With the development of fifth generation of wireless communication technologies that meet the requirements of high reliability and low latency, autonomous driving is getting closer to reality. In recent years, as interest in autonomous driving has greatly increased, an international consensus has emerged that a broadband spectrum is required for autonomous driving. However, despite this attention, research has not proposed how to estimate spectrum requirements for autonomous driving to ensures quality of service for all users in the cell. Thus, in this letter, we propose a methodology of estimating spectrum requirements based on signal-to-interference-plus-noise-ratio probability distributions. In addition, a practical environment with a traffic model is set up with data analysis for autonomous driving. Through these processes, spectrum requirements for autonomous driving are estimated.