In this study, we analyse regional cost differences of fibre-based access networks. Our data base comprises a complete sample of Very High Capacity Network (VHCN) investment figures. By matching this data with the internationally standardised EUROSTAT and BBSR urban/rural typology classification, we show that such classification criteria do not sufficiently account for a large share of geographical differences in fibre-based access network costs. In order to better explain and/or identify regional differences in VHCN investment, we turn to spatial regression models to identify alternative influencing factors solely on the basis of publicly available data. We show that a handful of geographical factors are capable of explaining 95% of the differences in fibre investment requirements; the most relevant being (1) the size of demand (as number of access lines), (2) the street-based household density (defined as the number of households per kilometre of road in built-up areas), (3) a dispersion measure (approximated by the main road length per built-up area) and (4) the degree of urbanisation (measured by the share of built-up area in relation to the overall area). These results are consistent at different levels of spatial aggregation (e.g. from access areas to NUTS-3 level) and even after controlling for neighbouring effects. Thus, it is capable of predicting costs more precisely and at the level of the territorial unit, at which funds are bounded to be allocated. From a public policy perspective, the proper identification of areas, where the commercial roll-out is unlikely to occur, is key in preventing the widening of a digital gap without having a wasteful use of public funds.