Humanoid robots are stepping off the demo stage and onto factory floors. In the past year, BMW, Mercedes-Benz, Amazon and logistics firm GXO have all launched pilots that put two-legged, AI-powered machines to work alongside human employees — a shift analysts describe as the most significant development in industrial robotics in decades. The move is driven by breakthroughs in machine learning that let robots understand language, see their surroundings and learn tasks on the fly.
Why Humanoids, and Why Now
Industrial robots have run assembly lines since the 1960s, but they perform fixed, repetitive tasks inside safety cages. They cannot navigate a warehouse built for people or handle the sheer variety of objects a human worker manages every shift.
That limitation is eroding fast. A new class of vision-language-action models, adapted from the large language models behind chatbots, allows robots to follow natural-language instructions, recognize unfamiliar objects and pick up new tasks from demonstration rather than explicit programming.
NVIDIA crystallized the trend in March 2024 with Project GR00T, a foundation model designed to give humanoid robots general-purpose skills. CEO Jensen Huang has repeatedly told developers that robotics represents the next wave of AI, a theme he now brands as physical AI.
From Pilot Programs to Payroll
Deployment is no longer hypothetical. Agility Robotics’ Digit became the first humanoid formally put to work for a customer when Amazon began testing it in warehouse roles, and GXO announced what it called the industry’s first commercial deployment at a Connecticut logistics site in 2024.
Figure AI placed its robots at a BMW plant in South Carolina, while Apptronik’s Apollo began trials with Mercedes-Benz and GXO. Tesla plans to use its Optimus robot inside its own factories first, with CEO Elon Musk projecting production in the thousands of units. Boston Dynamics, meanwhile, retired its hydraulic Atlas and unveiled an all-electric version built for commercial work.
Capital is following the machines. Figure AI raised $675 million in February 2024 from Microsoft, OpenAI, NVIDIA and Bezos Expeditions at a $2.6 billion valuation, then reportedly secured new funding in 2025 at roughly $39 billion. Agility Robotics opened a factory in Oregon with a stated capacity of 10,000 robots per year, and Apptronik closed a $350 million round backed by Google.
The Numbers Behind the Hype
The International Federation of Robotics counts about 4.3 million industrial robots operating worldwide, with roughly 540,000 new units installed in 2023 alone. Humanoids remain a rounding error in those figures — but forecasts suggest that will change.
Goldman Sachs, which once projected a $6 billion humanoid market by 2035, lifted its estimate to $38 billion after factoring in faster cost declines and broader use cases. Morgan Stanley goes further, estimating the market could reach $5 trillion by 2050 with close to a billion units in service.
Labor economics supply the demand. A Deloitte and Manufacturing Institute analysis projects U.S. manufacturers may need to fill 3.8 million jobs by 2033, with as many as half going unfilled. Warehouses face chronic turnover that can exceed 100 percent annually.
Experts Urge Caution on Timelines
Not everyone buys the speed. Robotics researchers note that today’s humanoids still struggle with fine manipulation, long-duration tasks and unpredictable environments, and that teleoperation — humans remotely guiding the machines — still backs many impressive demo videos.
Carnegie Mellon University robotics professor Christopher Atkeson has cautioned publicly that much of what circulates online is cherry-picked, urging observers to judge robots on reliability rather than spectacle. The gap between a viral clip and a dependable 24/7 workforce remains large.
Safety standards also lag the technology. No dedicated standard yet governs humanoids working in mixed human-robot spaces, though the Association for Advancing Automation has begun drafting guidance for the industry.
What It Means for Workers and Industry
For manufacturers, humanoids promise flexibility: a single machine that can be re-tasked overnight rather than a production line rebuilt over months. Early deployments target dull, dirty and dangerous jobs such as tote handling, machine tending and inspection.
For workers, the near-term effect looks more like redefinition than replacement. Companies piloting the robots consistently frame them as a response to vacancies they cannot fill, not as substitutes for existing staff.
For the industry itself, the race is consolidating around a handful of well-funded players, raising questions about whether smaller robotics startups can survive the capital intensity of hardware at scale.
What to Watch Next
The next 12 to 18 months will test whether humanoids can graduate from paid pilots to scaled fleets. Watch three signals: whether announced customers convert into repeat orders, whether reported uptime and mean time between failures improve, and whether unit costs fall toward the $30,000-to-$50,000 range many analysts consider necessary for mass adoption.
Regulators deserve equal attention. As humanoids begin sharing warehouses, hospitals and eventually sidewalks with the public, expect the first workplace-safety rulings and national pilot frameworks — decisions that will determine how quickly physical AI leaves the factory floor.