I study the role of industries position in supply chains on the transmission of final demand shocks. First, I use a shift-share design based on destination-specific final demand shocks and destination shares to show that shocks amplify upstream. Quantitatively, I find upstream industries respond to final demand shocks up to three times as much as final goods producers. To organize the reduced form results, I develop a tractable production network model with inventories and study how the properties of the network and the cyclicality of inventories interact to determine whether final demand shocks amplify or dissipate upstream. I test the mechanism both by directly estimating the model and in reduced form and I find evidence of the role of inventories in explaining heterogeneous output elasticities.