Humanoid robots are trading trade-show stages for paychecks. Over the past year, startups including Figure AI, Agility Robotics and Apptronik, along with Tesla, have moved their AI-powered machines from demonstration videos into paid deployments at automotive plants and logistics warehouses across North America and Europe. The shift, driven by persistent labor shortages and rapid advances in machine learning, marks the first time humanoid robots have held commercial contracts rather than research appointments. Analysts say the industry has crossed from promise to early reality, and the pace is accelerating.
Why Humanoid Robots, and Why Now
For decades, humanoid machines remained laboratory curiosities, confined to university labs and staged product reveals. Two developments changed the equation. First, so-called foundation models — large AI systems trained on video, language and simulated data — gave robots the ability to understand natural-language commands and adapt to tasks they had never seen. Second, component costs fell as electric actuators and sensors matured, pushing per-unit prices toward industrial budgets.
Wall Street has taken notice. Goldman Sachs analysts raised their humanoid forecast in 2024, projecting global shipments of roughly 1.4 million units by 2035 in a market worth about $38 billion. Morgan Stanley goes further, estimating the addressable market could reach $5 trillion by 2050. Bank of America analysts have pointed to the second half of the decade as the period when deployments could reach meaningful scale.
From Pilot to Payroll
The deployments are no longer symbolic. Figure AI’s Figure 02 robots have worked at BMW’s plant in Spartanburg, South Carolina, inserting sheet-metal parts in the chassis shop — tasks the company says are difficult to staff. Agility Robotics signed what it describes as the first commercial humanoid deployment contract with logistics operator GXO, placing its Digit robots in a warehouse handling flow-through work. Amazon has tested the same Digit model inside its own fulfillment network.
Tesla, meanwhile, is betting on scale. Chief Executive Elon Musk has said the company will produce Optimus units for internal use first, with external sales to follow, and has floated price targets as low as $20,000 to $30,000. Independent estimates put today’s humanoid prices closer to $150,000 to $250,000 per unit, a gap that illustrates how far the economics still must travel.
The Machine Learning Behind the Machines
What separates this generation from its predecessors is software. Rather than programming each motion, engineers now train vision-language-action models that map camera images and verbal instructions directly to motor commands. NVIDIA entered the field in March 2024 with Project GR00T, a foundation model for humanoid locomotion and manipulation, alongside simulation tools designed to train robots in virtual factories before they touch physical ones.
Figure announced its own Helix model in early 2025, saying a single neural network can control an entire robot, integrating vision, language and grasping. Rivals such as Physical Intelligence and Google DeepMind are pursuing similar generalist approaches. Data remains the bottleneck: most systems still learn from teleoperation, in which humans remotely pilot robots, and from simulation — methods that scale slowly and expensively.
Money Follows the Machines
Investors have poured in. Figure AI raised $675 million in February 2024 at a $2.6 billion valuation, with backing from Microsoft, NVIDIA, the OpenAI Startup Fund and Jeff Bezos. Physical Intelligence collected roughly $400 million later that year to pursue robot foundation models. Apptronik, maker of the Apollo humanoid, secured more than $350 million in early 2025, while Norway-based 1X Technologies is preparing a home robot for consumers.
Not everyone is convinced. Veteran roboticist Rodney Brooks, co-founder of iRobot, has publicly predicted that humanoid robots will take far longer to reach broad usefulness than their backers claim. Researchers at MIT and elsewhere note that human hands remain unmatched in fine dexterity, and that today’s robots still fail unpredictably at tasks children perform easily.
What It Means for Workers and Industry
The near-term impact falls on logistics and manufacturing, sectors that already struggle to hire. A 2024 study by Deloitte and the National Association of Manufacturers projected that 1.9 million U.S. manufacturing jobs could go unfilled by 2033, a shortage robot makers cite as their core market opportunity.
Economists caution that automation historically reshapes work rather than erasing it wholesale, but the transition can be painful for specific roles. Labor groups, including the United Auto Workers, have called for retraining guarantees as new technology spreads. Safety standards also lag: dedicated specifications for humanoids working near people are only now being drafted by industry bodies.
What to Watch Next
The next 18 months should reveal whether the economics hold. Watch for Tesla’s production claims on Optimus, Figure’s promised second commercial customer, and whether Agility can convert its first contract into multi-site rollouts. Unit costs, uptime and task reliability — not viral videos — will decide the sector’s fate.
National policy is also coming into focus. China has designated humanoid robotics a strategic industry, while U.S. and European regulators weigh standards and funding. If deployments compound as analysts expect, the question of a humanoid workforce will move from engineering blogs to boardrooms and bargaining tables well before the decade ends.