Based on the vast and elastic cloud computing resource, a cloud-based cognitive radio access networking (C2-RAN) service model is studied in this work. Integrating the cross-layer C2-RAN functions of computing and spectrum resource sharing, channel and power allocation, as well as user scheduling and medium access control in a public cloud platform, we explore the feasibility and potential of deploying a green yet broadband wireless RAN service in TV white space with tens of thousands of low-power cognitive radio (CR) access points (APs). Through an experimental and simulated study of C2-RAN in a wireless metropolitan area network (MAN), the system throughput and power consumption of the proposed C2-RAN MAN is compared with that of a simulated 3G MAN. Simulatio results show that the C2-RAN MAN can offer an aggregate 726.4 Gbps system capacity with a total power consumption of 203.6 kilowatt over a maximum of 180 MHz bandwidth, while the 3G system provides an 8 Gbps system capacity with 940.2 kilowatt over 15 MHz bandwidth. The proposed C2-RAN appears to be a promising solution for future broadband wireless access.