Humanoids by 2040
Yash@yash
−9.1%
since Oct 6
The idea
By 2040, I expect roughly one humanoid robot for every 100 people on Earth. They will soon make a lot of economic sense at scale. Factories with expensive or scarce labor will be among the first big users, especially as more manufacturing moves back to the US. Most humanoid companies are still private, so I’m investing around them instead: chips, motors, actuators, sensors and companies likely to use robots at scale. I’m leaning more toward the physical hardware side, with some robotics ETFs and cash to add over time.
What it owns
- Motors, gears, sensors and power electronics are the least priced-in layer, and every robot needs them no matter the manufacturer. · Regal RexnordIt sells the gears and motion parts that make up dozens of joints in every humanoid, getting paid regardless of which robot design wins., TimkenIt makes bearings and drives that move joints, and the body layer is the least priced-in part of this take if humanoid production scales., TE ConnectivityIt dominates connectors and sensors that link every joint to the brain, so higher robot volumes mean more content per unit regardless of manufacturer., AmphenolIt is the other half of the connector duopoly, wiring power and data through everything that moves, capturing content per robot as volumes scale., onsemiIt supplies image sensors for robot eyes and power chips for motors, capturing two layers of the body as humanoids move from demos to deployment., NovantaIt sells precision motion and sensing parts to robot makers, which humanoid hands and arms need for fine control., Monolithic Power SystemsIt makes power chips that move battery energy efficiently to motors and compute, so more robots means more power chips per unit., STMicroelectronicsIt makes motion sensors that give humanoids balance, the same parts in phones, and a walking robot needs them everywhere., Allegro MicroSystemsIt makes magnetic sensors that tell motors exactly where joints are, and each humanoid carries dozens of motors.40%
- Companies deploying robots into factories and warehouses where labor shortages and reshoring create the first real demand. · TeslaIt is building humanoids at car-factory scale with Optimus and plans to deploy them in its own US plants, aligning with reshored factories short on workers., TeradyneIt owns Universal Robots and MiR, so it already sells working robots into factories today while humanoids mature, bridging current automation revenue and the long-term bet on factory robot demand growth., Rockwell AutomationIt automates factories, and reshored US plants short on workers are the early buyer, so it gets paid before humanoids ship., SymboticIt runs robot fleets inside warehouses right now, proving companies will pay for labor done by machines at deployment scale., Zebra TechnologiesIt sells scanners, mobile robots and tracking gear that warehouses use to automate today, getting paid as labor-short operations buy machines before humanoids arrive.26%
- Chips and AI models that let humanoids see, think and move, sold to all robot makers regardless of brand. · NVIDIAMost humanoids reaching scale will think on its chips and train in its simulation tools, making it the default platform for robot brains., QualcommIt makes low-power chips built for on-device AI, which humanoids need to run on batteries without phoning a data center for every step., AmbarellaIt makes chips that turn camera feeds into actionable information for robots, a bet that robot vision becomes a real market.16%
- Two robotics funds spread the bet across the whole industry, steadying a take that runs toward a distant horizon. · ROBO Global Robotics and Automation ETFThis fund spreads money across the whole robotics industry, including winners not picked individually and companies not yet public., Global X Robotics & AI ETFIt holds major robotics and AI companies globally, including Japanese makers that are hard to own directly, providing ballast so the take does not depend on getting every single pick right over a long horizon.10%
- Cash8%