The robot boom is bigger than humanoids: mapping the companies behind it

Key takeaways

• Humanoids are the headline, but robotics already spans factories, warehouses, vision systems, motion control and artificial intelligence.

• Nvidia, Cognex, Fanuc and specialist component makers provide very different ways to gain exposure to the same theme.

• The key question is not only which robot wins, but which parts of the value chain remain scarce and valuable.

Robots can already dance. Investors appear equally capable of getting excited about it.

China's Unitree surged nearly sixfold on its Shanghai debut on 19 August, briefly valuing the humanoid robot maker at around 50 billion USD. The listing coincides with Beijing's World Robot Conference, where hundreds of companies are displaying everything from industrial machines to humanoid robots.

Unitree makes a useful hook, but it is not the whole investment story. Humanoids remain early, and relatively few perform productive commercial work today. The broader opportunity is what happens when artificial intelligence gives existing robots better brains, eyes and decision-making abilities.

For investors, that means looking underneath the robot's skin.

Humanoids are the billboard, not the industry

Factories installed more than 540,000 industrial robots globally in 2024, more than twice the number a decade earlier, according to the International Federation of Robotics. Manufacturing already uses automated arms extensively for welding, assembly, painting and moving materials.

The important change now is intelligence.

Traditional industrial robots are excellent at repeating precisely programmed movements. They struggle when an object moves, changes shape or something unexpected enters their workspace.

Artificial intelligence can make that environment less rigid. Machine vision helps robots see. New models help them interpret instructions. Simulation lets them practise tasks virtually before entering a real factory.

Nvidia says Fanuc, ABB Robotics, Yaskawa and Kuka, which together have more than two million installed robots, are integrating its simulation technology and artificial intelligence computing.

That suggests robotics may evolve more like semiconductors than smartphones. The final product gets the attention, but considerable value can sit with companies supplying essential components.

Build the robot, layer by layer

The easiest way to understand the opportunity is to imagine building an intelligent robot from scratch.

The brain: Nvidia

Robots need computing power to process cameras, understand their surroundings and decide what to do next. Nvidia is trying to become a common computing and software platform for robotics.

Its Jetson processors run artificial intelligence directly on machines, while Isaac provides tools for training and simulation. Its GR00T platform targets humanoid development.

Importantly, Nvidia does not need to predict whether Tesla, Unitree, Figure or another manufacturer wins. Its opportunity is supplying the intelligence layer across many platforms.

The eyes: Cognex and Keyence

A robot cannot adapt if it cannot see what changed.

Cognex specialises in machine vision. Its cameras, sensors and software identify objects, inspect products and guide robotic movements. Keyence offers similar systems, including three-dimensional vision that helps robots understand an object's position, shape and orientation.

As robots move from fixed tasks towards less predictable environments, better perception becomes increasingly important.

The muscles and joints: Harmonic Drive and Nabtesco

Making a robot think is only half the problem. It must also move accurately.

Motors, gears, bearings and actuators convert electrical instructions into movement. McKinsey identifies actuators as the largest hardware cost and one of the main performance bottlenecks for humanoids.

Japan's Harmonic Drive Systems produces precision gearing used in humanoid arms and legs. Nabtesco supplies reduction gears that allow industrial robots to move accurately under heavy loads.

If humanoid production scales, these less glamorous components could become strategically important. A robot with a brilliant brain but unreliable knees still has a problem.

The robot builders: Fanuc, Yaskawa, Teradyne and Tesla

Fanuc and Yaskawa bring decades of experience building industrial robots, controllers and motion systems. They also already have factory customers and experience keeping machines running reliably.

Teradyne offers another angle through Universal Robots, which makes collaborative robot arms, and Mobile Industrial Robots, which builds autonomous machines for moving goods through factories and warehouses.

Tesla sits further towards the speculative end. Optimus aims to become a general-purpose humanoid. Its challenge is turning artificial intelligence, vision and manufacturing expertise into dependable machines produced cheaply at scale.

ABB also belongs on the map, with one caveat. It agreed to sell its robotics division to SoftBank for 5.375 billion USD, shifting direct robotics exposure away from ABB if the transaction completes.

The warehouse: Kion and Symbotic

Robotics does not need legs to create economic value.

Kion combines warehouse equipment, automation and robotics and is working with Nvidia on artificial intelligence-powered autonomous industrial trucks. Symbotic automates the storage and movement of goods inside large warehouses.

These businesses offer a useful reality check. The first large-scale financial returns from smarter robots may emerge in controlled factories and warehouses long before humanoids start folding laundry at home.