The size and significance of public infrastructure investment impacts on costs and productivity of private enterprise, and thus on economic health and growth, has proven nebulous to empirically substantiate. Various studies using alternative theoretical and econometric methodologies, and for different time periods, sectors, and countries, have tentatively established that such a productive impact exists and is statistically significant. It also seems smaller and more variable over time, space, and sector than was implied by initial studies on the "public capital hypothesis". One piece of the puzzle that has received little attention, however, is the role of spatial spillovers in driving infrastructure investment benefits. Such spillovers are not only conceptually important, but could also shed light on discrepancies between studies for different data, and particularly aggregation levels. In this study we apply a cost-based model to state-level U.S. manufacturing data, for capital, production and non-production labor, and materials inputs, and for the 1982-96 time period, in an attempt to untangle the private cost-saving contributions of inter- and intra-state public infrastructure investment. We carry out two kinds of spatial adaptations - a spatial autocorrelation adjustment and a spatial spillover theoretical modification - to the estimating system consisting of a Generalized Leontief cost function and input demand equations, to address this issue. We find that intra-state public infrastructure benefits appear larger in magnitude when inter-state spillovers are directly recognized, as well as being invariably statistically significant. Inter-state spillovers are also directly beneficial to manufacturing firms, although their contribution appears smaller in size when temporal serial correlation is recognized in addition to spatial correlation.