Worth sharing: the real bottlenecks in physical AI
I came across a post from Inhi Cho Suh, CEO of Niantic Spatial, and it stuck with me enough to write down.
In her video, she makes a point I haven’t heard framed quite this way before: digital AI, for all its disruption, only operates on screens — and screens represent just 20% of global economic activity and workforce. The other 80% happens out in the physical world. Factories, airports, facilities, sports arenas, cities. Spaces where machines and people already interact every day, and where the real unlocks — robots, wearables, AR glasses, drones — are still largely ahead of us. Her vision: a world where physical AI simply makes that interaction safer.
I find that reframe genuinely clarifying. We’ve spent years talking about AI disrupting knowledge work, and that’s real — but it’s disrupting a fraction of what actually happens in the world. The bigger, messier, more consequential challenge is what happens when intelligence moves off the screen and into physical space.
I worked with Inhi for several years and have a lot of respect for how she thinks about this space. Physical AI feels like it could be the next real frontier — not the flashiest conversation in AI right now, but maybe the one with the most consequential unlocks still ahead of it. Curious to see how much of this gets solved in the next couple of years, and how much of it turns out to be structural rather than technical.