The LZW data compression algorithm is a dictionary-based compression algorithm in which the process of searching for a pattern and its corresponding code to compress the data is a time-consuming process. As the file size to be compressed becomes larger the size of the dictionary also increases which increase the time to search a pattern in the dictionary. This paper proposes a novel approach to increase the efficiency of the algorithm with GPU using CUDA. Thus, making the process of searching the dictionary parallel. The result of the proposed system outlines the improved performance of the LZW algorithm