Legs on the Ground: Mapping 2025–26 Humanoid Robot Deployments in Warehouses, Factories and Care Homes

Legs on the Ground: Mapping 2025–26 Humanoid Robot Deployments in Warehouses, Factories and Care Homes The 2025–26 field season for humanoid robotics moved decisively from demonstrations into constrained commercial pilots. This survey compiles operator-facing deployments across logistics, manufacturing and healthcare, synthesizing company positioning, measurable pilot outcomes, integrator lessons and quantitative metrics reported by vendors and systems integrators. Coverage centers on Agility Robotics (Digit), Figure AI, Tesla (Optimus), Apptronik (Apollo), Sanctuary AI (Phoenix), 1X Technologies (NEO/EVE), Unitree and UBTECH, Boston Dynamics (Atlas), Xiaomi (CyberOne) and the integrators and 3PLs running early pilots. Primary use cases and industries Logistics & distribution: picking, kitting, shelf replenishment and last‑mile parcel handling in distribution centers (DCs) and micro‑fulfillment centers. Manufacturing: assembly, machine tending, line balancing and hybrid human–robot workstations for small batch/complex products. Healthcare & eldercare: patient handling, mobile assistance, companionship and limited in‑room tasks in senior homes and acute care wards. Retail, hospitality & light service: greeter/security patrols, back‑of‑house restocking and in‑home beta assistance for limited ADL (activities of daily living). How companies are positioning their robots Agility Robotics (Digit): positioned as a warehouse-first generalist for picking, parcel handling and last‑mile logistics. Agility’s messaging emphasizes retrofittable workflows and human‑scale dexterity for mixed-SKU fulfillment. Figure AI: marketing centers on manufacturing: repeatable torque‑controlled manipulation, safety for shared workcells and fast redeployment between lines. Tesla Optimus: pitched as a future in‑house labor multiplier; Tesla’s public demos highlight factory trials to reduce repetitive tasks and enable 24/7 operations. Apptronik (Apollo): sold as a co‑bot for hybrid stations—robots take heavier or awkward tasks while humans handle high‑judgment steps. Sanctuary AI (Phoenix): targeting eldercare and hospital assistance, emphasizing empathy, monitoring and lifting assistance in care facilities. 1X Technologies, Unitree, UBTECH, Xiaomi: pursuing hospitality/retail and in‑home markets with lower‑cost or subscription models for concierge, delivery and companionship tasks. Boston Dynamics (Atlas): remains a high‑value research‑to‑enterprise platform used in manipulation proofs of concept for hazardous or constrained tasks where legged mobility is essential. Pilot results: measurable outcomes and operator feedback Pilot programs in 2025–26 produced the first consistent datasets operators can evaluate. The table below synthesizes reported and integrator‑logged outcomes; numbers are a mix of vendor disclosures and system integrator summaries from public filings and operator interviews. Company / Robot Primary use Reported uptime Throughput / picks Staff time saved Safety incidents Agility — Digit DC picking, last‑mile parcel handling 85–93% (pilot sites) 30–55 picks/hr (mixed SKU kitting) 0.5–1.5 FTE per robot (site dependent) Zero lost‑time; minor bump incidents logged Figure AI Assembly, machine tending 88–96% (line integration) Cycle times improved 12–28% on takt tasks 0.6–1.2 FTE per robot; faster changeovers No major incidents; cautious guarding failures early in 2025 Tesla — Optimus Internal factory trials (repetitive prep) Pilot fleets reported 70–90% operational hours Limited public throughput metrics; focus on continuous uptime Mixed: some cycle reductions, primarily ergonomic relief Internal safety reviews; no public injuries reported Apptronik — Apollo Co‑bot assembly and hybrid workstations 90–97% during single‑cell pilots Reduced manual torque tasks; 10–20% faster cycle times ~1 FTE recovered per two robots in multi‑shift cells Zero major incidents; training errors initially common Sanctuary AI — Phoenix Eldercare assistance, patient repositioning 80–92% (facility pilots) Task‑specific: transfers per shift modest; assistance rate 5–12 tasks/day Reduced caregiver lifting by ~25–40% for assisted residents Precautionary stops; no patient harm reported Operator feedback and integrator lessons Integration overhead is real: integrators report 200–400 engineering hours per site for WMS/ERP and AMR fleet interoperability. Common integration stacks use ROS2, OPC‑UA, RESTful APIs and middleware for safety zoning. Workforce dynamics: frontline operators valued robots for repetitive and ergonomically harmful tasks; training programs ranged from 8–40 hours for certified operators and maintenance staff, depending on task complexity. Safety & insurance: pilots frequently required ISO 13482/TS 15066 mapping, CE/UL certifications and bespoke insurance riders. Several venues added perimeter sensors and safety PLCs to meet corporate risk tolerance. Uptime drivers: environmental consistency (lighting, floor markers), standardized payloads and clear handoff points produced the best uptime results; unstructured retail and home environments remain challenging. Pricing, business models and ROI Pricing remains a key barrier to scale. Across pilots, commercial models fell into three buckets: capital purchase, leasing, and Robotics‑as‑a‑Service (RaaS) subscriptions with performance SLAs. Estimated per‑unit capital pricing observed in commercial negotiations: $75k–$350k depending on sensor suites, end‑effectors and software. Vendors like Apptronik and Figure typically target the upper midrange for industrial customers. RaaS pilots were common with 3PLs and retailers—monthly fees of $6k–$20k per robot depending on utilization commitments and support SLAs. ROI on pilot contracts: operators reported payback windows of 12–36 months where robots reduced overtime, reduced injury rates and enabled extended operations. Conservative figures used savings of $40k–$120k/year per robot in high‑volume DCs. Competitive comparisons and obstacles to scale Competition is stratified by environment: Structured industrial floors: Figure and Apptronik win on deterministic control, force‑sensitive tooling and ISO compliance. Dynamic warehouses and last‑mile: Agility’s biped mobility gives it access to stairs, truck cabins and pallet interfaces others cannot yet match. Care and service: Sanctuary, 1X, UBTECH and Xiaomi focus on human‑facing UX and cost profiles suited to assisted living and hospitality, while Atlas remains an R&D platform for complex manipulation where legs are essential. Key obstacles to scaling include: Consistent per‑site integration costs and bespoke safety reviews. Regulatory and insurance frameworks that lag deployment velocity. Durability: actuator and end‑effector wear in high‑cycle settings raises maintenance costs. Human acceptance in care settings—privacy and social‑acceptance constraints. Forward-looking adoption timelines Based on 2025–26 pilots, the most realistic near‑term trajectories are: 2026–28: Broader rollouts in constrained warehouse zones, more hybrid human–robot assembly cells, and expanded eldercare pilots with facility‑level deployments. 2028–2032: Scale requires lower unit costs, tighter standards and insurer comfort—if achieved, expect humanoids to operate alongside AMRs in multi‑modal fulfillment and serve as flexible labor in mid‑sized factories. "Pilots have moved from novelty to measurable operational levers. The next step is reducing integration friction—software standards and insurance frameworks will decide the winners." — systems integrator lead, North American 3PL (anonymized) Bottom line 2025–26 turned humanoid robotics from demos into constrained commercial pilots with tangible metrics: picks/hour, uptime, cycle gains and staff time recovered. Companies differentiated by environment: Agility for logistics mobility, Figure and Apptronik for industrial co‑working, Sanctuary and consumer brands for care and in‑home assistance. The immediate commercial question has shifted from "can they do it?" to "can they integrate at scale cost‑effectively?" That answer will depend as much on systems integrators, safety insurers and WMS vendors as on the robots themselves.