Finance Seminar Adair Morse
The paper studies „physical AI“: inventions outside the IT sector that build on AI advances, identified through patent citations. The authors show this kind of innovation is nearly as common as AI patenting itself, commands a valuation premium, and that European firms lag behind their U.S. counterparts. Using local AI knowledge from universities and federal labs as an instrument, they find that physical AI innovation is associated with higher employment, capital, and profit growth, driven mainly by output expansion rather than cost savings. One exception: when physical AI patents are also automation patents (so-called „smart automation“) they instead reduce costs per unit of output, concentrated in industries with high cognitive-task intensity.
The findings suggest AI's economic footprint extends well beyond AI-producing firms, diffusing into ordinary, non-tech inventions across the economy.
Takeaway: Outside of automation, AI’s economic value seems to come from increased output rather than higher cost efficiency.