In this paper, we propose a pipelined intra prediction using shuffled encoding order for H.264/AVC. Due to a dependency problem of 4×4 intra prediction mode, a new 4×4 block cannot be predicted before reconstruction of a previous 4×4 block. To minimize this latency, we shuffle the encoding order of 4×4 intra prediction while maintaining the consistency with the standard. To further reduce the processing time, two prediction modes are not used in some blocks. With the proposed method, we can implement the intra prediction block as a pipelined architecture. The processing time can be reduced by maximum 40%, because the latency caused by reconstruction of the previous 4×4 block was hidden. It was implemented into the hardware with some parallel processing architecture. The required cycles were 470 cycles.