We introduce in the theory of Gabaix and Maggiori (2015) a network structure to capture the complexity of the balance sheets of financial intermediaries, using the Leontief inverse-based centrality. We use this framework in a multi-country world with imperfect financial markets to study how currency risk premia are connected to financiers’ risk bearing capacity. Guided by the theory, we construct a Centrality Based Characteristic (CBC), based on the centrality of the trade imbalance network and variance–covariance matrix of currency returns. Sorting currencies on CBC generates a high Sharpe ratio, and the resulting excess returns reflect a novel source of predictability.