The present paper tries to identify the optimal size of a wind farm using North European data. An empirical analysis of 61 sites constructed between 2004 and 2014 suggests that economies-of-scale are highly heterogeneous across on-shore and off-shore projects. A Varying Coefficient Model captures such diversity by making the impact of the farm site on the amount of its potential capacity a non-linear function of the number of installed turbines. The resulting scale elasticities suggest that small on-shore farms have a bigger per-turbines output than off-shore ones, while the opposite is true for big projects.