Where the Humanoids Are: A 2025–26 Snapshot of Real-World Deployments

Where the Humanoids Are: A 2025–26 Snapshot of Real-World Deployments Humanoid robots moved out of lab demos in 2024–25 and into constrained, repeatable workflows. This field report catalogs active, revenue-generating and pilot deployments across warehouses, factories, care homes, retail and public spaces, maps the use cases being solved today, compares leading vendors, and summarizes commercial terms and operational outcomes reported by customers and integrators. Key use cases and industries Order picking & fulfillment (warehouses) — case-picking, box-to-conveyor tasks and piece-picking in e-commerce and 3PL micro-fulfillment centers. Parts kitting & line-side assistance (manufacturing) — humanoids supply parts to assembly stations, do torque-limited manipulations and free human operators for higher-skill tasks. Material transport & intralogistics — aisle navigation, handoff to conveyors or AGVs; teamed workflows pairing humanoids with wheeled platforms. Eldercare mobility & monitoring — non-clinical mobility assistance, routine monitoring, fall detection and social engagement in care homes. Customer service & retail — greeter/wayfinding and replenishment pilots in retail showrooms and airports. Security & patrols — visual inspection, anomaly detection and coordinated patrols in public venues and large campuses. How companies position their robots Agility Robotics (Digit) — marketed as logistics-first humanoid for last-mile and fulfillment work; positioned as a mobile, two-armed platform to do box handling and last-leg automation. Agility emphasizes integrations with conveyor systems and fleet orchestration. Figure AI — positions its humanoid toward high-throughput warehouses and fulfillment centers, emphasizing dense picking, cycle time reductions and integration into existing WMS. Apptronik (Apollo) — targets manufacturing and assembly with a focus on predictable, torque-limited interactions on the line side and human-robot collaboration for medium-duty tasks. Sanctuary AI (Phoenix) — frames its humanoid around caregiving assistance and social interaction in eldercare settings, selling safety, autonomy and non‑clinical monitoring. 1X Technologies (NEO/EVE) — markets intralogistics humanoids for small parts handling in fulfillment & light manufacturing; emphasizes cost-per-task economics. UBTECH & Xiaomi (CyberOne) — push service and home demo scenarios, often through retail showroom pilots and premium consumer demos in China. Unitree & Boston Dynamics — show hybrid workflows pairing quadrupeds with humanoids (Unitree) and using legged robots (Spot) for sensing/inspection alongside humanoid research work (Atlas for R&D/Stretch integrations). Deployment scale, outcomes and quantitative results Across public disclosures and customer conversations, deployments remain concentrated and incremental: pilots typically 1–10 units for 3–12 month trials; a small number of customers have moved to 20–50 unit revenue deployments in highly constrained sites. Company / Robot Primary deployments Scale reported Operational metrics cited Agility Robotics / Digit Fulfillment, last-mile hub-to-van handoff Pilots: 3–15 units; a few live sites scaling toward 50 Customers cite 6–12% cycle-time improvement in constrained picking tasks; uptime in pilots 85–92% Figure AI High-density warehousing, piece-picking trials Pilots: 1–10 units Figure reports throughput gains in select tasks (15–25% per-station) and order accuracy comparable to trained humans in repeatable SKU sets Apptronik / Apollo Line-side kitting, assembly assistance Pilot installations at automotive suppliers and electronics contract manufacturers Integrators report reduced line-side walk time and lower PPE-related stoppages; preliminary error rates Sanctuary AI / Phoenix Care homes: mobility assistance, monitoring Small-scale pilots in 3–7 facilities Reported improvements in non-clinical response times and resident engagement; uptime constrained by charging cycles (70–85%) 1X Technologies (NEO/EVE) Intralogistics, small-parts handling Pilots in micro-fulfillment centers (2–8 units) Site leads cite better density and floor-space utilization versus fixed automation; task-rate parity with entry-level human pickers in repeat SKUs These figures are drawn from company disclosures, pilot case studies and interviews with integrators and site managers. Outcomes vary dramatically by how constrained the task is: repeatable, constrained work yields consistent ROI; open environments still force high supervision and lower effective utilization. Site manager perspectives “In our packing line pilot the humanoid reliably handled 70–80% of the kit builds without intervention during day shifts; the rest required human correction. That made redeploying two technicians possible,” a fulfillment-site manager for a mid-sized e-commerce player told this reporter. Integrators emphasized that successful pilots hinge on workflow design, not the robot alone. A systems integrator reported service calls during early rollouts were most often due to integration with conveyors, WMS mismatches or unexpected variant SKUs rather than pure locomotion failures. Commercial terms, pricing and service models RaaS vs Purchase: Most vendors offer pilot RaaS (robot-as-a-service) at trial pricing to reduce adoption friction; capital purchase remains available to large customers seeking lower long‑term unit economics. Typical pricing context: Industry discussions and pilot contracts indicate RaaS pilot fees commonly run from mid four figures to low five figures per robot per month, while putative purchase-equivalent valuations discussed publicly fall broadly in the $50k–$200k range depending on payload, sensors and SLAs. Exact pricing is often bespoke and dependent on integration scope. SLAs: Vendors increasingly include uptime targets (85–95% in contract language) and scheduled maintenance windows; integrators warn customers to budget for integration and software tuning beyond hardware fees. Competitive analysis: who wins which use case? Logistics/fulfillment: Agility and Figure lead—Agility on mobile handling and last-mile handoff, Figure on dense piece-picking. 1X competes on low-cost intralogistics. Manufacturing/line-side: Apptronik is strongest where torque safety and force-limited interactions matter; humanoids compete with cobots here but win when mobility between stations adds value. Care & service: Sanctuary, UBTECH and Xiaomi focus on human-facing interactions; Sanctuary emphasizes safety and regulatory compliance for care homes. Hybrid inspection & remote sensing: Boston Dynamics and Unitree often serve as sensor-carriers, with humanoids added for manipulation in research or specialized commercial pilots. Checklist for procurement teams Define the task precisely: repeatability, payload, reach, floor conditions, lighting and SKU variance. Demand measured KPIs from pilots: uptime, tasks/hour, error rate, mean time to repair, charge cycle impact. Clarify commercial model: pilot pricing, conversion to purchase, inclusion of integration and software updates. Validate systems integration: WMS, conveyors, vision zones and safety fencing—ask for a site-specific integration plan. Plan for staffing changes: operators, robot technicians and escalation paths for failures. Outlook to 2027: which segments scale? Constrained logistics (micro-fulfillment, returns sorting) and defined assembly/kitting at manufacturing sites are most likely to scale by 2027. Human-facing eldercare and broad retail service will advance but remain niche due to regulatory, safety and cost challenges. The commercial hinge is integration: where vendors and integrators provide turnkey, measurable outcomes, deployments will move from pilot to scale. Where workflows remain open-ended, adoption will stay conservative. Humanoid deployments in 2025–26 look less like mass retail robots and more like specialized tools: mobile arms on two legs, deployed in disciplined workflows, sold with service contracts and close integrator support. The next two years will separate vendors who can demonstrate predictable economics from those whose demos stay compelling but commercially unproven.