At China’s World Humanoid Robot Games last month, the technological progress of humanoid robots were on full display. Some shattered human records in 100-metre sprints, others flexed their biomimetic muscles in boxing, football, high jump, and even breakdancing.
The event highlighted the rise of machines and the underlying physical AI layer that enables them to perform certain tasks such as picking, assembly, machine handling, and warehouse work.
Meanwhile, in the US, NVIDIA, through its Isaac GR00T models, simulation tools, and robot hardware, is building the training and computing infrastructure for humanoid robots. Microsoft has partnered with Hexagon Robotics to power its industrial humanoid for manipulation and inspection in manufacturing and logistics; Meta is testing robots from vendors like Watney Robotics, Kinova, and ABB for physical tasks; and Amazon‘s robotic systems are optimising operations across fulfilment operations.
However, their increasing capabilities are now putting pressure on lower-skilled workers, many of whom are immigrants.
According to a MIT Economics study, introducing one additional industrial robot resulted in a reduction in employment equivalent to several workers and a drop in wages by about 0.4%.
Moreover, millions of workers are uncertain about future employment as businesses take interest in physical AI systems amid labour shortages and rising labour costs.
JP Morgan recently estimated that the US manufacturing sector has about 462,000 unfilled jobs, a figure that could rise to 1.6 million by 2030. Of them, humanoid robots could perform about 25% of those roles with current technology, potentially rising to 50% as the technology improves.
Automation at the Local Level
Long before humanoid robots and AI, machines have transformed labour-intensive work into automated processes, reducing the need for human workers while boosting productivity. USDA data show that hired farm employment fell from 2.33 million in 1950 to 1.13 million in 2000 as farms became increasingly mechanised.
However, for a low-skilled worker, moving to a different job is not always straightforward. A factory worker cannot automatically become a robotics technician, as they may not have the education or training required.
“But it could happen temporarily for companies that struggle financially as they look to cut costs and stay afloat. Every healthy company will not experience job loss. If anything, due to the need for specialised skill in maintenance and AI usage, more hiring will happen,” an Indian-origin immigrant in the US working in the construction sector tells AIM on the condition of anonymity.
Jobs Created May Not Be Jobs Lost
Automation does not mean fewer jobs, nor is it a threat to all jobs. Robotics might take over physical and labour-intensive jobs, but it is likely to create demand for engineers, technicians, programmers, maintenance workers, and other professionals.
Immigrants are heavily represented in several low-skilled occupations such as construction, transportation, routine work, and food delivery. The Migration Policy Institute (MPI) analysis in 2020 found that only 26% of immigrant-origin workers were employed in occupations projected to decline. This shows that automation will not completely eliminate jobs that are highly populated by immigrants; rather, it is likely to change the types of immigrant labour the US economy demands.
However, as political pushback against immigrants grows globally, with several OECD countries tightening migration policies, the entry of humanoid robots into the labour market could worsen the situation. The International Monetary Fund has warned that automation can lower labour demand and hiring in occupations where AI substitutes for human tasks.
However, replacing humans with humanoids on the factory floor is easier said than done.
Take China, the world’s largest robotics market. In 2024, the country installed about 295,000 industrial robots, accounting for 54% of global installations, while its operational stock crossed 2 million units. The next year, China grew its robotics prowess, accounting for more than 80% of the roughly 16,000 humanoid robots installed globally.
However, a Reuters investigation found that machines being developed for factory work still struggled with dexterity, adaptability, and autonomous operation. Some robots remain dependent on choreographed routines and can be slower and less adaptable than conventional robotic arms.
As technology evolves, the labour market may increasingly favour workers with specialised skills relevant to building, maintaining, and developing automated systems.
The Migration Policy Institute’s 2024 research found that 72% of US jobs are expected to require postsecondary education or training by 2032. Among college-educated immigrants in 2022, 13% worked in computer and mathematics occupations, compared to 6% of college-educated US-born workers. An increase in automation means less demand for some routine/manual labour and more demand for workers who can build, operate, maintain, and develop AI systems.
“The engineering profile for humanoid robotics deployment is not a robotics engineer in the traditional sense. It is not someone who programmes FANUC or KUKA arms,” said Charles Cureton, an automation talent specialist, wrote on LinkedIn.
“The humanoid robotics challenge is fundamentally different—autonomous operation in unstructured environments, safe interaction with untrained humans, physics-based constraints on AI-driven actions, real-time adaptation—to conditions that cannot be pre-programmed,” he observed.
A New Workforce Divide
The problem is the transition, creating a skill gap that could become one of the biggest economic challenges of automation. If companies adopt robots faster than workers can be retrained, this could result in uneven productivity. Highly-skilled workers could gain new opportunities while workers performing repetitive or predictable tasks could face job loss.
“Immigrants primarily bring technical, highly skilled labour and hard-working driven people in physical labour jobs like construction. Both cannot be substituted by AI. So I don't think it will affect the immigrants at all,” the immigrant notes.
Retraining workers is often seen as an obvious solution, but learning new technologies comes with significant challenges. The World Economic Forum estimates that 59% of workers will need reskilling or upskilling by 2030, while 11% may not have access to the training they need. Robotics training can also require specialised, hands-on education.
For workers, however, spending months or years acquiring these skills can mean tuition costs, time away from existing jobs and income, and the challenge of balancing training with family and financial commitments. That makes employer-funded upskilling critical to the transition.
Some companies are already ahead of the curve. Amazon’s Mechatronics and Robotics Apprenticeship combines classroom instruction with on-the-job training, with the company claiming that participants see an average 23% wage increase after classroom training and another 26% after on-the-job training.
The Missing Manufacturing Jobs
Who actually manufactures the machines? The manufacturing gap is becoming a heated economic issue in the physical AI and robotics ecosystem.
According to the Observer Research Foundation, much of the industry’s manufacturing and deployment capacity remains concentrated in Asia, despite the US having strong capabilities in software and robotics design. China has the largest number of robot installations, while Japan and South Korea are responsible for manufacturing and deployment.
According to Visual Capitalist data, the US manufactures a fragment of global humanoid robots, accounting for less than 5% of global shipments. In 2025 alone, US-based Figure AI, Agility Robotics, and Tesla roughly shipped 150 humanoids each. This is a fraction of 5,500 humanoids Unitree shipped in the same year.
The US’ robotics challenge then is not only about automation, but whether the US can create a manufacturing ecosystem and jobs needed to build those machines domestically.
Automation can eventually change labour demand in sectors populated by immigrants. The important question is whether American workers will be able to adapt quickly enough.
Your reaction