We study the causal effect of conflict on investment using a unique administrative dataset from a large bank serving rural producers in Colombia. Our difference-in-difference strategy exploits the 2016 peace agreement between the Colombian government and insurgent group FARC, combined with pre-existing differences in FARC exposure across municipalities. We show that the number of business loans increases in municipalities with historical FARC presence after the peace agreement. More loan applications drive this increase, with no change in supply-side variables. However, higher investment is only observed in municipalities located close to markets and does not materialize before the peace agreement is finalized, despite a large decline in violence during the preceding negotiations period. A simple theoretical framework combined with rich information on the characteristics of loan applicants and projects (including credit scores and delinquency rates), as well as night-time lights, suggests that conflict hinders investment mostly by lowering project returns.