Robotics & Physical AI

Robotics and physical AI is the bet on machines that perceive and act in the real world: humanoids, collaborative robots, factory automation, and autonomous mobility. As AI models move from screens to bodies, the software that ran chatbots is learning to control motors and hands. This pillar maps the theme and how to invest in it.

Key takeaways
  • Robotics and physical AI is the bet on machines that perceive and act in the real world: humanoids, cobots, factory automation, and autonomous mobility. As AI models move from screens to bodies, the software that has run chatbots is now learning to control motors and hands.
  • The base is already large and growing. The world ran 4,664,000 industrial robots in 2024, up 9% in a year, with 542,000 new units installed (IFR World Robotics 2025).
  • The new layer is the humanoid. Goldman Sachs raised its 2035 total addressable market for humanoid robots to about $38 billion, more than six times its earlier $6 billion call, on roughly 1.4 million units shipped.
  • NVIDIA frames this as 'physical AI' and has built the brain (Jetson Thor, Isaac GR00T) that many robot makers run on, the same playbook it used in data-center compute.
  • Korea is structurally central: it has the world's highest robot density at 1,220 robots per 10,000 workers, and Hyundai's Boston Dynamics Atlas is committed to deployment in 2026 with a 30,000-unit-a-year factory targeted by 2028.

What is physical AI?

For a decade, the headline AI lived on a screen. It wrote, drew, and answered questions. Physical AI is the next step: the same kind of foundation model, but wired into a body that senses the world and moves things in it. NVIDIA’s framing is blunt. “Physical AI and robotics will bring about the next industrial revolution,” says CEO Jensen Huang (NVIDIA). The bet here is on the companies that build those bodies and the chips, sensors, and models that run them.

What are the parts of the robotics market?

The theme is not one thing. It splits into layers, each with its own incumbents and its own economics:

  • Industrial robots and cobots: the arms that weld, paint, and pick on factory lines, plus the lighter collaborative robots that work next to people. This is the mature, cash-generating base of the market.
  • Humanoids: general-purpose machines with legs and hands, meant to do work designed for humans. Still early, but the fastest-moving frontier, led by Boston Dynamics and a wave of startups.
  • The robot brain: the compute and AI models that turn sensor data into action. This is where NVIDIA sits, via its Jetson Thor onboard computer and Isaac GR00T robot foundation models.
  • Autonomous mobility: self-driving and automated logistics, the moving robots that carry goods and people.

The investable bottleneck shifts between these. Right now the scarce input is the brain, a model good enough to generalize across tasks, which is why the compute and software layer commands the most attention.

Why does it matter now?

The base is already enormous. The world ran 4,664,000 industrial robots in 2024, a 9% rise in a single year, with 542,000 new units installed (IFR World Robotics 2025). That figure has roughly doubled over a decade. The IFR expects installations to keep climbing, to about 575,000 units in 2025 and past 700,000 by 2028.

The total number of industrial robots in operational use worldwide was 4,664,000 units in 2024, an increase of 9% compared to the previous year.

— Takayuki Ito, President, International Federation of Robotics

On top of that base sits a new and much-debated layer: the humanoid. Goldman Sachs raised its 2035 total addressable market for humanoid robots to about $38 billion, more than six times its earlier $6 billion estimate, on roughly 1.4 million units shipped (Goldman Sachs Research). The driver behind the upgrade is AI itself: models good enough to let a robot learn a task instead of being hand-coded for it.

Why is NVIDIA central to robotics?

NVIDIA is running the same play in robotics that it ran in data centers: sell the brain. Its Jetson Thor onboard computer and Isaac GR00T foundation models are the stack many robot makers build on, so NVIDIA captures value whoever wins the hardware race (NVIDIA). It is the connective tissue between the compute theme and the robotics theme, which is why the chip leader appears in both clusters.

Why is Korea a robotics leader?

Korea is the most automated economy on earth. It runs 1,220 robots per 10,000 manufacturing workers, the highest density in the world and far above the global average of 132, ahead of Singapore and Germany (IFR). Its electronics and auto industries are the two biggest buyers of industrial robots anywhere. Hyundai sits at the center of the humanoid push: it owns Boston Dynamics, whose electric Atlas robot has 56 degrees of freedom and can lift up to 50 kg, with deployments committed in 2026 to Hyundai’s own plant and to Google DeepMind (Boston Dynamics). Hyundai is building toward a factory that makes 30,000 robots a year by 2028, and frames its strategy around owning the full robotics value chain, from components to software (Hyundai Motor Group).

Explore this theme

Concepts

Companies

Key terms

How can you invest in it?

The cleanest exposures span the brain and the body. On the brain side, NVIDIA supplies the compute and models. On the body side, Korea’s Hyundai robotics value chain concept covers the listed names tied to Boston Dynamics and its suppliers. The Akros Hyundai Motor Robotics Value Chain TOP3 Plus Index groups those Hyundai-affiliated robotics names (Akros).

What are the risks?

Humanoids are early. Demos run ahead of revenue, and the gap between a robot that walks on stage and one that pays for itself on a line is wide. Many suppliers lean on a few large buyers, so orders can be lumpy. Costs and AI capability are both still moving fast, which can reprice the winners as quickly as it created them. Industrial automation is also a capital-cycle business that slows when factories pause spending.

Frequently asked questions

What is robotics and physical AI investing?

It is investing in companies that build machines which perceive and act in the physical world: humanoids, collaborative robots, factory automation, and autonomous mobility, plus the chips, sensors, and AI models that control them. The thesis is that AI is moving from software-only tasks into bodies that do physical work.

How big is the robotics market?

The installed base of industrial robots reached 4,664,000 units in 2024 with 542,000 new installations that year, per the IFR. The newer humanoid segment is forecast separately: Goldman Sachs puts the 2035 humanoid total addressable market at about $38 billion on roughly 1.4 million units.

Why is robotics accelerating now?

AI foundation models can now generalize to physical tasks, so a robot can learn from data rather than be hand-coded for each job. NVIDIA supplies the 'physical AI' compute and software (Jetson Thor, Isaac GR00T) that many robot makers use, and aging, costly labor pushes manufacturers toward automation.

Which countries lead in robotics?

China installs the most industrial robots (54% of 2024 deployments) and Korea has the highest robot density at 1,220 per 10,000 workers, ahead of Singapore and Germany. The United States leads in robot 'brains' through NVIDIA and in humanoids through Boston Dynamics, which Hyundai owns.

Sources & references

  1. World Robotics 2025 report — global robot demand in factories doubles over 10 years · International Federation of Robotics (IFR), 2025-09-25
  2. Robot density surges in Europe, Asia and the Americas · International Federation of Robotics (IFR), 2026-04-08
  3. NVIDIA powers humanoid robot industry with cloud-to-robot computing platforms for physical AI · NVIDIA Corporation, 2025-05-18
  4. The global market for humanoid robots could reach $38 billion by 2035 · Goldman Sachs Research, 2024-01-08
  5. Boston Dynamics unveils new Atlas robot to revolutionize industry · Boston Dynamics, 2026-01-05
  6. Hyundai Motor Group to unveil AI Robotics Strategy at CES 2026 · Hyundai Motor Group, 2025-12-22