The article is devoted to a spatial econometric analysis of the impact of enterprises’ innovation activity on the long-term economic growth of regions, taking into account inter-territorial external effects and the structural heterogeneity of the economic space. The purpose of the study is to provide a quantitative assessment of the direct and indirect effects of R&D expenditures, innovative product output, and patent activity on the growth rate of real GRP per capita in Russian regions, as well as to decompose the aggregate effect into own and neighboring components. The empirical base consists of balanced panel data for 82 subjects of the Russian Federation for 2012–2022, using statistics from Rosstat, specialized patent and scientometric databases, and additional macroeconomic sources; the model includes control variables reflecting investment in fixed capital, the level of urbanization, the structure of employment, and the openness of the regional economy. The methodology relies on the tools of spatial econometrics: construction of spatial weight matrices based on adjacency and inverse distances, diagnostics of the global Moran’s index, estimation of SAR, SEM, and spatial Durbin (SDM) models by maximum likelihood with fixed effects, as well as dynamic panel estimation using the system GMM to control for endogeneity. It is shown that innovation activity is characterized by high variation and stable positive spatial autocorrelation, strengthening over time, which indicates the formation of clusters of “growth poles.” The results of the SDM model demonstrate a statistically significant positive impact of own R&D expenditures and patent activity on GRP growth rates, while the indirect (spillover) effects of innovations from neighboring regions are comparable to or even exceed the direct effects, whereas the spatial effects of physical capital investment are partly competitive in nature. The decomposition of spatial effects shows that more than half of the total return on R&D is realized outside the investor region, creating a “free rider” problem and justifying the need for cluster-oriented and differentiated regional policy. It is concluded that ignoring spatial interactions leads to biased estimates of innovation policy effectiveness, while accounting for interregional spillovers is critical for calibrating instruments of inter-budgetary equalization, supporting clusters, and developing transport and digital infrastructure that expand the radius of innovation impulses.