To quickly and accurately obtain residential project cost time series variation inherent laws, improve the effectiveness of residential project cost prediction, a model of the gray Markov is established by combing the gray GM (1,1) prediction methods with the Markov probability transfer method effectively to forecast the residential project cost. The residential project cost of Chengdu city is analyzed and predicted by this model, the predicted results agree well with the actual value. Practical application verifies that it is feasible to establish this prediction model, which has popularization and application value in residential project cost prediction.