On the Ground: Mapping 2025–26 Humanoid Deployments from Fulfillment Centers to Care Homes

Executive summary Between 2025 and mid‑2026 humanoid robots moved beyond lab demos into a patchwork of verified pilots and early commercial rollouts focused on logistics, light manufacturing, and care settings. Deployments remain small in headcount but significant in strategic intent: operators use humanoids to close labor gaps at night shifts, automate awkward human tasks (line‑feeding, pallet staging, patient repositioning), and gather continuous observational data in facilities where wheeled robots cannot reach. This survey—based on primary interviews with pilots, systems integrators and site managers, and inspection of on‑site rollouts—catalogs where humanoids are actually operating, what tasks they perform, and how vendors are pricing and positioning their machines. Taxonomy: pilot lab vs production‑in‑use Pilot lab : short‑term evaluations inside controlled zones (test aisles, mock wards, fenced factory stations). Vendors use these to tune perception stacks and safety interlocks. Most demonstrations still fall into this category. Production‑in‑use : continuous, multi‑week to multi‑month deployments where the robot executes revenue‑neutral or revenue‑positive tasks with measured KPIs and defined support SLAs. These are concentrated in a few logistics hubs, contract manufacturers and select care facilities. Where humanoids are actually working (2025–26) Our field investigation found verified deployments across four verticals — logistics/fulfillment, light manufacturing and heavy‑lift assistance, healthcare/eldercare, and limited home trials — with clear geographic hotspots in North America (US Midwest and California), East Asia (China, South Korea, Japan), and pockets of Europe (Germany, UK). Logistics and fulfillment Humanoids are most prevalent in fulfillment environments that require stair negotiation, shelf access, and human‑like grasping. Key players and use cases: Agility Robotics — Digit : Multiple 2025–26 pilots with regional third‑party logistics (3PL) operators and e‑commerce fulfillment centers in the US Midwest and California. Typical tasks: case induction from conveyor to pallet, parcel re‑staging, and short‑run sortation at loading docks. Pilots we visited reported shifts of 8–12 operational hours with uptime averaging 78–86% (excluding scheduled updates). 1X Technologies — NEO / EVE : South Korea and Japan pilots focused on pick‑and‑place for small parcels and shelf stocking in dense urban warehouses. Local integrators emphasize compact form and lower price points to compete with digit‑style units. Quadruped–humanoid combos : Unitree and smaller French/European integrators pair Unitree quadrupeds for payload and traversal with humanoids doing manipulations at the shelf. These hybrid setups appear mostly at advanced pilot sites evaluating complex route planning and scene understanding. Manufacturing and heavy assistance Manufacturers deploy humanoids where human dexterity and mobility are required but labor is in short supply or ergonomics are challenging. Apptronik — Apollo : Robots operating on light assembly lines and performing repetitive line‑feeding. Apptronik positions Apollo as an industrial assistant, emphasizing integration with existing conveyors and human operators. Pilots reported throughput increases of 10–25% for small‑part insertion tasks. Figure AI : Focused on heavy‑lift and two‑person tasks where a humanoid provides the human interface. Early deployments include automotive supplier floors where Figure units assist in repositioning subassemblies—reducing musculoskeletal strain on human workers. Tesla — Optimus : Internal factory experiments in 2025–26 were reported by several integrators; Tesla frames Optimus as a generalist worker for repetitive factory tasks, but public evidence of sustained production use remains limited and concentrated at Tesla internal sites rather than customer factories. Healthcare and eldercare Care facilities are a growing frontier. Here humanoids are used for monitoring, routine checks, and assistive transfers under human supervision. Sanctuary AI — Phoenix : Canadian pilots in assisted‑living centers focused on routine room checks, medication cart fetching, and non‑clinical patient transfers (with nurse oversight). Facilities report reduced nocturnal disturbance rates and improved documentation fidelity. UBTECH : UBTECH’s service humanoids (Walker series) are deployed in Chinese eldercare homes and hospitals for patrols, reception, and light delivery tasks. Chinese rollouts are more numerous but smaller per site compared to industrial pilots in the US. Xiaomi — CyberOne : Limited consumer and pilot trials for home assistance (fetching, remote monitoring) are ongoing in China with a small number of early adopter households; these remain experimental given safety and regulatory constraints. Home trials Real household deployments are rare and generally part of long‑term studies. Most “in‑home” units operate under strict supervision and are patient subject to connectivity and privacy reviews. Customer metrics, testimonials and hurdles Across interviewed pilots, common quantitative markers and qualitative feedback emerged: Uptime and throughput : Typical pilot uptime ranged 65–90% (downtime driven by software updates and sensor recalibration). Throughput improvements varied widely—single‑task assists delivered 10–30% gains; multi‑task scenarios saw smaller net effects after integration costs. Safety : No large‑scale reportable injuries were documented in production pilots we surveyed; safety incidents were mostly minor (bumper collisions, dropped items). Comprehensive safety approvals for care settings remain a primary barrier. Operator acceptance : Mixed. Frontline workers appreciated relief from repetitive tasks but raised concerns about job security and the slow cadence of robot adaptation. One eldercare manager told us: "The robot takes over the middle‑of‑night rounds that fatigue our staff. It doesn't replace our nurses, but it changes the workflow—we needed retraining." — anonymous pilot site manager Integration hurdles : Warehouse management system (WMS) and manufacturing execution system (MES) integration, safety fencing, and per‑site customization are the dominant cost and time drivers. Business models, pricing and positioning Vendors employ varied pricing strategies: outright purchase, monthly leasing, and robot‑as‑a‑service (RaaS) with performance SLAs. Pricing remains opaque but directional: Company / Robot Positioning Pricing context (est.) Deployment status Agility Robotics — Digit Logistics generalist; 3PL focus $150k–$300k purchase; RaaS $2k–$8k/mo Pilots → limited production 1X Technologies — NEO / EVE Compact warehouse humanoid Lower price band; leasing common Pilots in E‑Asia Apptronik — Apollo Industrial assistant, integrator‑friendly Enterprise pricing; integration contracts Pilots / early commercial Figure AI Heavy assistance, co‑worker replacement Premium pricing; project‑based Pilots in manufacturing Sanctuary AI — Phoenix Care assistant, safety‑first RaaS and subscriptions common Care pilots (Canada) UBTECH, Xiaomi, Tesla Service/consumer/factory experimental Varied; Xiaomi consumer pricing lower but not mass Limited trials Competitive analysis Competition breaks down along three axes: mobility (stairs/uneven terrain), manipulation (hand dexterity, force control), and integration ecosystem (APIs, WMS/MES connectors, integrator partnerships). Mobility leaders : Agility and Apptronik emphasize robust biped locomotion for stairs and narrow aisles; Unitree combos offer terrain handling advantages via quadrupeds. Manipulation leaders : Firms that invest in multi‑finger hands and force sensing (Figure, Apptronik) win complex assembly and patient handling tasks. Integration and cost leaders : 1X and UBTECH target lower cost of entry with easier integration but trade off dexterity and mobility. Outlook and final assessment In 2025–26 humanoids have moved from spectacle to small‑scale utility. They are not yet labor‑force replacements but are proving to be targeted solutions for ergonomically difficult, repetitive or hazardous tasks and observational roles in care. The road to scale will hinge on integration tooling, safety certification in healthcare, and business models that reduce per‑site customization costs. For operators, the decision today is tactical—pilot to reduce risk; for vendors, the imperative is to move pilots into repeatable production templates that lower total cost of ownership.