Since the Lisbon agenda (2000), European policies are increasingly oriented towards R&D and innovation. In this context, we analyze the effects of a centralized R&D subsidy policy upon steady state and welfare using an agglomeration and growth model composed of two asymmetric countries. If the policy leads to a steady state characterized by a higher growth rate and lower inequalities, the welfare analysis provides a more contrasted vision on the effects of the policy. Indeed, even though such a policy can eliminate some distortions and improve global welfare, it implies a zero-sum game at national level (the policy increases the welfare in the periphery country and decreases the welfare in the core country) when knowledge spillovers are sufficiently localized. Consequently, a centralized R&D subsidy policy conducted in an economy composed of two asymmetric countries would be a positive-sum game only if international knowledge spillovers are sufficiently large.