When AI meets the physical world: Robotics

Jay Jacobs Nov 20, 2026 Equity

Key takeaways

  • Structural forces are driving demand for robotics. Aging populations, operational challenges, and workplace safety needs could accelerate robotics adoption.
  • Physical AI is expanding what robots can do and where they operate. We believe robots will be moving beyond factories into logistics, transportation, humanoid applications, and other industries.
  • The investment opportunity extends beyond the finished robot. Investors may explore opportunities across the robotics value chain, including semiconductors, enabling technologies and components, via thematic ETFs such as the iShares Future Robotics and Related Technologies Active ETF (RBTX).

For decades, Hollywood has asked audiences to imagine a world of robots. The Jetsons gave us Rosie, the friendly household helper who seemed to do everything. The Terminator offered a less reassuring vision of intelligent machines. And Blade Runner asked what happens when robots become almost indistinguishable from humans.

Today, large language models have made extraordinary strides, bringing us closer to the intelligence Hollywood envisioned. But teaching machines to move with the same ease and precision as humans presents an entirely different challenge.

Consider the humble coffee mug. Roughly two-thirds of American adults drink coffee as part of their daily routine.1 Reaching for a cup and taking a sip mid-conversation takes almost no thought. Teaching a robot to perform that same seemingly effortless task is far more complicated than it looks. It must gather sensory data, recognize the mug, judge the distance, and apply the right force, all while avoiding other humans or objects nearby. Change the lighting, move the mug, or replace it with a less commonly shaped one, and the problem changes again.

Here’s the plot twist: just as machines are learning to navigate more of the physical world, the world is finding more reasons to put them to work.

What’s driving robotic adoption

Factories can’t always find the workers they need, particularly for repetitive, physically demanding jobs. With populations aging across major manufacturing economies, finding those workers could become even harder. Meanwhile, companies looking to diversify supply chains or bring production closer to customers often face higher costs. Robots aren’t a cure-all for labor shortages, but they could take on tasks involving heavy lifting, sharp objects, or hazardous environments. In doing so, they could help businesses improve workplace safety and make labor more productive.

Robotics is hardly starting from zero. The International Federation of Robotics counted 603,000 new industrial robot installations worldwide in 2025, bringing the operating fleet to roughly 5.1 million.2 The more surprising part may be how unevenly those machines are distributed. A look at robot density reveals just how much. In 2024, South Korea had 1,220 industrial robots per 10,000 manufacturing employees, compared with just 307 in the United States.3 So, while the robotics revolution may well be underway, not everyone is moving at the same speed. As technology improves and the economics become more compelling, robotics adoption could broaden further.

Robot density in manufacturing, 2024

Industrial robots per 10,000 manufacturing employees

Source: International Federation of Robots, “Robot density surges in Europe, Asia, and Americas,” 4/8/2026. Data is based on industrial robots, not humanoids.

Chart description: Horizontal bar chart comparing industrial robot density across ten countries in 2024, measured as robots per 10,000 manufacturing employees.


How is physical AI changing robotics?

A robot that can dance, do backflips, or run a half marathon may grab headlines. But the real breakthroughs could be far less glamorous, like a robot that reliably handles dangerous materials. While traditional industrial robots are great at repeating the same task thousands of times, physical AI aims to give machines more flexibility: the ability to see, interpret, and respond to their surroundings rather than follow a specific script.

So, what happens when you take a robot trained in a virtual world and throw it into a real one? Robots can practice thousands of times in a simulation without breaking a single coffee mug. Put them in a real warehouse, where packages tumble and people walk unexpectedly into view, and things get complicated. Engineers call this the “sim-to-real gap.”4 Better sensors, AI models, simulation, and real-world training are helping bridge the divide.

Despite those challenges, some robots are already earning their keep. At logistics provider DHL, a purpose-built robot named Stretch can unload cartons from shipping containers at rates of up to 700 cases per hour.5 DHL reports that the technology reduces physically demanding work in hot or cold trailers and has announced plans to expand deployments. No backflips required. Just a practical, demanding job, executed by a robot.

Where are robots being used today?

The story doesn’t end at the loading dock. Self-driving cars, once another staple of science fiction, are increasingly becoming part of everyday transportation. Waymo reported 15 million fully autonomous rides in 2025, more than triple its 2024 volume, with plans to expand into additional cities.6

While a self-car may look nothing like a warehouse robot, both rely on sensors, AI computing, and software to perceive their surroundings and respond to changing conditions. Different machines. Different missions. But many of the same technologies.

What will it take for humanoid robots to scale?

And yet, for all the places robotics could take us, there’s one machine that continues to capture our imagination: the humanoid robot. Built to operate in spaces designed for people, humanoids could open the door to an entirely new range of applications. One industry forecast envisions annual humanoid shipments growing from 20,000 units in 2025 to 1.4 million units by 2035.7

Getting there will take more than an impressive demonstration. Affordability, reliability, and safety could all determine how widely humanoids are deployed. McKinsey estimates that humanoid prototypes typically cost $150,000-$500,000 today, while costs may need to fall to $20,000-$50,000 to compete in broader commercial applications.8

So where might economic breakthroughs matter most? It isn’t necessarily the robot’s brain. The biggest hardware cost is often what makes it move. Motors, gearboxes, and joint assemblies, collectively called actuations, can account for 40-60% of a humanoid’s bill of materials.8

What drives the cost of humanoid robots?

Source: Humanoid illustration adapted from McKinsey, “Turning humanoid supply chain into billion-dollar wins,” 4/1/2026; cost estimates from McKinsey, “Humanoid robots: Crossing the chasm from concept to commercial reality,” 10/15/2025. Estimated component cost ranges as shares of the bill of materials. Ranges vary by robot design and should not be added together. Forward looking estimates may not come to pass.

Chart description: Illustration of a humanoid robot identifying key hardware components, alongside a horizontal bar chart showing estimated component costs as a percentage of total bill of materials. Actuation, including motors, joints, and gears represent the largest cost at 40-60%.


That leaves plenty of room for innovation in less photogenic parts of the machine, such as smaller motors, more precise joints, smarter sensors, and more efficient power systems. Improvements in these components could make a growing range of applications practical long before a general-purpose robot is folding laundry at home.

Where are the investment opportunities in robotics?

The companies developing tomorrow’s robots may be only part of the story. Behind each machine lies a broader ecosystem of technologies working together to bring physical AI to life.

  • Semiconductors provide computing power.
  • AI models help machines interpret their surroundings and make decisions.
  • Components such as sensors, motors, and actuators help turn those decisions into physical action.

Together, these technologies support robotics systems across a growing range of applications, from industrial automation and autonomous vehicles to humanoids and space-related technologies.

The physical AI technology stack explained

Source: BlackRock FE Technology Team, 9/2026. For illustrative purposes only.

Chart description: Infographic illustrating the physical AI technology stack and the technologies enabling robotics applications.


This may create two potential growth engines: more machines performing existing tasks and increasingly capable robots taking on entirely new ones. Of course, not every technology or application will advance at the same pace. High development costs, uncertain adoption timelines, and intense competition could pose risks to profitability and investment returns, making it especially important to identify innovations that deliver meaningful improvements in performance, reliability, and cost.

That’s why we believe an active approach to robotics investing — as found in the iShares Future Robotics and Related Technologies Active ETF (RBTX) — may be appropriate at this relatively early stage. The fund is managed by an experienced team led by Tony Kim, Head of BlackRock’s Global Fundamental Technology Team, and relies on bottom-up company research, detailed data modeling and risk-aware position sizing as it seeks to maximize total return in the global robotics theme.

Hollywood imagined a future where robots could do almost anything. The next chapter may be even more interesting as we discover just how much more they can help us accomplish.

Frequently asked questions

Physical AI refers to artificial intelligence that helps machines perceive, interpret, and respond to their physical surroundings. Unlike traditional robots that follow predefined instructions, physical AI aims to help robots adapt to changing conditions and perform more complex tasks.

Demand for robotics is being driven by aging populations, workplace safety needs, and pressure to improve productivity. As technology advances, robots could perform more tasks across manufacturing, logistics, transportation, and other industries.

Industrial robots are typically designed to perform specific, repetitive tasks in controlled environments, such as factories. Humanoid robots are built to operate in spaces designed for people and could eventually perform a broader range of tasks, although cost, reliability, and safety remain challenges.

Potential robotics investment opportunities extend beyond robot manufacturers to companies developing semiconductors, AI models, sensors, motors, actuators, and other enabling technologies. These components support a range of applications, including industrial automation, autonomous vehicles, and humanoid robots.

Robotics investing involves risks related to technology development, manufacturing costs, competition, and the pace of adoption. Not every robotics application or company will succeed, and commercial deployment may take longer than anticipated. These uncertainties could affect investment outcomes.

Featured funds

Photo of Jay Jacobs

Jay Jacobs

U.S. Head of Equity ETFs at BlackRock

Titania Hanrahan

U.S. Head of Sectors and Thematic Strategy, Fundamental Equities

Contributor

Gareth Feighery

Director, Product Strategy, Fundamental Equities

Contributor

Thi Hoang

Product Strategist

Contributor