Robots-as-a-Service (RaaS) vs. Leasing vs. Buying Humanoid Robots: The Enterprise Acquisition Decision Framework (2026–2027)
Physical AI Journal Research Team
Independent analysis by Sekason Research Limited, United Kingdom
physicalaijournal.org · Decision Intelligence for the Robotics Decade
Executive Summary
Three acquisition models now define enterprise humanoid procurement in 2026: outright purchase, leasing, and Robots-as-a-Service (RaaS) — each allocating capital exposure, maintenance obligation, technology-refresh risk, and long-term financial liability differently. The choice between these models is a capital strategy decision, not a technology preference. The global RaaS fleet grew 31% in 2024 (International Federation of Robotics), confirming that enterprises across manufacturing, logistics, and automotive operations are already selecting acquisition structures — and that procurement teams without a defined model are operating without a framework.
The market evidence confirms commercial scale: the global professional service robot market sold nearly 200,000 units in 2024, up 9% year-on-year (IFR). Announced deployments reinforce the trajectory: Schaeffler has committed to up to 2,000 humanoid robots by 2032, BMW has operated Figure robots inside a live production plant, and Apptronik has raised US$935 million+ in Series A funding to scale Apollo manufacturing (company-claimed).
What the market does not yet provide is pricing transparency. As of mid-2026, Unitree is the only major humanoid manufacturer with publicly verifiable list prices — 39,900 yuan for the R1 and 99,000 yuan for the G1 (Reuters). All industrial platform vendors — Agility, Apptronik, Figure, Boston Dynamics, and Humanoid — negotiate commercial pricing privately, making competitive TCO benchmarking impossible without direct vendor engagement.
This report provides the analytical framework procurement buyers require: lifecycle economics, break-even scenarios across three deployment scales, contract risk allocation, vendor commercial readiness, and the specific conditions under which each model — buy, lease, or RaaS — produces the best financial outcome. Factual claims are classified throughout as Verified or Company-Claimed using the Physical AI Journal evidence hierarchy.
1. Executive Intelligence Synthesis
▌ DIRECT ANSWER
The central enterprise decision for humanoid robotics in 2026–2027 is acquisition model, not platform selection. Procurement teams must determine whether purchasing, leasing, or adopting RaaS delivers the best balance of capital efficiency, operational flexibility, technology-refresh risk allocation, and long-term ownership cost. Deployment activity is accelerating — announced commitments now include fleets of up to 2,000 robots — while pricing transparency and independently verified lifecycle cost data remain structurally absent across the industry.
Signal 1 — Enterprise procurement is replacing technology experimentation
Commercial humanoid deployments in manufacturing, automotive, and logistics are no longer proof-of-concept exercises. BMW has operated Figure robots inside a live production facility handling sheet-metal components at scale. Schaeffler has committed contractually to a multi-year deployment roadmap. The procurement question has shifted from "does the technology work?" to "which acquisition model best fits our capital strategy?"
Signal 2 — Commercial financing is as strategically important as robot capability
The choice between CapEx (purchase), OpEx (RaaS/lease), and hybrid models affects balance-sheet treatment, capital allocation efficiency, maintenance liability, and technology-refresh exposure. Agility Robotics' SEC-filed investor presentation is the only industrial humanoid vendor that publicly references both ownership and RaaS models, citing an illustrative subscription rate of US$8,500 per month per Digit robot (company-claimed). All other industrial vendors negotiate privately. Procurement teams that do not model acquisition structures before selecting a platform risk misaligning financial strategy with operational deployment reality.
Signal 3 — Technology depreciation risk is accelerating
Rapid iteration in embodied AI, battery technology, and robot dexterity means platforms purchased today carry meaningful obsolescence risk within 3–5 years. Apptronik introduced Apollo 2 in June 2026, underscoring the pace at which successive generations arrive. Buyers who purchase outright carry the full depreciation risk. RaaS and lease structures transfer some of this risk back to the vendor, though contract terms vary and are rarely published.
Signal 4 — Vendor financial strength is a procurement criterion, not a background factor
A vendor's ability to honour long-term RaaS contracts, provide software updates, maintain spare-parts pipelines, and deliver field service depends directly on its financial position. Apptronik raised US$520 million in a Series A extension in February 2026, reaching an approximate valuation of US$5 billion (Reuters). Agility Robotics announced a SPAC merger in June 2026 valuing the company at approximately US$2.5 billion. Humanoid raised US$152 million in July 2026. Vendor financial stability must be weighted as a procurement criterion alongside hardware capability — especially for multi-year RaaS or lease commitments.
Signal 5 — Most published ROI claims remain vendor-generated
Across all four major enterprise deployments reviewed for this report, independently verified productivity, ROI, or lifecycle cost data is either absent or incomplete. BMW has disclosed more operational detail than most — including over 90,000 sheet-metal components handled and approximately 1,250 operating hours (company-claimed) — but these figures originate from the deploying company rather than independent audit. Buyers should request independently verified uptime, MTBF, and productivity data from vendors before committing to enterprise-scale contracts.
Executive Takeaways:
(1) Acquisition model selection is a strategic CFO-level decision.
(2) RaaS fleet growth of 31% (IFR) signals enterprise preference for OpEx flexibility.
(3) Only Agility Robotics publicly references a RaaS price in investor materials.
(4) Technology depreciation risk favours flexible acquisition models for early-stage deployments.
(5) No vendor currently provides published, independently verified lifecycle cost data across purchase, lease, and RaaS.
2. Enterprise Humanoid Market Landscape
▌ DIRECT ANSWER
Enterprise humanoid deployments in 2026 are led by Agility Robotics, Apptronik, Figure AI, Humanoid, and Unitree. Only Agility Robotics publicly references a RaaS commercial model in investor materials. Most other vendors operate through pilot programmes or negotiated enterprise agreements with undisclosed terms. Buyers must evaluate commercial readiness — pricing, contract availability, service infrastructure, and upgrade policy — separately from robot specifications.
2.1 Industry Evolution: From Prototype to Production
Between 2023 and 2026, the humanoid robotics sector completed the transition from laboratory demonstration to commercial deployment. IFR data places the professional service robot market at nearly 200,000 units sold in 2024 (+9% year-on-year), with the separately tracked RaaS fleet expanding 31% over the same period — evidence that enterprises are selecting subscription and rental structures over capital ownership for robotic assets at a meaningful rate.
By mid-2026, Reuters was reporting active manufacturing deployments at BMW, Schaeffler, and Mercedes-Benz, and logistics deployments at GXO Logistics. These deployments are inside operating production facilities, not controlled demonstrations — they represent the first wave of enterprise humanoid procurement decisions reaching commercial scale.
2.2 Major Vendors and Commercial Positioning
Vendor | Platform | Primary Market | Purchase | Lease | RaaS | Public Pricing |
Agility Robotics | Digit | Warehousing / Logistics | ✔ | Not disclosed | ✔ (illustrative only) | No (SEC illustrative only) |
Apptronik | Apollo / Apollo 2 | Manufacturing / Automotive | ✔ (pilots) | Not disclosed | Not announced | No |
Figure AI | Figure 02 / 03 | Automotive Manufacturing | Enterprise agreements | Not disclosed | Not announced | No |
Humanoid | HMND 01 | Industrial Manufacturing | Enterprise agreements | Not disclosed | Not disclosed | No |
Unitree | G1, R1 | Research / Light Industrial | ✔ | No evidence | No evidence | Yes (list price) |
Tesla | Optimus | Manufacturing (future) | Not available | Not available | Not available | No |
2.3 The Acquisition Model Landscape
Three commercial structures dominate enterprise humanoid acquisition in 2026:
Purchase (CapEx): The buyer acquires the robot outright. Maintenance, software licences, battery replacement, and upgrades are either bundled under warranty or purchased separately. The buyer carries the full depreciation and residual-value risk. Currently available from Agility Robotics, Figure AI, and Unitree.
Leasing: The buyer accesses the robot under a fixed-term contract, with ownership remaining with the vendor or a third-party finance provider. Lease structures vary — some include maintenance, some do not. No industrial humanoid vendor currently publishes standard lease terms.
Robots-as-a-Service (RaaS): The buyer pays a subscription or per-unit-time fee. The vendor retains ownership, handles maintenance, and may upgrade hardware during the contract term. IFR formally defines RaaS as business models where robot ownership remains with the supplier. Agility Robotics is the only industrial humanoid vendor to reference this model in public filings, citing an illustrative rate of US$8,500/month per Digit robot (company-claimed).
2.4 Vendor Commercial Readiness Assessment
Vendor | Pilot Stage | Commercial Deployment | Enterprise Scale | RaaS Offered | Pricing Transparent |
Agility Robotics | ✔ Confirmed | ✔ Confirmed (GXO; Toyota, Schaeffler, Mercado Libre — per Agility SPAC filing, company-claimed) | Expanding (SPAC-funded) | ✔ Referenced in SEC filing (company-claimed) | Partial (illustrative only) |
Apptronik | ✔ Confirmed (Mercedes-Benz) | Expected 2027+ | Scaling (Robot Park) | Not announced | No |
Figure AI | ✔ Confirmed (BMW) | ✔ Confirmed (BMW production) | Enterprise agreements | Not announced | No |
Humanoid | Announced (Schaeffler) | Initial deployments Dec 2026 | Up to 2,000 by 2032 | Not disclosed | No |
Unitree | Research / education | Light industrial | Limited enterprise track record | No evidence found | Yes (list price only) |
2.5 The Pricing Transparency Gap
As of mid-2026, Unitree is the only major humanoid manufacturer with verifiable list prices — 39,900 yuan (R1) and 99,000 yuan (G1), both confirmed by Reuters. Every other industrial platform vendor — Agility, Apptronik, Figure, Boston Dynamics, and Humanoid — negotiates pricing privately.
This is a structural procurement challenge: buyers cannot perform competitive TCO benchmarking without requesting full lifecycle cost disclosures — hardware price, software licences, maintenance schedules, battery replacement frequency, upgrade cost, and end-of-contract obligations. Any business case built on hardware price alone will be incomplete. The Physical AI Journal Humanoid Platform Index at physicalaijournal.org contains the most current platform comparison data available.
3. What Has Actually Been Deployed?
▌ DIRECT ANSWER
Commercial humanoid deployments at enterprise scale are confirmed across four organisations in 2024–2026, but verified productivity, ROI, or lifecycle cost data from those deployments is either absent or incomplete. Named projects involving BMW, Mercedes-Benz, Schaeffler, and GXO Logistics demonstrate genuine enterprise adoption. Procurement teams must distinguish between what has been contractually announced and what has been independently measured.
3.1 Deployment Evidence Overview
Deployment | Outcome Data Available | Independent Verification | Evidence Quality |
BMW × Figure AI | High (operational metrics disclosed) | Partial — BMW company-reported | ★★★★★ |
Schaeffler × Humanoid | None — announcement only | Reuters confirmed contract | ★★★★☆ |
Mercedes-Benz × Apptronik | Limited — no metrics published | Reuters confirms partnership | ★★★☆☆ |
GXO × Agility (Digit) | Moderate — vendor-reported | SEC filing only, not independent | ★★★☆☆ |
3.2 Deployment Outcome Table
Deployment | Productivity | Labour Impact | ROI Disclosed | Independently Verified |
BMW × Figure AI | >90,000 components handled (company-claimed) | Not disclosed | No | No — BMW-reported |
Schaeffler × Humanoid | Not yet reported | Not disclosed | No | N/A — pre-deployment |
Mercedes-Benz × Apollo | Not published | Not disclosed | No | No |
GXO × Digit | 100,000+ totes moved (company-claimed) | Not disclosed | No | No — Agility-reported |
3.3 The Deployment Evidence Gap
Across all four enterprise deployments, independently verified productivity or ROI data is unavailable — meaning procurement teams have no independent benchmark for expected uptime, task-completion rate, maintenance frequency, or payback period.
IEEE Spectrum has specifically cautioned that many highly publicised humanoid demonstrations involve scripted environments, while enterprise buyers require repeatable performance under real operating conditions. Former Agility Robotics Chief Product Officer Melonee Wise told IEEE Spectrum: "The bigger problem is demand" — noting that no enterprise has yet demonstrated applications requiring several thousand humanoids in a single facility.
Physical AI Journal tracks deployment outcome evidence in its Deployment Outcome Tracker at physicalaijournal.org. Procurement teams should monitor that resource for independently verified outcome data before committing to long-term contracts.
3.4 What the Evidence Supports and What It Does Not
What the evidence supports: Commercial humanoid robots can perform structured logistics and manufacturing tasks — tote movement, sheet-metal handling, material logistics — in real production environments. This is a confirmed commercial fact, not a laboratory claim.
What the evidence does not support: No independently verified lifecycle cost, payback period, sustained uptime rate, or comparative ROI across acquisition models is currently available in the public domain. Any vendor-presented ROI figure should be treated as company-claimed until independently audited.
4. RaaS vs. Leasing vs. Buying: Which Model Delivers the Lowest Enterprise Cost?
▌ DIRECT ANSWER
No universal best acquisition model exists. RaaS reduces upfront investment and technology-obsolescence risk, making it appropriate for pilots and uncertain workloads. Purchase typically delivers lower lifetime cost when robots operate at high, predictable utilisation over several years. Leasing sits between the two — preserving capital while providing partial ownership benefits. The correct decision depends on utilisation rate, deployment duration, capital availability, contract terms, and risk allocation, not hardware price alone.
4.1 The Physical AI Journal Enterprise TCO Framework
Hardware acquisition cost is one line item in a humanoid robot deployment. The complete enterprise total cost of ownership (TCO) includes the following cost categories, each of which is treated differently across purchase, lease, and RaaS structures:
Cost Category | Description | Varies by Acquisition Model? |
Hardware | Purchase price or present value of lease/subscription payments | Yes — highest under purchase |
Integration | Facility modification, IT/OT connectivity, workflow redesign | No — comparable across models |
Software & AI licences | Ongoing OS, AI model, and autonomy stack fees | Yes — may be bundled in RaaS |
Maintenance & Repair | Scheduled servicing, unplanned downtime, spare parts | Yes — vendor bears this in RaaS |
Battery replacement | Degradation cycle replacement costs | Yes — typically buyer cost in purchase |
Cloud & connectivity | Remote operation, data transmission, fleet management | Yes — may be included in RaaS fee |
Teleoperation | Human supervisory cost per robot (where required) | No — operational cost throughout |
Training & change management | Workforce adaptation, process redesign | No — one-time but significant |
Downtime costs | Lost productivity during outages or maintenance | Yes — SLA-covered in RaaS |
End-of-life / residual value | Resale, decommission, or contract return value | Yes — buyer bears in purchase; none in RaaS |
4.2 Financial Comparison: Buy vs. Lease vs. RaaS
Variable | Buy (Purchase) | Lease | RaaS |
Upfront cost | High — full hardware price | Low to medium — first payment only | Low — first subscription period |
Ongoing monthly cost | Maintenance + software separately | Fixed lease payment (maintenance variable) | Single subscription (maintenance bundled) |
Balance sheet treatment | CapEx asset | Depends on IFRS 16 treatment | Operating expenditure (OpEx) |
Depreciation exposure | Buyer bears full risk | Shared — residual value agreed | None — vendor retains asset |
Technology refresh | Buyer funds upgrade | Limited — end of term only | Vendor may upgrade during contract |
Maintenance obligation | Buyer or separate contract | Variable by contract | Typically vendor-owned |
Battery replacement | Buyer cost | Variable by contract | Typically vendor obligation |
Vendor lock-in risk | Low — asset owned outright | Medium — contract term binds | High — operational dependency |
Capital efficiency | Low — capital tied up | Medium | High — capital preserved |
Residual value | Buyer retains (uncertain) | Fixed at contract end | None — no asset ownership |
4.3 Break-Even Analysis: Three Deployment Scenarios
The following scenarios use publicly available and editorially estimated inputs. Agility Robotics' SEC filing provides the only industrial humanoid RaaS reference in public sources — an illustrative US$8,500/month subscription rate per Digit robot (company-claimed). All other purchase and lease figures are editorial estimates based on available market evidence and disclosed manufacturing cost indications, and should be treated as indicative frameworks for internal modelling only, not vendor-confirmed pricing.
Scenario A — Pilot Deployment (1–5 robots)
Metric | Buy | Lease (Est.) | RaaS (Agility illustrative) |
Upfront cost (per robot) | ~US$150,000 (editorial estimate — based on disclosed manufacturing cost indications; not a vendor-confirmed selling price) | ~US$15,000–20,000 first payment | Nil |
Monthly cost (per robot) | Maintenance ~US$2,000–4,000 est. | ~US$6,000–8,000 est. | US$8,500 (company-claimed) |
12-month total (per robot) | ~US$174,000–198,000 | ~US$87,000–116,000 | ~US$102,000 |
Technology refresh risk | Full buyer exposure | Partial | Vendor absorbs |
Recommended model | — | — | RaaS or Lease |
At pilot scale (1–5 robots), the capital preservation argument favours RaaS or lease. The upfront cash commitment of outright purchase is disproportionate to the validation objective — and technology obsolescence risk during a pilot is highest, making vendor asset-ownership the more rational allocation.
Scenario B — Medium Deployment (10–50 robots)
Metric | Buy | Lease (Est.) | RaaS (Est.) |
Total fleet upfront cost (20 robots) | ~US$3,000,000 est. | ~US$300,000–400,000 est. | Nil |
Annual fleet OpEx (20 robots) | ~US$480,000–960,000 est. | ~US$1,440,000–1,920,000 est. | ~US$2,040,000 (illustrative) |
3-year total (20 robots) | ~US$4,440,000–5,880,000 est. | ~US$4,620,000–6,160,000 est. | ~US$6,120,000 illustrative |
Break-even vs. RaaS | Month 18–24 est. | Month 24–30 est. | Baseline |
Recommended model | If utilisation >70% and deployment confirmed | If capital constrained | If deployment uncertain |
At medium deployment scale, purchase begins to become competitive when utilisation is consistently above 70% and the deployment workflow is validated. Below that threshold, or where the deployment is still being optimised, RaaS or lease remains the lower-risk option.
Scenario C — Enterprise Fleet (100+ robots)
Metric | Buy | Lease (Est.) | RaaS (Est.) |
Total fleet upfront cost (100 robots) | ~US$15,000,000 est. | ~US$1,500,000–2,000,000 est. | Nil |
Annual fleet OpEx (100 robots) | ~US$2,400,000–4,800,000 est. | ~US$7,200,000–9,600,000 est. | ~US$10,200,000 illustrative |
5-year total (100 robots) | ~US$27,000,000–39,000,000 est. | ~US$37,500,000–50,000,000 est. | ~US$51,000,000 illustrative |
Technology refresh advantage | None — buyer funds upgrades | End of term only | Vendor may upgrade |
Recommended model | High-utilisation, stable workflow | Capital constrained, multi-site | Uncertain or evolving workflow |
At enterprise fleet scale, purchase delivers the lowest lifetime cost under stable, high-utilisation conditions. However, the five-year technology refresh risk is significant — platforms purchased in 2026 may face meaningful obsolescence pressure by 2029–2030 given the current pace of AI and hardware iteration.
4.4 Physical AI Journal CFO Decision Matrix
The Physical AI Journal CFO Decision Matrix maps enterprise procurement conditions to a recommended acquisition model:
Enterprise Situation | Recommended Model | Confidence Level | Primary Rationale |
First pilot deployment | RaaS | High | Capital preservation; exit flexibility; vendor carries obsolescence risk |
Uncertain demand or seasonal workload | RaaS or Lease | High | Avoid stranded capital in unvalidated workflow |
Capital-constrained organisation | Lease | High | Preserves CapEx budget; fixed monthly cost |
Stable, high-utilisation manufacturing workflow | Buy (Purchase) | High | Lowest lifetime cost at >70% utilisation over 3–5 years |
Rapidly evolving use case or AI iteration expected | RaaS | High | Vendor absorbs depreciation risk; upgrade flexibility |
Multi-site, phased rollout | Lease or RaaS | Medium | Scalable commercial structure; contract flexibility |
Proven workflow, confirmed long-term demand | Buy (Purchase) | High | Superior NPV over 5-year horizon vs. subscription |
4.5 Hidden Costs Procurement Teams Consistently Underestimate
Based on publicly reported deployment challenges and expert statements from IEEE Spectrum and IFR, the following cost categories are most frequently absent from initial humanoid robot business cases:
Integration and facility modification: Connectivity infrastructure, workflow redesign, and facility adaptation are consistently underestimated in early business cases and are not covered under RaaS subscriptions.
Teleoperation cost: Many current humanoid deployments rely on partial human remote supervision. The per-robot cost of teleoperation must be included in any TCO model.
Battery replacement: Battery degradation cycles are rarely disclosed publicly by vendors. Buyers purchasing outright should request replacement schedules and costs upfront.
AI licence fees: Autonomy stack, software update, and AI model licences are now charged separately from hardware by a number of vendors. These can constitute 15–25% of annual operating cost (editorial estimate — no verified benchmark currently available).
Downtime cost: No industrial humanoid vendor publishes independently verified MTBF (Mean Time Between Failures) data. Procurement teams should require MTBF commitments and SLA guarantees as a contract condition.
5. What Enterprise Deployments Reveal About Acquisition Strategy
▌ DIRECT ANSWER
Four enterprise deployments — at BMW, Schaeffler, Mercedes-Benz, and GXO Logistics — confirm that humanoid robots are performing structured logistics and manufacturing tasks in live production environments. Most organisations are still validating operational performance rather than publishing ROI data. Procurement teams should evaluate deployment maturity, commercial support, and evidence quality alongside technical capability before committing to long-term acquisition contracts.
Case Study 1 — BMW Group × Figure AI: The Strongest Current Evidence Base
BMW Group deployed Figure 02 robots inside its Spartanburg, South Carolina manufacturing plant, with the programme later expanding to Figure 03 in 2026. BMW became one of the earliest automotive manufacturers to operate a commercial humanoid robot inside a running production facility, beginning in 2025. BMW disclosed that the deployment handled over 90,000 sheet-metal components and accumulated approximately 1,250 operating hours (company-claimed), with robots operating across 5-day weeks on 10-hour shifts (company-claimed). Over 30,000 BMW X3 vehicles were produced during the deployment period (company-claimed).
BMW noted that factory integration — including 5G network connectivity and human-robot workflow configuration — required specific planning and investment beyond robot acquisition cost.
Procurement Takeaway: BMW's disclosure of operational metrics — rather than just a partnership announcement — makes this the strongest publicly documented enterprise humanoid case study available. The acquisition model (enterprise agreement with Figure AI) remains undisclosed. Buyers should model integration and connectivity costs separately from hardware acquisition cost.
Case Study 2 — Schaeffler × Humanoid: Scale Without Outcome Data
In May 2026, Reuters confirmed that Schaeffler, the German industrial manufacturer, had signed an agreement with UK-based Humanoid to deploy between 1,000 and 2,000 HMND 01 robots across its global manufacturing network by 2032. Initial deployments at Herzogenaurach and Schweinfurt are scheduled for December 2026. The deployment focuses on box handling and factory logistics. No productivity, ROI, or financial outcome data has been published.
Procurement Takeaway: The Schaeffler programme is the largest announced humanoid deployment commitment globally. It demonstrates enterprise willingness to commit to scaled acquisition — but buyers should not interpret an announced deployment plan as validated operational performance. Define measurable KPIs (uptime, task-completion rate, cost per productive hour) and contractual milestones before committing to enterprise-scale agreements. An announced programme without outcome data cannot anchor another organisation's ROI case.
Case Study 3 — Mercedes-Benz × Apptronik: The Enterprise Validation Role
Mercedes-Benz entered a commercial pilot with Apptronik to evaluate Apollo humanoid robots for material movement and logistics across production facilities in Hungary and Germany. Reuters confirmed the partnership remains active following Apptronik's US$520 million Series A extension in February 2026. No productivity metrics, robot count, or ROI figures have been disclosed publicly.
Procurement Takeaway: Mercedes-Benz's role is that of a validation partner rather than a commercial deployer — a strategically important distinction. Being named as an enterprise pilot customer is commercially valuable to the vendor. Buyers should ensure that pilot agreements include independent performance measurement, defined escalation criteria for production commitment, and explicit exit clauses if performance thresholds are not met.
Case Study 4 — GXO Logistics × Agility Robotics: The RaaS Reference
GXO Logistics has deployed Digit robots in US logistics operations, performing repetitive tote and material movement. This deployment is cited by Agility Robotics as one of its primary commercial references and is referenced extensively in its SEC-filed investor presentation. Agility reports that over 100,000 totes have been moved (company-claimed) under multi-year commercial operation (company-claimed). No independent third party has verified these figures.
Procurement Takeaway: The GXO deployment represents the longest-running published commercial humanoid deployment and is the primary real-world reference for Agility's RaaS model. All performance metrics originate from vendor investor materials. Buyers evaluating Digit for logistics should request independent references, engage GXO directly about operational experience, and benchmark the illustrative US$8,500/month RaaS rate (company-claimed) against their own task economics before building a business case.
6. What Could Prevent Expected ROI?
▌ DIRECT ANSWER
The primary acquisition risks for humanoid robots are commercial, not purely technical. Buyers face uncertainty around long-term reliability, technology obsolescence, pricing transparency, software licensing, maintenance obligations, and vendor financial stability. Most deployments remain pilots, meaning independently verified lifecycle cost and productivity data are structurally limited. Procurement teams should address these uncertainties through phased deployment strategies and contractually defined acquisition terms.
6.1 Technology Risk
Rapid AI iteration is the defining technology risk for humanoid robot procurement. Apptronik introduced Apollo 2 in June 2026, demonstrating the pace at which successive platform generations arrive. IEEE Spectrum cautions that rapid advances in generative AI do not automatically translate into commercially capable humanoid robots — current AI systems are not yet robust enough to satisfy many industrial requirements for general-purpose autonomy.
Battery technology improvement and declining costs add a parallel depreciation pressure. Buyers who purchase robots outright today may find significantly more capable platforms available at lower cost within 3–5 years. RaaS and lease structures transfer some of this risk to the vendor, though upgrade terms are rarely published.
6.2 Commercial Risk: Vendor Financial Stability
A vendor that cannot sustain long-term service, software updates, spare-parts pipelines, and RaaS contract commitments creates significant operational and financial risk.
Three recent developments illustrate both opportunity and risk:
Apptronik has raised approximately US$935 million+ (company-claimed) but does not expect production-scale commercial deployments until 2027 onward (CEO statement to Reuters, June 2026).
Agility Robotics' planned SPAC merger at approximately US$2.5 billion provides potential public-market access and capital transparency, but a SPAC is not a revenue guarantee.
Humanoid raised US$152 million in July 2026 but remains a pre-production enterprise deployer. Its Schaeffler rollout beginning December 2026 will be its earliest commercial test.
6.3 Operational Risk: Teleoperation and Reliability Requirements
IEEE Spectrum reports that factory customers expect approximately 99.99% reliability from production equipment. No major humanoid vendor has published independently verified uptime data approaching this level. Industry experts and Davos 2026 robotics discussions confirm that many current humanoid deployments still rely partly on remote human operators — a teleoperation cost that is absent from most initial business cases.
Buyers should require:
(a) independently verified MTBF data;
(b) contractual uptime guarantees with SLA penalty clauses;
(c) teleoperation cost disclosure as a separate line item in any RaaS or service contract.
6.4 Contract Risk
Contract terms for RaaS and lease agreements are not publicly available from any industrial humanoid vendor. Buyers have no market benchmark for commercially fair terms. The following elements should be treated as minimum requirements in any enterprise acquisition contract:
Contract Element | Purchase | Lease | RaaS |
SLA / uptime guarantee | Optional | Should be included | Must be contractual |
Upgrade rights | Buyer funds | End of term typically | Vendor obligation — define scope |
Software ownership / data rights | Negotiate upfront | Negotiate upfront | Critical — define explicitly |
Battery replacement obligation | Buyer bears | Define in contract | Vendor should bear |
Termination / exit clause | N/A — owned | Define notice period | Define minimum term and penalties |
Cybersecurity responsibility | Buyer | Define explicitly | Define explicitly |
Performance guarantee | Warranty only | Define clearly | Must be core to SLA |
6.5 Acquisition Risk Matrix — Physical AI Journal
Risk Category | Buy (Purchase) | Lease | RaaS |
Capital exposure | HIGH | MEDIUM | LOW |
Technology obsolescence | HIGH | MEDIUM | LOW |
Vendor lock-in | LOW | MEDIUM | HIGH |
Upgrade flexibility | LOW | MEDIUM | HIGH |
Maintenance obligation | HIGH (buyer) | MEDIUM (variable) | LOW (vendor) |
Residual / asset value | HIGH (uncertain) | FIXED (contract) | NONE |
Cash flow impact | HIGH upfront | MEDIUM (fixed monthly) | LOW (subscription) |
SLA / performance risk | LOW (owned) | MEDIUM | HIGH (vendor dependency) |
7. Which Vendor Offers the Strongest Commercial Acquisition Proposition?
▌ DIRECT ANSWER
Procurement teams should compare vendors on commercial readiness, not technical specifications alone. Key criteria are deployment maturity, pricing transparency, available financing options, service infrastructure, software support, contract flexibility, and independently verified enterprise deployments. A technically capable platform with limited commercial support may present greater long-term operational risk than a less advanced platform backed by stronger financial and service capability.
7.1 Physical AI Journal Commercial Readiness Index
The Physical AI Journal Commercial Readiness Index™ rates vendors across six commercial procurement criteria based exclusively on publicly available, independently reported information. Criteria sourced only from vendor materials carry the (company-claimed) designation.
Vendor | Deployment Evidence | RaaS Availability | Pricing Transparency | Service Infrastructure | Vendor Financial Stability | Overall Commercial Readiness |
Agility Robotics (Digit) | High (GXO confirmed per SEC filing; Toyota, Schaeffler, Mercado Libre — per Agility SPAC materials, company-claimed) | ✔ Referenced in SEC (company-claimed) | Low — SEC illustrative only | Commercial (expanding via SPAC) | High — SPAC merger announced | MEDIUM-HIGH |
Apptronik (Apollo) | Medium (Mercedes-Benz pilot) | Not announced | None public | Robot Park launched Jun 2026 | High — US$5B valuation | MEDIUM |
Figure AI (Figure 02/03) | High (BMW production deployment) | Not announced | None public | Enterprise support (undisclosed) | Medium — private funding | MEDIUM |
Humanoid (HMND 01) | Announced (Schaeffler, Dec 2026) | Not disclosed | None public | Early stage | Medium — US$1.35B valuation | LOW-MEDIUM |
Unitree (G1, R1) | Light industrial / research | No evidence found | High — published list price | Limited enterprise support | Privately held | LOW (enterprise) |
Tesla (Optimus) | None (no commercial programme) | Not available | None | N/A | Very high — public company | NOT APPLICABLE |
7.2 Best Fit by Enterprise Deployment Type
Enterprise Deployment Type | Recommended Vendor(s) | Rationale |
Warehouse / logistics automation | Agility Robotics (Digit) | Only vendor with publicly referenced RaaS; GXO and Toyota deployments as references. |
Automotive / manufacturing pilot | Figure AI or Apptronik | BMW (Figure) and Mercedes-Benz (Apptronik) provide most relevant enterprise reference base. |
Industrial manufacturing at scale | Agility Robotics or Humanoid | Both have announced large-scale industrial programmes; Humanoid's Schaeffler rollout is earliest test. |
Price-transparent research / light industrial | Unitree | Only publicly priced platform; suitable where enterprise support infrastructure is less critical. |
Capital-constrained first pilot | Agility Robotics (RaaS) | Only industrial vendor with a public reference to a RaaS commercial model. |
7.3 A Critical Pricing Observation
Enterprise procurement teams cannot perform a true competitive TCO comparison across industrial humanoid vendors as of mid-2026. Pricing, lease terms, RaaS subscription rates, maintenance costs, battery schedules, and upgrade policies are all negotiated privately. The only public industrial price reference is Agility Robotics' illustrative US$8,500/month RaaS rate (company-claimed) from its SEC-filed investor presentation — explicitly presented by Agility as an illustrative management assumption, not a contractual list price.
8. Strategic Recommendations for Automation & Procurement Buyers
▌ DIRECT ANSWER
Selective, phased adoption is the evidence-supported position for enterprise humanoid procurement in mid-2026. Organisations with repetitive, structured workflows and multi-year automation strategies should pursue structured pilots with pre-defined performance metrics. Enterprise-wide acquisition commitments should be deferred until operational performance, utilisation rates, and lifecycle economics have been validated against pre-agreed financial thresholds.
Recommendation 1 — Pilot Before Any Enterprise Commitment
No enterprise-scale humanoid acquisition commitment should precede a validated pilot demonstrating task-specific performance, uptime, and workflow integration under real operating conditions. Define what "success" means in measurable terms before signing a multi-year lease or RaaS contract. The BMW × Figure AI case study is instructive: BMW disclosed operational metrics — hours operated, components handled — before expanding the programme.
Recommendation 2 — Build the Business Case Before Selecting a Vendor
Procurement sequence matters: financial model → acquisition model selection → vendor shortlist → pilot → scale.
Selecting a vendor before defining financial thresholds creates pressure to justify the choice post-hoc. The CFO Decision Matrix in Section 4 provides the analytical starting point.
Recommendation 3 — Define Procurement KPIs in the Contract
Every acquisition contract — purchase, lease, or RaaS — should include contractually defined performance metrics: utilisation rate target, uptime guarantee with SLA penalty structure, cost per productive hour benchmark, MTBF commitment, and maintenance response time. None of the major vendors currently publish standard SLA terms. Buyers who do not negotiate these terms explicitly will have limited contractual remedy if performance falls short.
Recommendation 4 — Compare Humanoid Robots Against Alternative Automation
Before committing to any humanoid platform, model the same workflow using alternative automation: Autonomous Mobile Robots (AMRs) for logistics, collaborative robots (cobots) for assembly, Automated Storage and Retrieval Systems (AS/RS) for warehousing, and fixed-arm automation for structured manufacturing tasks. In many current deployments, these alternatives offer better-proven ROI, more transparent TCO, and lower vendor risk. Humanoid robots are justified where bipedal mobility and dexterous manipulation are operationally required — not as a default automation choice.
Recommendation 5 — Negotiate Commercial Protections Upfront
The following contract terms should be non-negotiable in any enterprise humanoid acquisition:
Software ownership and data rights — establish which party owns operational data generated by the robot during your deployment
Technology upgrade rights — define when and how hardware or software upgrades are provided and at what cost
Exit / termination clause — ensure contract exit terms do not lock the organisation into a non-performing deployment
Performance guarantee with SLA penalties — require measurable uptime and performance guarantees with defined commercial consequences for shortfall
Cybersecurity responsibility — clarify who is responsible for security patching, data transmission, and cloud connectivity obligations
Recommendation 6 — Establish a Phased Investment Roadmap
The most commercially rational approach in mid-2026 is: Monitor → Pilot → Limited Production → Scale. Commit capital in proportion to validated performance. Use RaaS or short-term lease structures for the Monitor and Pilot phases to preserve capital and maintain exit flexibility. Move toward purchase or longer-term lease only when deployment performance has been independently validated against your own financial thresholds.
9. Executive FAQ
Q1 — Is Robots-as-a-Service (RaaS) cheaper than buying a humanoid robot?
RaaS is lower-cost in the short term but typically more expensive over a long, high-utilisation deployment horizon. Using the only public reference — Agility Robotics' illustrative US$8,500/month (company-claimed) — a single robot costs approximately US$102,000 per year under subscription, compared to an estimated US$150,000–175,000 fully loaded purchase cost (editorial estimate). The break-even point — where purchase becomes cheaper than RaaS — is estimated at 18–28 months depending on maintenance costs and utilisation; no verified independent benchmark exists.
Q2 — When should an enterprise lease instead of purchase a humanoid robot?
Leasing suits organisations with capital constraints, unvalidated deployment workflows, or concerns about technology refresh where full RaaS dependency is undesirable. It provides a middle position: lower upfront commitment than purchase, more predictable cost than RaaS, and partial ownership benefit. No industrial humanoid vendor publishes standard lease terms; buyers must request them directly.
Q3 — How do I calculate the total cost of ownership (TCO) of a humanoid robot?
TCO = Hardware + Integration + Software licences + Maintenance + Battery replacement + Cloud/connectivity + Teleoperation + Training + Downtime cost − Residual value. Hardware price is typically less than 50% of total five-year TCO (editorial estimate). The Physical AI Journal Sector ROI Model Library at physicalaijournal.org provides sector-specific TCO frameworks for warehouse, manufacturing, logistics, and automotive deployments.
Q4 — Which humanoid robot vendors currently offer RaaS or leasing?
As of mid-2026, Agility Robotics (Digit) is the only industrial humanoid vendor that publicly references a RaaS commercial model, citing an illustrative rate of US$8,500/month (company-claimed). No other industrial vendor — Apptronik, Figure AI, Humanoid, or Boston Dynamics — has published RaaS pricing or standard lease terms. Unitree offers transparent purchase pricing (39,900 yuan for R1) but has no evidence of enterprise lease or RaaS programmes.
Q5 — What contract clauses should every RaaS agreement include?
Seven non-negotiable elements:
(1) Uptime guarantee with SLA metrics and penalty structure;
(2) Software and AI model upgrade commitment — frequency and cost;
(3) Data ownership — who retains operational data generated during deployment;
(4) Battery maintenance and replacement obligation;
(5) Cybersecurity responsibility allocation;
(6) Termination clause — minimum term, notice period, and early-exit cost;
(7) Performance benchmark — defined task-completion rate and cost per productive hour. No industrial vendor currently publishes standard RaaS contract terms; independent legal review of any proposed contract is essential before signature.
Q6 — At what deployment scale does buying become more economical than RaaS?
At single-robot scale, using Agility Robotics' illustrative US$8,500/month (company-claimed), outright purchase becomes economically preferable at approximately 18–28 months of sustained high-utilisation deployment (editorial estimate). At fleet scale (20+ robots), five-year total ownership cost is estimated at 15–25% lower than equivalent RaaS subscription at consistent 70%+ utilisation (editorial framework — no verified benchmark available). Below 70% utilisation, RaaS preserves capital without stranded-asset risk.
10. Scope & Disclaimer
Scope of This Report
This report evaluates the commercial, financial, and procurement dimensions of humanoid robot acquisition, including RaaS, leasing, and purchase models, enterprise deployment evidence, total cost of ownership frameworks, vendor commercial readiness, and capital allocation decision support.
This report does not evaluate, certify, or provide opinion on: engineering performance, functional safety, ISO compliance or certification status, CE conformity, AI model quality, cybersecurity certification, or regulatory compliance of any named platform. Those subjects require independent technical assessment and fall outside the scope of this commercial decision framework.
Evidence Classification
Label | Meaning |
VERIFIED | Confirmed by an independent Tier 1 or Tier 2 source (Reuters, IFR, SEC filing, IEEE, AP) |
COMPANY-CLAIMED | Originated from a manufacturer, vendor, or investor presentation — not independently verified |
Editorial estimate | Physical AI Journal analytical input based on available evidence — no verified benchmark source |
Proprietary framework | Physical AI Journal analytical tool (Acquisition Score, CFO Decision Matrix, etc.) |
This report is produced by Physical AI Journal, an independent research publication operated by Sekason Research Limited, United Kingdom (Company No. 14339910). It provides market and adoption intelligence only and does not constitute legal, financial, investment, engineering, safety-certification, or professional advice of any kind. No endorsement of any named platform or manufacturer is implied. For the full disclaimer, see physicalaijournal.org/disclaimer.
11. References & Strategic Sources
Tier 1 — Primary & Verified Sources
International Federation of Robotics (IFR)
World Robotics 2025 – Service Robots: Sources & Methods (2025) – https://ifr.org/img/worldrobotics/Sources___Methods_WR_2025_Service_Robots.pdf (Verified)
World Robotics – Service Robots (2025) – https://ifr.org/wr-service-robots (Verified)
Service Robots See Global Growth Boom (7 October 2025) – https://ifr.org/news/service-robots-see-global-growth-boom (Verified)
Reuters
Humanoid to Deploy up to 2,000 Robots at Schaeffler Plants (13 May 2026) – https://www.reuters.com/business/humanoid-deploy-up-2000-robots-schaeffler-plants-2026-05-13/ (Verified)
Humanoid Startup Apptronik Raises $520 Million (11 February 2026) – https://www.reuters.com/technology/humanoid-startup-apptronik-raises-520-million-with-backing-google-mercedes-benz-2026-02-11/ (Verified)
Apptronik Raises $350 Million to Scale Production (13 February 2025) – https://www.reuters.com/technology/apptronik-raises-350-million-scale-production-humanoid-robots-2025-02-13/ (Verified)
Apptronik Launches Robot Park and Apollo 2 (30 June 2026) – https://www.reuters.com/ (Verified)
Robotics Startup Humanoid Raises $152 Million Series A (21 July 2026) – https://www.reuters.com/ (Verified)
Unitree Launches R1 Humanoid Robot (25 July 2025) – https://www.reuters.com/ (Verified)
SEC / Agility Robotics
Agility Robotics Investor Presentation (SPAC Filing) (June 2026) – https://www.sec.gov/Archives/edgar/data/2074973/000121390026071287/ea029548401ex99-2.htm (Verified – contains company-claimed information)
IEEE
IEEE Spectrum – Humanoid Robot Scaling (2025–2026) – https://spectrum.ieee.org/ (Verified)
IEEE Robotics & Automation Society – Robots in Society, Business and Culture (May 2026) – https://www.ieee-ras.org/ (Verified)
Tier 2 — Institutional & Specialist Sources
Associated Press
Agility Robotics to Go Public Through SPAC (June 2026) – https://apnews.com/article/39f2356b9c1e167d0985b821f70079c5 (Verified)
Fortune
Agility Robotics CEO on Task Augmentation (11 June 2026) – https://fortune.com/2026/06/11/agility-robotics-c-h-robinson-ceo-task-augmentation-not-mass-layoffs/ (Verified)
Business Insider
Humanoid Robots Face Real Deployment Challenges (January 2026) – https://www.businessinsider.com/ (Verified)
Tier 3 — Company-Claimed Sources
BMW Group
BMW Group Advances the Use of Physical AI in Production with Figure 03 (2026) – https://www.press.bmwgroup.com/ (Company-Claimed)
Humanoid Robot in Leipzig (2026) – https://www.bmwgroup.com/ (Company-Claimed)
Figure AI
BMW Manufacturing Partnership (2025–2026) – https://www.figure.ai/ (Company-Claimed)
Apptronik
Apptronik Closes Over $935 Million Series A (11 February 2026) – https://apptronik.com/news-collection/apptronik-closes-over-935-million-series-a (Company-Claimed)
Agility Robotics
Digit Product Information (2026) – https://agilityrobotics.com/ (Company-Claimed)
Unitree Robotics
Unitree G1 Product Page (2025) – https://www.unitree.com/ (Company-Claimed)
Tesla
Optimus Product Updates (2025–2026) – https://www.tesla.com/AI (Company-Claimed)
Humanoid
HMND 01 Platform Information (2026) – https://www.humanoid.ai/ (Company-Claimed)
Conclusion
Automation and procurement buyers can now make one of three informed decisions:
Commission an internal humanoid robot pilot.
Request detailed RaaS and lease proposals from shortlisted vendors.
Defer large-scale investment until independently verified deployment data is available for comparable use cases.
The available evidence supports beginning structured evaluations and pilot deployments, but does not yet support enterprise-wide commitments without validated operational and financial performance.
Before making a purchasing decision, organizations should:
Define clear financial success criteria.
Build a comprehensive Total Cost of Ownership (TCO) model.
Request transparent vendor pricing and commercial terms.
Compare RaaS, leasing, and purchasing scenarios using identical assumptions.
Validate business outcomes through a controlled pilot before scaling.
This report is supported by:
Tier 1 authoritative and independently verified sources.
Tier 2 institutional and specialist publications.
Tier 3 manufacturer disclosures that are explicitly identified as Company-Claimed.
A structured, evidence-based analytical methodology designed to support enterprise capital-allocation and procurement decisions.
© 2026 Sekason Research Limited · Physical AI Journal · All rights reserved · physicalaijournal.org
