Outlook theme 2 of 4
Physical AI & robotics
AI is learning to act in the real world - not just on a screen.
The big picture
For most of its history, AI lived inside computers - recognising text, images and speech. Physical AI is different: models that perceive the world through cameras and sensors, reason about it, and then move something - a robotic arm, a forklift, a vehicle, a drone.
The same breakthroughs that made chatbots fluent are now teaching machines to grasp unfamiliar objects, navigate messy environments and learn new tasks from demonstration rather than months of hand-written code.
Why now
Three trends are converging: AI models that can understand the physical world, falling costs for sensors and actuators, and labour shortages in manufacturing, logistics and aged care. Together they make robots economically sensible in places they never were before.
Where the opportunity sits
The layers I’m researching
Industrial automation
Established automation suppliers adding AI to arms, vision systems and controllers.
Components & sensors
Motors, actuators, cameras and lidar that every robot needs - the shovels of this theme.
Simulation & training
Software that trains robots in virtual worlds before they touch a real one.
Autonomy in heavy industry
Mining, agriculture and logistics - sectors where Australia already leads in automation.
Signals I watch
How I’ll know it’s working
- ↗Warehouses and factories moving from pilots to fleet-scale robot deployments
- ↗Falling cost of robot hardware and sensors
- ↗Progress in general-purpose ‘foundation models’ for robotics
- ↗Growth of autonomous vehicles and machinery in mining and agriculture
Risks
What could go wrong
- !Timelines for general-purpose and humanoid robots may prove much longer than the hype suggests
- !Hardware businesses are capital-intensive and often lower-margin than software
- !Safety, regulation and public acceptance may slow deployment
How I’m approaching it
This is an early-stage theme, so I keep it small and diversified - favouring profitable suppliers of components and automation over single-bet start-ups - and add as real-world deployments scale.
“The first chapter of AI learned to talk. The next chapter learns to work with its hands.”
These are personal, forward-looking views, not predictions or recommendations. They may turn out to be wrong.