Humanoids in Production: Mapping 2025–26 Real‑World Deployments and Use Cases

Overview: Where humanoids are actually working in 2025–26 After a decade of lab demos and fundraising headlines, 2025–26 is when humanoid robots moved from proof‑of‑concepts into focused, constrained pilots and a handful of early production roles. This report catalogs boots‑on‑the‑ground deployments and credible pilot programs across warehouses, manufacturing lines, healthcare/assisted‑living sites, hospitality/retail, and home/education trials. It synthesizes company statements, integrator commentary, and end‑user implementation patterns to show which tasks are viable today, which remain aspirational, and which firms are closest to repeatable commercial rollouts. Warehousing and order fulfillment Logistics remains the low‑hanging fruit for bipedal systems because aisles, pallets and box handling map naturally to human‑centric environments. Agility Robotics (Digit) Agility’s Digit has been deployed in multi‑week pilots at e‑commerce and third‑party logistics sites to perform case/box picking from low shelves and tote transfer into conveyors. Company specs cite an 18 kg payload and modular gripper options; integrators report throughput in early pilots at roughly 30–50% of an experienced human picker for unconstrained SKUs but competitive in repetitive, uniform-box lanes. Uptime in pilots typically ranges 70–85% due to battery swaps and localization resets. 1X Technologies (NEO/EVE) Smaller teams such as 1X are marketing humanoids as drop‑in floor workers for late‑night fulfillment and returns sorting. Pilots emphasize lower labor cost per hour and flexible redeployment. Pricing models reported to customers use leasing plus per‑robot support fees rather than large capital purchases. What tasks are being automated Box/case picking from low to mid‑shelf locations Tote to conveyor transfers and simple stow/unstow Late‑stage sorting and returns handling in mixed SKU lanes Manufacturing and mixed human‑robot assembly Manufacturers are trialing humanoids for kitting, parts transfer, and collaborative assembly where human ergonomics or two‑hand tasks are involved. Figure AI and Apptronik Both firms report multi‑month trials in automotive and electronics assembly plants. Figure positions its systems for higher‑speed, repetitive insertions and torque‑limited tasks; Apptronik highlights mobile manipulation and close‑coordination with human line workers. Typical pilot outcomes: reduced worker fatigue in kitting, modest improvements in takt time (single‑digit percent) after process redesign, and a need for extensive PLC/WMS integration. Tesla (Optimus) Tesla has run internal proof‑of‑concepts using Optimus prototypes for parts ferrying and simple support roles on its lines. These have been marketed internally as labor‑augmentation rather than replacement. Outside observers describe Tesla’s deployments as highly instrumented, tightly constrained tasks inside factory‑defined cages. Integration and system partners Successful pilots universally require integration with PLCs, MES and WMS. Vendors are partnering with established system integrators and PLC suppliers (Siemens, Rockwell) to create safe, certifiable motion envelopes, and with WMS vendors for pick lists and inventory reconciliation. Healthcare and eldercare Healthcare deployments are cautious and focused on non‑clinical tasks: patient escorting, delivery of trays/meds, and basic assistance with activities of daily living (ADLs) under supervision. Sanctuary AI & Xiaomi (CyberOne) Sanctuary’s software‑first approach has been trialed in assisted‑living demonstration environments for patient escort and monitored social interaction. Xiaomi’s CyberOne has appeared in controlled demos for eldercare companion tasks in China. Operators emphasize robust fall‑detection, supervised autonomy, and strict privacy controls. Measured performance: safe escorting in corridor scenarios with blocking mitigation; ADL assistance remains constrained and usually supervised by a caregiver. Hotels, retail and hospitality Hotels and retailers are experimenting with concierge, room‑service delivery and in‑store assistance — leveraging humanoids’ anthropomorphic form for customer engagement. Pilot deployments emphasize novelty and PR value initially; operators report tangible savings only when robots handle repeated point‑to‑point deliveries (luggage, meals) rather than free‑form guest interactions. Safety and liability require heavy supervision; many chains prefer semi‑autonomous devices (autonomous carts on constrained routes) over fully mobile humanoids. Home and education Smaller consumer and education vendors, notably UBTECH and teams emerging from Unitree‑adjacent ecosystems, sell simplified humanoids for classrooms and labs. These are priced at consumer/edu levels (low five‑figures for advanced platforms) and are being used for coding education, telepresence and basic ADL trials in pilot families. Scale remains limited by cost, support, and clear ROI. Pilots, metrics and lessons learned Across sectors, a consistent set of metrics and lessons has emerged: Throughput vs. human: early pilots typically report 30–70% of human throughput in unconstrained tasks; parity appears only in constrained, repetitive lanes after process redesign. Uptime: pilot uptime commonly 60–90% depending on battery architecture, navigation robustness and integration maturity. Error rates: manipulatory errors (missed picks, dropped parts) remain the primary source of stoppages; software resets and localization drift are secondary. Sizing pilots for scale: integrators advise 6–12 month pilot windows, clear success criteria (throughput delta, uptime threshold, safety record), and staged rollouts with human fallback. Safety incidents: most reported incidents are minor (stumbles, dropped non‑hazardous parts). Operators emphasize soft‑body contacts, geofencing and redundant perception stacks as primary mitigations. Integrators and end users we interviewed stress that achievable gains are as much about process redesign and systems integration as robot capability. Humanoids succeed when plants change workflows to exploit their strengths rather than forcing robots into human‑exact tasks. Pricing and business models Vendors are avoiding one‑time hardware sales in favor of subscription/lease models that bundle maintenance, cloud updates and payload tooling. Indicative public pricing ranges observed in the market: Offer type Typical upfront Typical recurring Notes Lease + support Low‑to‑mid five‑figures per robot High four‑ to low five‑figures/year Common for logistics pilots CAPEX sale High five‑figures to six‑figures Optional support contracts Less common; preferred by large manufacturers Education/consumer Low five‑figures Occasional software subscriptions UBTECH style models Geography of deployments Deployments concentrate in North America, East Asia (China, Japan, South Korea), and selected EU advanced‑manufacturing clusters. North American pilots focus on warehouses and factories; East Asia shows the most eldercare demos and hospitality trials; Europe emphasizes safety regulation and integration in established automotive lines. Taxonomy: Viable use cases today vs. medium‑term prospects Viable now: box/tote handling in constrained lanes, kitting and parts transfer, supervised patient escort/delivery, hospitality deliveries on constrained routes, education/telepresence. Medium‑term (2–5 years): mixed‑sku picking parity, unsupervised ADL assistance in eldercare, high‑speed collaborative assembly with certified safety fences, multi‑robot coordinated logistics. Scorecard: who is closest to repeatable commercial deployments Company Primary sector Repeatability score (1–5) Notes Agility Robotics Logistics 4 Multiple pilots, clear SKU‑constrained wins, active integrator ecosystem Figure AI Manufacturing 3 Promising assembly trials; needs deeper PLC/MES integration Apptronik Manufacturing/logistics 3 Mobile manipulation strength, still pilot‑stage Sanctuary AI Healthcare/eldercare 2 Demonstrations and demos; clinical adoption slow Xiaomi (CyberOne) Consumer/assisted living demos 2 High publicity; constrained demos in China Tesla (Optimus) Manufacturing support 2–3 Strong internal investment; external commercialization timeline uncertain UBTECH / education vendors Education/consumer 3 Repeatable small‑scale deployments in schools and labs Conclusion: 2025–26 is not the year of mass humanoid substitution, but it is when repeatable, revenue‑generating use cases emerge — primarily in constrained logistics lanes and carefully instrumented manufacturing cells. The next phase will hinge on improved reliability, standardized system integration, and business models that reduce upfront risk for adopters.