Firms Accumulating Medical-AI Patents Are Valued More Highly, and Incumbent Health Firms Own the Medical-AI Frontier

Abstract

Artificial intelligence is moving into medicine, from diagnostic imaging to drug discovery and clinical prediction. Yet it remains unclear which firms are building this capability and whether financial markets reward them. This paper measures the diffusion of AI into medicine using patents. Patents flagged as AI by the USPTO Artificial Intelligence Patent Dataset and classified as medical under CPC subclass A61 are linked to public firms through the Kogan et al. (2017) crosswalk and merged with Compustat, producing a panel of 7,087 firms and 83,976 firm-years from 1995 to 2024. Firm-matched medical-AI patenting roughly doubled between 2013 and 2020, rising from about 920 to 1,800 patents per year. Ownership is concentrated among incumbent health firms, which hold 47 percent of firm-matched medical-AI patents across 482 firms, compared with 19 percent across 135 technology entrants. In regressions with firm and year fixed effects, a larger cumulative medical-AI patent stock is associated with higher firm value, measured by both Tobin’s Q and market capitalization. This association remains stable with industry-by-year fixed effects, a one-year-lagged patent stock, a matched sample of comparable non-adopters, and randomization inference. Future medical-AI patenting also does not predict current firm value after controlling for current patent stock. However, the estimated value premium is not detectably larger for incumbents than for entrants. Overall, medical AI has diffused primarily through established health firms, although the market appears to reward medical-AI innovation across firm types.