Cognitive radio based smart grid communication allows smart grid users to utilize spectrum bands opportunistically. Cognitive radio does not require spectrum leasing cost, saves users' energy consumption, keeps interference level low, and improves interoperability. In low traffic load environment, cognitive radio is able to support smart grid data transmission. However, the traffic load in Advanced Metering Infrastructure (AMI) can be as high as 1000 kbps or 500 kbps for backhaul. In this paper, the performance of cognitive radio in high traffic load AMI is investigated. Iterative Water Filling Algorithm (IWFA) is implemented upon cognitive radio to maximize the capacity of smart grid users through computing optimal transmit power for each user. The performances of sequential IWFA and asynchronous IWFA in smart grid communication are investigated. Simulation is conducted using Network Simulator 2. Simulation results show that IWFA indeed is able to maximize the capacity of smart grid users. Results also show that the asynchronous IWFA performs better than sequential IWFA. Asynchronous IWFA improves delivery ratio and throughput of cognitive radio based AMI communication by 5.95% and 8.67%, respectively.