Standard analyses of economic policy assume exponential discounting, even though empirical and experimental evidence shows that preferences are time-inconsistent and discounting is hyperbolic. When policy makers-or the voters they must satisfy-apply smaller discount rates for long-term than for short-term decisions, they benefit from investing in infrastructure and technologies that will influence future decisions. This paper analyzes the equilibrium investment strategy and policy as a function of the technology's type and position in the production chain. The strategic concern can be measured by the subsidy a sophisticated decision maker would impose on a naive agent, or on a perfect market. Two main results are pro- vided. First, I derive a formula for how the optimal investment subsidy depends on the investment lags and the technology's complementarity with future investments. When applied to climate change, it implies that investments in "green" technology should be subsidized while adaptation and "brown" technology should be taxed, even when laissez faire is first best under exponential discounting. Second, I show that fundamental technologies (i.e., those further upstream in the production chain) should be invested in and subsidized to a larger extent. This result also reveals that quasi-hyperbolic discounting is a poor approximation for strictly decreasing discount rates.