Gene expression is essential for biological functions but also incurs a fitness cost, measured by the relative reduction in growth rate. However, the limiting factors generating the fitness cost remain elusive due to the complexity of cell growth. In this work, we disentangle the fitness cost into components generated by different limiting resources based on a resource competition model. We derive analytical formulas connecting the fitness cost to the copy numbers of the ribosome, RNA polymerase, and transcription factor. Comparing our model predictions with experimental data of Saccharomyces cerevisiae, we demonstrate that the competition for ribosomes dominates the translation cost and the competition for transcription factors dominates the transcription cost. The model also systematically connects the fitness cost to genetic and environmental properties, making predictions quantitatively consistent with various experimental observations. Our work builds a systematic framework for gene expression cost, guiding synthetic biology to optimize genetic design.