Most research on global governance has focused either on theoretical accounts of the overall phenomenon or on empirical studies of distinct institutions that serve to solve particular governance challenges. In this article we analyze instead “governance architectures,” defined as the overarching system of public and private institutions, principles, norms, regulations, decision-making procedures and organizations that are valid or active in a given issue area of world politics. We focus on one aspect that is turning into a major source of concern for scholars and policy-makers alike: the “fragmentation” of governance architectures in important policy domains. The article offers a typology of different degrees of fragmentation, which we describe as synergistic, cooperative, and conflictive fragmentation. We then systematically assess alternative hypotheses over the relative advantages and disadvantages of different degrees of fragmentation. We argue that moderate degrees of fragmentation may entail both significant costs and benefits, while higher degrees of fragmentation are likely to decrease the overall performance of a governance architecture. The article concludes with policy options on how high degrees of fragmentation could be reduced. Fragmentation is prevalent in particular in the current governance of climate change, which we have hence chosen as illustration for our discussion.