 operating hours over ten months — is the strongest publicly available evidence of industrial humanoid deployment as of mid-2026 (verified — BMW customer disclosure). Agility Robotics' 65,000 cumulative operating hours (company-reported / SEC-filed) and more than 100,000 totes moved at GXO Logistics (company-claimed) represent significant operational depth, though both figures originate from the vendor. The Deployment Outcome Tracker at physicalaijournal.org/deployment-tracker maintains current deployment status across named commercial programmes.
What happens if my humanoid robot vendor goes out of business after deployment?
Without contractual protections established before signing, five operational consequences follow: software updates and AI-model improvements cease; spare-parts supply becomes dependent on remaining inventory or third-party sourcing; cloud services and remote-monitoring infrastructure go offline; API access and data portability may be restricted; and field-service availability ends. Contractual mitigation requires source-code escrow, data portability guarantees, third-party servicing permissions, spare-parts commitments, and change-of-control provisions — all established before the contract is signed, not after a problem occurs.
What contractual protections should buyers require when purchasing or leasing a humanoid robot?
The five highest-priority provisions are:
(1) source-code and software escrow triggered by acquisition, insolvency, or product discontinuation;
(2) hardware refresh rights and upgrade pricing for when the contracted platform is superseded;
(3) spare-parts availability for a minimum of five years post-model discontinuation;
(4) data portability guarantees covering all operational data, robot logs, and API access; and
(5) change-of-control provisions giving the buyer exit rights or contract continuity guarantees if the vendor is acquired or restructured. Agility's SEC-filed investor materials (company-claimed / management model) represent the closest currently available public reference for humanoid commercial contract structure.
Should companies buy a humanoid robot now or wait until the technology matures?
Vendor selection and operational context determine the answer. Proceed to Procurement applies to Agility Robotics and Figure AI for buyers who can verify operating-hours data and negotiate service-level and contract-continuity terms. Pilot with Conditions applies to Apptronik for buyers who want production-scale deployment evidence before committing at fleet scale. Contract Protection Required applies to Unitree and UBTECH for Western buyers until geopolitical exposure and service-infrastructure availability are contractually addressed. Wait applies to Tesla Optimus until external commercial availability is established. IFR President Takayuki Ito's assessment that the timing of mass humanoid adoption "remains uncertain" (August 2025) remains the most accurate institutional summary of the sector-wide position. Buyers who apply the PAJ Procurement Classification to their specific vendor shortlist can make a staged capital commitment that accounts for current evidence and remaining uncertainty — without waiting for sector-wide maturity.
Scope & Disclaimer
This report provides market and adoption intelligence only. It does not constitute legal, financial, investment, engineering, safety-certification, or technical-conformity advice of any kind. Named companies and platforms are referenced for analytical purposes only; no endorsement of any vendor, platform, or product is implied. Company-claimed figures are labelled as such throughout and are not independently verified by Physical AI Journal; readers should verify all such figures directly with the relevant manufacturer or vendor. No claim in this report constitutes an assessment of the safety, technical soundness, engineering conformity, or fitness-for-purpose of any robotic platform or deployment. Readers must conduct independent professional, technical, legal, and financial due diligence before making procurement, deployment, investment, or capital-allocation decisions.
Full disclaimer: physicalaijournal.org/disclaimer
References & Strategic Sources
International Federation of Robotics – Humanoid Robots: Vision and Reality (14 Aug 2025) | Classification: Verified | Link
Interact Analysis – Humanoid Robot Production Surges Tenfold in 2025 (Jun 2026) | Classification: Verified / Analyst Estimate | Link
Interact Analysis – Humanoid Robot Revenue to Reach $15B by 2035 (May 2026) | Classification: Verified / Analyst Estimate | Link
Reuters – Apptronik Raises $520 Million (11 Feb 2026) | Classification: Verified | Link
Reuters – Figure Valued at $39 Billion (16 Sep 2025) | Classification: Verified | Link
Reuters – Unitree Shanghai IPO Filing (20 Mar 2026) | Classification: Verified | Link
Reuters – Unitree IPO Pricing (6 Aug 2026) | Classification: Verified | Link
Reuters – Unitree Post-IPO Volatility (25 Aug 2026) | Classification: Verified | Link
Reuters – Unitree CEO — Capability Statement (20 Aug 2026) | Classification: Verified | Link
Reuters – Schaeffler Humanoid Robotics Orders (5 May 2026) | Classification: Verified | Link
Reuters – China Humanoid Robot Market Investigation (27 Aug 2026) | Classification: Verified | Link
Reuters – Apptronik Apollo 2 Launch (30 Jun 2026) | Classification: Verified | Link
SEC – Agility / Churchill Transaction Disclosure (24 Jun 2026) | Classification: Verified (SEC-Filed) | Link
SEC – Agility Investor Presentation — RaaS Economics (Jun 2026) | Classification: Company-Claimed (SEC-Filed) | Link
BMW Group – Figure 02 and Figure 03 Spartanburg Deployment (27 Feb & 25 Jun 2026) | Classification: Verified | Link
Agility Robotics – Digit Moves Over 100,000 Totes at GXO (20 Nov 2025) | Classification: Company-Claimed | Link
Agility Robotics – Toyota Motor Manufacturing Canada Commercial Agreement (19 Feb 2026) | Classification: Verified (Announcement) | Link
Agility Robotics – Fremont Facility Announcement (16 Jul 2026) | Classification: Company-Claimed | Link
Figure AI – Ramping Figure 03 Production (29 Apr 2026) | Classification: Company-Claimed | Link
Figure AI – Figure 03 Specifications (Current) | Classification: Company-Claimed | Link
Apptronik – Apollo Specifications (23 Aug 2023) | Classification: Company-Claimed | Link
Mercedes-Benz Group – Apollo Production Testing (18 Mar 2025) | Classification: Verified (Deployment Confirmation) | Link
Unitree – G1 Humanoid Robot — Price and Specification (Current) | Classification: Company-Claimed | Link
Reuters / UBTECH – Airbus Industrial Humanoid Deal (21 Jan 2026) | Classification: Verified / Company-Reported | URL unconfirmed
This report is backed by authoritative research, independent verification, and structured analytical methodology.
© 2026 Physical AI Journal. All rights reserved. Operated by Sekason Research Limited, London, United Kingdom (Company Registration No. 14339910).
