Some fun robot math

He is bullish on robots as a category rather than on any one shape, because every important input has got cheaper and better at once over the past decade: battery energy density, graphics-processor energy efficiency, LiDAR resolution and whatever stands in for intelligence in AI. Numbers like that, he says, tend to herald the beginning of a super-cycle. Shape is the open question. Robots look to him like organisms radiating into task-specific forms; the rival view, which he attributes to Tesla and Figure, is that most human spaces and jobs were designed around human-shaped workers, so a humanoid is the most versatile machine available. He is not convinced, but is intrigued enough to run two numbers. The energy one, offered mostly for fun, goes against the robots. The cost one would decide whether they substitute for people, and it goes for them, on his own medium-term assumptions about capital cost, lifetime, energy cost and utilization.