This paper examines knowledge sharing processes in digital government projects (DGPs). Although knowledge sharing processes are a central feature of the functioning of government, they have received little attention in the literature. The importance of knowledge sharing has become even more evident with the rise of digital government initiatives, as these have a networking effect on bureaucracies. With multiple agencies and multidisciplinary knowledge coming together, it is necessary to combine and reconnect the required knowledge. Based on empirical data from four DGPs in Switzerland and the United States, a theoretical model for knowledge sharing in DGPs is proposed. The model ties together processes and content of knowledge sharing through two task dimensions, each of them subdivided into two phases: (1) Conception and implementation of the project, and (2) Exploration and exploitation of knowledge. I show that the implementation of DG requires a balanced mix of exploration and exploitation of knowledge, where exploration is more important in the conceptual phase, and exploitation becomes more fruitful in the implementation phase. I found that different configurations of knowledge networks in DGPs are related to different stages in the advancement of the project: DGPs rely on sparse networks during the conceptual phase of the project, and tend to employ teams or tightly-bound groups when the project moves towards implementation. The study concludes with implications for research and practice.