The question everyone is asking in 2026 is simple: are humanoid robots actually working in factories, or is this still science fiction? The answer is specific and documented: yes, they are working. Not everywhere. Not perfectly. But on real production lines, during real shifts, contributing to real vehicle manufacturing at BMW, handling real warehouse logistics for Amazon, and being deployed at automotive plants across three continents.
The International Federation of Robotics confirmed in 2026 that humanoids have crossed the threshold from prototype demonstrations to production deployment. This blog strips away the hype and documents exactly what is happening: which companies are deploying robots where, what tasks the robots are actually performing, how they are performing them, and what the honest limitations still are.
BMW Spartanburg: The First Verified Production Deployment
Figure 02 ran at BMW Group Plant Spartanburg for 11 months on an active assembly line, running every single working day. Within 6 months of bringing up Figure 02, robots were delivered to the plant and began testing. Within 10 months, full deployment launched on the line. The robots contributed to the production of over 30,000 BMW X3 vehicles, logged over 1,250 operational hours, and loaded more than 90,000 sheet-metal parts.
This is not a demo. This is shift work.
The task was sheet-metal loading — a classic pick-and-place operation in automotive manufacturing. An associate picks sheet-metal parts from racks or bins and places them on a welding fixture, after which six-axis industrial robots weld and feed the parts into the main production line. Figure 02 replaced the human in that first step, working 10-hour shifts Monday through Friday.
One learning that informed Figure 03 design was the robot's forearm, the top hardware failure point at BMW. The forearm is a challenging subsystem due to its tight packaging, dexterity requirements (three degrees of freedom), and thermal constraints. For Figure 03, the wrist electronics were completely re-architected, with each wrist motor controller now communicating directly with the main computer, reducing complexity, improving reliability, and simplifying thermal management.
This level of engineering transparency is unusual in the robotics industry and genuinely valuable. Companies don't typically publish detailed descriptions of their failure modes unless they are confident the next generation has addressed them. It also illustrates something critical about the state of humanoid deployment in 2026: the hardware is capable enough to run for months, but reliability engineering — identifying what breaks under sustained use and redesigning those components — is still happening in real time.
BMW Leipzig: Europe's First Humanoid Deployment
BMW is not stopping at Spartanburg. On February 27, 2026, BMW confirmed it is deploying humanoid robots at its plant in Leipzig, Germany, marking the first time Physical AI of this kind has entered a European automotive production environment. Following an initial theoretical evaluation phase and successful laboratory tests, there was an initial test deployment in December 2025. A further test deployment is planned from April 2026, ahead of the full pilot phase beginning in summer 2026.
The Leipzig deployment uses Hexagon's AEON humanoid rather than Figure's platform. The deployment in Leipzig is focusing on testing a multifunctional application of the robot. It is based on AEON's design, whose human-like body allows a wide range of hand and gripper elements or scanning tools to be flexibly attached and enables dynamic use on wheels.
The tasks targeted for Leipzig are broader than Spartanburg's single operation: battery production for energy modules and component manufacturing for exterior parts. BMW's approach is methodical — theoretical assessment, laboratory testing using real production use cases, first deployment under live conditions, then the pilot phase proper.
Artificial intelligence is already an integral part of the BMW Group's production system. A prerequisite for the effective use of artificial intelligence in production is a unified IT and data model across the entire production system. BMW has consistently transformed isolated data silos into a unified data platform, meaning all data is consistent, standardized, and available at all times.
That data infrastructure is what allows BMW to deploy humanoids at all. Without a unified platform where robot sensor data, production metrics, quality control data, and logistics data all flow into a common model, the feedback loops needed to train and refine robot performance would not exist.
Hyundai and Boston Dynamics: The Largest Commitment Yet
Hyundai Motor Group announced at CES 2026 that it is preparing to deploy tens of thousands of Boston Dynamics Atlas humanoid robots across its global manufacturing facilities, has committed its entire 2026 Atlas production run exclusively to internal deployment at its Robotics Metaplant Application Center and to Google DeepMind, and is investing $26 billion in US operations that includes a new robotics factory targeting 30,000 Atlas units per year by 2028.
The scale of Hyundai's commitment is unprecedented. This is not a pilot. This is industrial infrastructure. Boston Dynamics develops the platform, Hyundai Mobis manufactures the actuators, Hyundai Motor Group operates the factories where the robots work, and the Robotics Metaplant Application Center in the US serves as the supervised training environment where Atlas learns manufacturing tasks before being deployed at scale.
Boston Dynamics unveiled the production version of Atlas at CES 2026 — a fully electric, enterprise-grade humanoid robot now entering manufacturing. All 2026 units are committed to Hyundai's Robotics Metaplant Application Center and Google DeepMind, with broader customer orders opening in 2027.
The vertical integration is the structural fact that makes this different from every other humanoid announcement. When Tesla talks about deploying Optimus in its factories, it is describing a scenario where a single company is both robot maker and robot operator. Hyundai has constructed the same closed-loop feedback system through acquisition and subsidiary relationships: the company that builds the robot is the company that operates the factories where the robot works, which is the company that manufactures the actuators that power the robot.
Amazon and Agility Robotics: The Warehouse Leader
Agility Robotics' Digit continues to lead real-world humanoid deployments. In February 2026, Agility signed a Robots-as-a-Service agreement with Toyota Motor Manufacturing Canada following a successful pilot. Seven or more commercial units are now active supporting RAV4 logistics at the Woodstock plant.
Agility's Digit has been operating in warehouse logistics longer than any other humanoid platform. Figure 02 works at BMW's South Carolina plant; Agility Robotics' Digit operates in GXO's Georgia warehouse as of 2024. These are real deployments, not demonstrations.
Agility CEO Peggy Johnson recently confirmed that the company's Salem, Oregon RoboFab facility maintains an annual capacity of 10,000 units. Johnson's vision for late 2026 includes a safety-certified Digit that can break out of the cage, moving material across facilities alongside human staff without protective barriers.
The "break out of the cage" phrasing is significant. Traditional industrial robots operate inside safety cages precisely because they are not designed to operate around humans — they move too fast, with too much force, and with no capacity to detect human presence. A safety-certified humanoid that can work on an open warehouse floor alongside human workers represents a fundamentally different deployment model: the robot goes where the work is, rather than the work being brought to the robot.
Mercedes-Benz and Apollo: The German Pilot
At Mercedes' Digital Factory Campus in Berlin and its plant in Kecskemét, Hungary, Apollo handles moving components between stations, logistics shuttling, and quality-check routines for certain parts. The robots learn tasks through teleoperation first, then repeat them autonomously once trained. Roll-out is cautious. Mercedes still treats Apollo as a way to test automation of repetitive, physically taxing tasks, especially in areas with labor shortages, rather than a wholesale replacement of workers.
Mercedes's approach reflects a specific deployment philosophy: humanoids as a solution to labor shortages and ergonomic challenges, not as a cost-reduction play that displaces existing workers. This framing matters for worker acceptance. If the introduction of robots is positioned as eliminating the most physically demanding tasks that are already difficult to staff, resistance is lower than if it is positioned as headcount reduction.
Tesla Optimus: The Most Ambitious Internal Bet
Tesla aims to install first Optimus production lines and build 5,000 to 10,000 robots in 2025, scaling to 50,000 in 2026 and eventually one million units per year. Price target is around $20,000 per unit, with CEO Elon Musk saying Optimus and autonomy will make up most of Tesla's future value. Current lines are R&D lines, with a completely different scaled production line still to come, which implies the mass-production process is not locked in yet.
Tesla Optimus Gen 3 with redesigned hands and autonomous capabilities is in final stages ahead of production ramp, with all deployments remaining internal to Tesla factories through 2026 and external deployment planned for later.
Tesla's Optimus program is the largest gap between stated ambition and documented external deployment. Tesla Optimus is in internal factory testing with external deployment planned for later in 2026, but no public sales, no pre-orders, and no confirmed shipping dates for consumers exist as of early 2026.
The honest picture of Tesla Optimus in 2026 is: an internal research and development program converting portions of the Fremont factory to robot production, targeting deployment within Tesla's own operations, with commercial availability to external customers still a year away at minimum.
What the Robots Are Actually Doing — And What They're Not
The deployment timeline above organizes the verified use cases. Here is what matters about that list: the tasks are narrow, structured, and repetitive. The pattern is consistent: real work in 2025 lives in logistics, simple handling, and repetitive motions on the factory floor, not in high-precision assembly or human-dense public spaces.
Sheet-metal loading at BMW. Material handling at Amazon. Component logistics at Mercedes. Battery module assembly testing at BMW Leipzig. These are not high-dexterity tasks requiring sub-millimeter precision. They are not tasks requiring real-time judgment about variable part quality or unexpected assembly problems. They are tasks with consistent inputs, consistent motions, and measurable output quality.
Proven tasks include material handling (moving boxes, totes), basic assembly (inserting parts), visual inspection, beverage service, and walking on flat surfaces. Tasks must occur in structured environments with predictable layouts and lighting.
The tasks humanoids are not doing are equally revealing. No final assembly of complex components requiring tactile feedback and fine motor control. No real-time diagnosis and repair of equipment failures. No quality troubleshooting that requires interpreting visual defects in variable lighting. No operation of tools designed for human hands in unpredictable orientations.
The shared constraint across all of these efforts is not hardware capability, where multiple platforms have now demonstrated credible industrial manipulation, but AI reliability: the ability to complete a manipulation task correctly, safely, and consistently across thousands of repetitions in uncontrolled environments with variable part positioning, lighting conditions, and human co-workers.
The Hardware-vs-Software Bottleneck
While the hardware is increasingly mature, the software brains are still catching up. Most live deployments — like Digit's work at GXO Logistics or Figure's pilot at BMW — focus on repetitive material handling. The dream of a general-purpose machine that can walk into any home or warehouse and begin working zero-shot is being chased by a specialized layer of AI startups.
The AI constraint is the binding one. Actuators are cheaper, lighter, and more reliable than they were two years ago. Battery energy density has improved. Motion planning for bipedal locomotion on flat surfaces is a solved problem. What remains unsolved is generalization: the ability to encounter a novel object, in a novel pose, under novel lighting, and manipulate it correctly on the first attempt without human teleoperation or extensive retraining.
By running larger fleets for longer durations, firms can encounter and resolve invisible failures that only manifest after thousands of cumulative operating hours — the long tail of edge cases that laboratory prototypes or simulation simply cannot replicate. In theory, this strategy creates a data flywheel, where real-world deployments provide the diverse data needed to harden systems for messy, unstructured environments.
This explains why every major humanoid company is racing to deploy as many robots as possible in as many environments as possible: they need the data. A robot that has handled 10,000 parts in 10,000 different poses builds a dataset that cannot be replicated in any lab or simulator. That dataset is what trains the next model, which performs better, which deploys to more sites, which generates more data. The companies that get this flywheel spinning first have a compounding advantage.
The 2026 Reality Check
Expect several hundred humanoid robots deployed industrially in 2025, scaling to low thousands by 2026–2027, concentrating in automotive manufacturing (Tesla, BMW, Mercedes-Benz, Chinese EV makers), logistics and warehousing (Amazon, GXO, third-party logistics providers), and specialized service roles.
Given the announcements and investments on record, 2026 will not be the year every household buys an android. It will be the year several bets are tested at larger scale. Agility Robotics intends to ramp production at its RoboFab facility from hundreds of Digit units towards thousands, while deepening deployments in logistics centers. Apptronik and Mercedes plan to expand Apollo testing beyond initial German and Hungarian plants if pilot KPIs hold, and other carmakers are watching closely.
The market is real. The deployments are documented. The tasks are narrow but economically significant. What is not happening in 2026 is the general-purpose humanoid assistant that handles arbitrary tasks in arbitrary environments with human-level dexterity. That capability remains years away, and anyone claiming otherwise is either misinformed or selling something.
The Bottom Line
The term manless factory often circulates when discussing the future of the automotive industry. Currently in focus: humanoid robots. Hyundai announced at CES 2026 that it will gradually deploy humanoid robots in real factory environments and qualify them for series processes there. The goal is not only to demonstrate robotics but to integrate them on a large scale into existing production systems.
Humanoid robots are working in real factories in 2026. That sentence is factually accurate. The follow-up sentence is equally important: they are working on a small number of narrowly defined tasks, in structured environments, at a scale measured in dozens to low hundreds of units per deployment, with active human oversight and engineering support.
The gap between that reality and the science fiction vision of fully autonomous factories is still substantial. But the gap is closing measurably, deployment by deployment, shift by shift, part by part.
The companies that are documenting what works, publishing what fails, and iterating in public — BMW, Figure AI, Agility, Hyundai — are building the data and the credibility that will define the industry for the next decade. The robots are here. The question is no longer if they work. It is how fast they improve.