Humanoid robots are leaving the demo floor. In 2026 they started showing up on actual production lines and in distribution centers, doing the kind of repetitive, physically demanding work that factories have struggled to staff. The question now is whether they can keep doing it for a full shift.
The strongest example so far comes from BMW. In June 2026, Figure 03 completed an automotive sequencing workflow at BMW Group Plant Spartanburg, selecting parts, manipulating a heavy cart, and adjusting its position as objects moved or arrived in different orientations. That is the kind of task that looks simple until a robot has to do it without supervision.
The timing matters because the bottleneck in automation has shifted. Fixed robotic arms have handled repetitive work for decades, but they need a redesigned cell around them. A humanoid can, in theory, walk up to an existing workstation, use the same shelves and tools, and leave the line intact. According to the Global Trade Magazine analysis of humanoid deployments, that is the design's real selling point: working in spaces built for people without rebuilding them.
The numbers behind the push are large. Figure produced more than 350 third-generation humanoids by April 2026 and improved its production cycle from one robot per day to one robot per hour. Apptronik announced a $520 million extension to its Series A in February 2026, bringing the total above $935 million. Figure closed a Series C above $1 billion in September 2025. Mercedes-Benz put a low double-digit-million-euro investment into Apptronik while testing Apollo robots for component transport and quality checks.
Boston Dynamics introduced the product version of its electric Atlas in January 2026 and began manufacturing it, with initial deployments at Hyundai's Robotics Metaplant Application Center and Google DeepMind. Schaeffler opened a 7,500-square-meter humanoid robotics facility in Taicang, China in October 2026, installing six humanoid robots across five stations to sort, assemble, inspect, label and package components used in other humanoid robots.
None of this guarantees the robots will hold up. The biggest open question is reliability. Industrial buyers need predictable cycle times, high task-completion rates, and limited human intervention. A robot that performs well in a controlled demonstration can still struggle with shifted objects, damaged packaging, reflective surfaces, or changing floor conditions.
Where the Robots Are Actually Working
Automotive plants are the first real test bed. They have established automation budgets, structured processes, and measurable production targets. Humanoids there handle part transport, loading components into fixtures, replenishing workstations, and initial quality inspections. BMW has tested them for sheet-metal loading and parts sequencing. Mercedes-Benz is evaluating Apollo for repetitive intralogistics work. These deployments are task-specific, not general-purpose.
Warehousing is the second front. In May 2026, Figure announced an agreement to deploy humanoids at a distribution and logistics center in Reno, Nevada, focused on physically demanding supply-chain tasks. Distribution centers contain endless repetitive movement of products between pallets, containers, shelves, and conveyor systems. Whether humanoids beat specialized equipment depends on whether they can work with existing containers and shelving without infrastructure changes.
Schaeffler plans to deploy at least 1,000 AEON humanoids across its global production network over the next seven years. That is the kind of commitment that only makes sense if the per-unit cost and uptime are already trending toward viable.
The Money Behind the Machines
Capital is moving past prototypes. Apptronik's $520 million extension is being used to expand Apollo production, establish robot-training facilities, and support deployments in manufacturing, retail, and logistics. Figure's funding is supporting its Helix AI platform and BotQ manufacturing operations. Industrial users are becoming strategic investors rather than just customers.
The investment is also flowing into the supply chain around the robots: actuators, gearboxes, sensors, batteries, robotic hands, AI processors, and training infrastructure like simulation platforms and digital twins. A supporting market is forming, which is usually a sign that buyers expect this to be a real category rather than a science project.
What Could Slow It Down
Battery life is a real constraint. Agility Robotics selected an ultra-fast charging system for Digit 5 designed to support more than 20 working hours within a 24-hour period, and longer availability would make humanoids more suitable for multi-shift operations. But walking, lifting, balancing, and processing visual information all consume energy, and charging has to be planned without creating production delays.
Cost is the other barrier. The purchase price is only one part of the expense. Integration, training, safety systems, maintenance, spare parts, software subscriptions, and teleoperation support all count toward total cost of ownership. Robot-as-a-service and leasing models may help by letting customers evaluate performance before a large hardware purchase, but they also put maintenance and software-update responsibility on the provider.
Safety evaluation is its own problem. Industrial humanoids must be assessed as complete systems, including hardware, software, tools, workspaces, and human interaction. Current robotics standards cover safe robot design, system integration, legged locomotion, and service-robot operation, but individual deployments still require site-specific risk assessments.
The Comparison That Matters
The honest comparison is not humanoid versus human. It is humanoid versus fixed robotic arm, cobot, or autonomous mobile robot. If the process does not require a human-like form, conventional automation often offers lower cost and higher reliability. Humanoids make sense where the workspace is already built for people and retrofitting a fixed cell would be expensive.
That is why the advice from deployment analysts is to select the task before selecting the robot. The best starting points are repetitive, measurable, physically demanding activities with clear operating boundaries. Establish measurable pilot targets: cycle time, task-completion rate, human interventions, uptime, energy use, safety events, and maintenance requirements.
Most humanoids remain at the prototype or pilot stage, and conventional industrial robots are still economically stronger for many repetitive production tasks. Wider adoption will require improvements in reliability, safety, operating time, and total cost. The transition will be gradual, with adoption determined by proven operating results rather than demonstrations.
What to Watch Next
The metric to track is not how many robots get announced but how many complete a full shift without human intervention. Figure's Reno deployment and Schaeffler's 1,000-unit plan are the kind of tests that will produce that number. If uptime holds, the category moves from interesting to essential. If it doesn't, the money keeps flowing but the warehouses keep using what they already have.
For broader context on deployment trends, see Global Trade Magazine's humanoid robotics coverage. For related coverage on this site, visit our robotics and drones category.