Transportation networks worldwide suffer from heavy congestion. This paper provides the first estimates of congestion’s effect on the production side of the economy, combining firm survey data with traffic data from Indian Railways. Geographic variation in congestion comes from a recent wave of passenger trains which were planned according to certain rigid rules, making it possible to identify the costs the additional traffic imposes on firms using the railways to ship goods. In estimating this congestion externality, the empirical strategy accounts for both direct and spillover effects of congestion. It also draws on a traffic model from operations research to disentangle a mean effect (congestion makes the average shipment slower) from a variance effect (congestion makes shipping times less predictable). In response especially to the unpredictability, firms simplify operations in several ways, leading to lower productivity and substantial revenue loss. While affected firms suffer, however, I draw on a general equilibrium model of competition to identify gains to their competitors. Policy implications of these results concern both the management of traffic on existing infrastructure, and the construction of new infrastructure.