EMU was a major step towards deeper financial integration among member states. However, diversification of equity portfolios remained limited while banking integration surged. We argue that the nature of banking integration is of first-order importance for understanding the patterns and channels of risk sharing. While EMU was associated with the creation of an integrated interbank market, as witnessed by an explosion in cross-border interbank flows, “real” banking integration (in terms of cross-border bank-to-real sector flows or banking-consolidation) remained limited. But we find that real banking integration is associated with more risk sharing, while indirect integration via interbank flows is not. Further, indirect banking integration proved to be highly procyclical, which contributed to the freeze in risk sharing after 2008. Based on this evidence, and a stylized DSGE model that allows us to explain these patterns in the data, we discuss implications for banking union. Our results show that real banking integration and capital market union are complements and robust risk sharing in the EMU requires both.