More than 20,000 humanoid robots were produced worldwide in 2025. Only about 10% entered real-world applications, according to industrial research firm Interact Analysis.
Most of the remaining machines went into research, data collection and entertainment. Even among those deployed commercially, many were still operating in small-scale pilot programs rather than established, economically justified roles.
Production had increased roughly tenfold in a year. Commercial adoption had not kept pace.
BMW Proved the Technology. The Economics Remain Less Clear.
One of the strongest demonstrations of humanoid robotics in an industrial setting came from Figure AI's deployment at BMW's Spartanburg factory.
In 2025, Figure 02 robots handled more than 90,000 sheet-metal components over approximately 1,250 operating hours, contributing to the production of more than 30,000 BMW X3 vehicles.
The deployment was meaningful: these were real production tasks, not staged demonstrations.
But Figure's published performance criteria reveal an important distinction between proving capability and establishing commercial viability.
The company identified three operational targets:
An 84-second total production cycle, including 37 seconds for loading.
More than 99% successful component placement per shift.
Zero human interventions per shift.
Figure disclosed these targets alongside its cumulative deployment results, but did not publish corresponding achieved rates for all three metrics or the total cost per completed task.
That missing information matters to manufacturers. A robot's technical capability is only valuable relative to the speed, reliability and cost of the process it replaces.
There was another revealing development: following the deployment, Figure began retiring the Figure 02 platform in favor of Figure 03.
Figure identified the forearm as its most problematic hardware subsystem during the BMW deployment and redesigned the wrist electronics for the next generation.
Rapid hardware iteration is normal in an emerging technology. But it complicates the economics of industrial equipment, where customers must account for maintenance, replacement parts and the useful life of their investments.
BMW has outlined plans to test the successor model on more complex logistics workflows. The question is whether those improvements can eventually deliver a competitive cost per operation over years of service.
The Hidden Cost of Human Intervention
Agility Robotics offers an important counterexample to the idea that humanoids remain confined to research laboratories.
Its Digit robots have been deployed commercially at GXO Logistics under a multiyear Robots-as-a-Service (RaaS) agreement since 2024.
At GXO's facility near Atlanta, Digit has reached a cumulative 100,000-tote milestone. Agility reported approximately 98% accuracy while the robots were performing assigned tasks.
These are tangible operating results.
But 98% on-task accuracy is not the same as 98% autonomous workforce replacement.
A comprehensive economic assessment would also need to capture how frequently robots require human intervention, their productive utilization, maintenance costs and the number of tasks completed per paid operating hour.
This becomes especially important when teleoperation or technical supervision remains necessary.
The International Federation of Robotics noted in its September 2026 industry report that most humanoid applications remain specialized and often require human teleoperation.
Remote assistance does not necessarily destroy the business case. One technician supervising a large fleet could be economically efficient. One technician continuously assisting a single robot could make the same deployment unattractive.
The ratio matters more than whether a robot can operate autonomously for an impressive demonstration.
RaaS reduces the customer's upfront capital commitment, but it does not eliminate these costs. It transfers more of the hardware, maintenance and utilization risk to the robotics provider.
General-Purpose Capability Has Yet to Prove Its Premium
Humanoid developers frequently emphasize their robots' potential to operate across multiple tasks in environments designed for humans.
That flexibility could offer significant value. It is also difficult to price before customers can demonstrate they need it.
At GXO, Digit has primarily performed tote-handling operations, including transferring containers between mobile robots, conveyors and other locations. At BMW, Figure 02 initially focused on positioning sheet-metal parts.
Both are legitimate automation applications. Neither, by itself, establishes the economics of a broadly adaptable robotic worker.
For an individual task, a humanoid must compete against existing automation—not just human labor.
Fixed robotic arms, conveyor systems and autonomous mobile robots often solve narrower problems with simpler hardware and mature operating procedures.
Photo: M P Hennessey / Wikimedia Commons, CC BY-SA 4.0.
A humanoid's additional mobility and manipulation capabilities justify their cost only when they deliver savings that specialized alternatives cannot, such as avoiding substantial facility modifications or switching economically between different jobs.
That creates a difficult commercialization requirement: flexibility must translate into measurable utilization, not merely a longer list of possible demonstrations.
Why More Orders Are Not Yet Proof of Scale
Agility reported more than $300 million in multiyear orders for Digit 5 as of May 2026, subject to contractual milestones.
It is significant evidence of customer interest. But it is not equivalent to recognized revenue or demonstrated customer returns.
The company expects early access to Digit 5 in the first half of 2027, with general availability targeted for the end of that year.
The new generation is designed to address problems identified in earlier deployments, including safer operation around people, longer effective working time and expanded manipulation capabilities.
These improvements are commercially relevant because the cost of deploying humanoids extends beyond the robots themselves.
BMW's pilot required revised safety arrangements, including barriers and partitions, as well as improved connectivity on the factory floor. Similar integration requirements can erode the advantage of using machines supposedly designed to operate in existing human environments.
As deployment volumes increase, manufacturers will need to show that installation, supervision and maintenance requirements decrease rather than scale proportionally with each additional robot.
The Next Commercial Milestone
The most revealing evidence will not be another shipment target or an announcement about production capacity.
It will be customers expanding paid deployments beyond their initial pilot sites, supported by operating data that demonstrates:
Sustained throughput and uptime across full production schedules.
Low human-intervention rates and predictable maintenance costs.
Competitive cost per successfully completed task, including integration and supervision.
The ability to perform multiple economically useful tasks without extensive retraining or hardware modification.
These measures would establish whether the additional complexity of a humanoid platform delivers a return that conventional automation cannot.
Until then, the industry's central commercialization challenge remains unresolved: companies have demonstrated that humanoid robots can do useful work. They have not yet demonstrated, at scale, that humanoids are the most economical way to do it.
References
Interact Analysis (June 2026). Humanoid Robot Production Surges Tenfold in 2025, but Commercial Deployments Remain Limited.
International Federation of Robotics (September 30, 2026). Global Sales of Professional Service Robots Surge 24% — World Robotics 2026.
BMW Group (September 21, 2026). Leipzig Debut: BMW Group Introduces Humanoid Robots.
Figure AI (November 19, 2025). Figure 02 Contributed to the Production of 30,000 Cars at BMW.
GXO Logistics (June 27, 2024). GXO Signs Industry-First Multi-Year Agreement with Agility Robotics.
Agility Robotics (September 15, 2026). Agility Unveils Digit 5 Humanoid Robot Built for Cooperatively Safe Work at Scale.
Image Credits
Cover: Boston Dynamics Atlas at Hyundai Motorstudio Goyang. Photo: Damian B Oh / Wikimedia Commons. CC BY-SA 4.0.
.jpg){kind=link}
Inline: OTTO 1500 autonomous mobile robot. Photo: M P Hennessey / Wikimedia Commons. CC BY-SA 4.0.
Both photographs have been proportionally resized and converted to WebP for web delivery. These photographic derivatives remain licensed under CC BY-SA 4.0; that license does not apply to the GlobalFuse article text.