The strict latency constraints of emerging vehicular applications make it unfeasible to forward sensing data from vehicles to the cloud for processing. To shorten network latency, Vehicular fog computing (VFC) moves computation to the edge of the Internet, with the extension to support the mobility of distributed computing entities. In other words, VFC proposes to complement stationary fog nodes co-located with cellular base stations with mobile ones carried by moving vehicles. Previous works of VFC mainly focus on optimizing the assignments of computing tasks among available fog nodes. However, capacity planning, which decides where and how much capacity to deploy, remains an open and challenging issue. The complexity of this problem comes from the mobility of vehicles, the spatio-temporal dynamics of vehicular traffic, and the computing resource demand generated by varying vehicular applications. To solve the above challenges, we propose a data-driven capacity planning framework that optimizes the deployment of stationary and mobile fog nodes to minimize the installation and operational costs under the quality-of-service constraints, taking into account the spatio-temporal variation in computing demand. Through real-world experiments, we analyze the cost efficiency potential of VFC in long term and demonstrate that the performance loss of VFC is below $6\%$ compared to stationary deployment with equal network capacity. We also analyze the impacts of traffic patterns on the potential cost saving. The results show when the traffic density is higher, more operational costs will be saved in the long run due to more dense deployment of mobile fog nodes.