The emergence of hyperspectral remote sensing has greatly promoted the development of the remote sensing technology. Endmember extraction is an important task in hyperspectral data processing. Pixel purity index (PPI)[1] algorithm has been widely used for endmember extraction in hyperspectral images. With the development of hyperspectral sensors, the resolution of hyperspectral images increases and the traditional hyperspectral processing algorithm is highly time consuming as its precision increases asymptotically. In order to process massive hyperspectral data efficiently, this paper proposes a distributed parallel implementation of PPI algorithm (PPI_DP) on cloud computing architecture. The realization of the proposed method using Spark framework and MapReduce model is described and evaluated. Experimental results demonstrate that the proposed method can effectively extract the endmembers of large quantity hyperspectral data.