H.264/AVC is adopted as a next generation moving picture compression standard. Context-based Adaptive Binary Arithmetic Coding (CABAC) is the major entropy coding algorithm employed in H.264/AVC. Although the performance gain of H.264/AVC is mostly resulted from CABAC, it is difficult to implement a high-throughput decoder due to its decoding complexity. Although CABAC excludes a multiplication, the algorithm is basically sequential and needs large computations to compute some important variables, which are range, offset and context variables. Therefore, it is difficult to achieve fast decoding performance. In this paper, a prediction scheme is proposed to decode maximally two bits at a time and thus to reduce overall decoding time. A CABAC decoder based on the proposed prediction scheme reduces total cycles by 24% compared to conventional decoders.