Humanoid robotics: beyond the viral videos
Actualizado: 2026-05-03
For three years, videos of humanoid robots doing backflips or opening doors dominated social networks. The winning narrative was that we were months away from seeing humanoid fleets in factories and warehouses. Reality arrived more slowly, but it arrived. During 2025 and Q1 2026, several makers finally crossed the line between demo and verifiable productive deployment.
Key takeaways
- Figure AI, Agility Robotics, Apptronik, and Unitree have moved from video to signed contracts with verifiable industrial clients.
- Tasks that work: standard container handling, press loading, defined-route visual inspection, material transfer between nearby stations.
- Real per-unit cost on direct purchase ranges 150,000–300,000 dollars, plus 15–25% annual maintenance.
- The Robot-as-a-Service model (3,000–10,000 $/month) accelerates adoption by converting capex to opex.
- Effective availability in the most successful deployments: 70–85% of the time, achievable only with well-bounded tasks.
The actors that left the lab
Four companies concentrate most verifiable real deployments.
Figure AI has Figure 02 deployed at a BMW plant in Spartanburg (South Carolina) since Q3 2024, and since 2025 also at a non-publicly-identified German auto maker. Tasks are concrete and limited: sheet-metal handling in pre-assembly, specific container load/unload, repetitive positioning work.
Agility Robotics has Digit working in GXO logistics centers since late 2024 and at a Spanx plant since 2025. Its specialty is moving standard containers between conveyors and shelves, benefiting from Digit being purpose-built for that context rather than a general humanoid platform.
Apptronik, backed by Mercedes-Benz and Google, has Apollo at Mercedes Hungary since 2025 and at an SKT plant since early 2026. Apollo positions as a human collaborator: works alongside operators on inspection and parts-transfer. Contract model is leasing more than direct purchase.
Unitree, from China, followed a different route selling G1 and H1 to universities, integrators, and some industrial clients at significantly lower prices (20,000 to 70,000 dollars depending on config). For mobility and inspection tasks performance is sufficient and price opens markets US companies don’t touch.
Which tasks actually fit
Three characteristics make a task a reasonable humanoid candidate:
- Predictable structure: stable environment, known objects, bounded variability.
- Duration: tasks filling full hours, not short interactions where activation cost doesn’t amortize.
- Failure tolerance: systems where a robot error doesn’t break critical chain nor endanger nearby people.
Tasks meeting these three conditions in real plants: – Standard container handling in logistics. – Loading parts into presses and machine tools. – Visual inspection with defined routes. – Material transfer between nearby stations.
Tasks that don’t fit: variable fine assembly, ad-hoc repair, customer interaction, and anything combining non-trivial human judgment with physical execution.
Real cost and return
Declared cost for Figure, Agility, and Apptronik industrial robots ranges 150,000 to 300,000 dollars per unit on direct purchase, plus annual maintenance estimated at 15–25% of purchase price. Real client cost is higher because there’s specific integration, staff training, and production-flow adaptation paid separately.
The leasing model (Robot-as-a-Service) changes the math: monthly fees between 3,000 and 10,000 dollars per unit with maintenance included, converting capex to opex. This model accelerated adoption among clients unwilling to accept fast-obsolescence risk.
ROI depends on task type. In logistics with high rotation and long shifts, recovering investment in 18–30 months is viable with current numbers. In manufacturing with varied tasks, return takes longer and often depends on factors not known until the robot is on-site for months.
The data point many companies discover late is that amortization depends less on hourly robot throughput and more on reliability: how many hours per month it runs without human intervention.
Where humans remain unbeatable
Despite enthusiasm, humans maintain clear advantage in four dimensions:
- Fine manual dexterity, especially with soft or deformable objects (cloth, cables, unpackaged food): still out of current humanoid robotics’ reach by a considerable margin.
- Adaptation to unexpected changes: when a belt jams in an unusual way, the robot asks for help, the human improvises.
- Judgment on imperfect quality: deciding whether a marginal part passes or fails in an edge case is still better with an expert operator.
- Interaction with people: a humanoid robot on-site still requires strict safety protocols and limited interaction zones.
What will change soon
Three lines will move in the next 18 months per conversations with integrators:
- Foundation models for motor control that several makers are training will enable behavior transfer between tasks without reprogramming, lowering commissioning cost.
- Fine manipulation will improve with new-generation hands already in testing phase.
- Per-unit cost will drop as volume production grows, likely below 100,000 dollars for midrange models by 2027.
What probably won’t change is the problem’s nature: the general humanoid replacing the operator across all tasks remains distant because it’s not just hardware or software.
When it pays off to enter
For a company considering whether to enter humanoid robotics, three questions filter the decision:
- Are there concrete tasks with sufficient volume (minimum one full shift per day) and repeatable structure?
- Is physical integration possible without full line redesign or construction work?
- Does the operations department have technical capacity to absorb the robot as a system requiring management, not passive machinery?
If the three answers are yes, it’s worth engaging with makers that have similar cases and requesting verifiable references. If any answer is no, it’s probably too early. The cost of entering too early is high and cases ending well are few.
The general lesson from the first real enterprise humanoid robotics cycle is consistent with previous tech waves: it works where well-bounded and where there’s discipline to measure outcome.
