Every AI system in wide commercial deployment has operated on the same side of a single boundary: data comes in, output goes out, and the system's reach ends there. Physical AI is built to cross that line. The category, referring to AI designed to interact with and act within the real world, is drawing high expectations from Wall Street and Silicon Valley while unions and economists are raising warnings about its effects on jobs and wages.
The action surface
The constraint that makes physical AI a distinct engineering category is the shift from generating output to taking physical action. Software-based AI perceives and classifies data; physical AI is built to perceive and operate in actual environments. That extension of the action surface is the mechanism behind the investment case: categories of work that required human physical presence become, in principle, addressable.
Wall Street and Silicon Valley's interest reflects a view that this expansion is substantial. Work that was insulated from software-based automation by the requirement of physical presence would, under that reading, no longer carry the same protection.
Unions and economists are not disputing the technology's reach. Their concern is about what that reach produces: warnings specifically about jobs and wages, and about the scope of the worker population that physical AI could affect as the technology moves from early stage toward wider deployment.