Cloud computing has undergone rapid expansion throughout the last decade. Many companies and organizations have made the transition from traditional data centers to the cloud due to its flexibility and lower cost. However, traditional data centers are still being relied upon by those who are less certain about the security of cloud. This problem is highlighted by the fact that there only exist limited efforts on threat modeling for cloud data centers. In this paper, we conduct comprehensive threat modeling exercises based on two representative cloud infrastructures using several popular threat modeling methods, including attack surface, attack trees, attack graphs, and security metrics based on attack trees and attack graphs, respectively. Those threat modeling efforts provide cloud providers practical lessons and means toward better evaluating, understanding, and improving their cloud infrastructures. Our results may also imbed more confidence in potential cloud tenants by providing them a clearer picture about potential threats in cloud infrastructures and corresponding solutions.