In our duopoly, an irreversible investment incorporates a significant amount of R&D, so that the improvement it introduces in production processes generates a spillover lowering the second comer's investment cost. The presence of the inter-firm spillover substantially affects the equilibrium of the dynamic game: for low - and hence realistic - spillover values, the leader delays her investment until the stochastic fundamental has reached a level such that the follower's optimal strategy is to invest as soon as he attains the spillover. This bears several interesting implications. First, because the follower invests as soon as he benefits from the spillover, in our equilibrium the average time delay between the two investments is short, which is realistic. Second, we show that in case of a major innovation, an optimal public policy requires a sub- stantial intervention in favour of the investment activity; moreover, an increase in uncertainty -delaying the equilibrium- calls for higher subsidization rates. Third, we find, by means of numerical simulations, that the spillover reduces the difference in the leader's and in the follower's maximum value function. Accordingly, our model can help generating realistic market betas.