Abstract: This paper studies the risk of large deviations in GDP in the context of a general nonlinear production model. It derives the probability of extreme events conditional on the structure of the economy and the distribution of the shocks. Tail risk is driven by complementarities in production. Increases in interconnectedness in the presence of complementarity can simultaneously reduce the sensitivity of the economy to small shocks while increasing the sensitivity to large shocks. Tail risk is strongest in economies that display conditional granularity, where some sectors become highly influential following negative shocks. For a wide class of shock distributions, all crashes are identical, in the sense that they come with probability one from a particular combination of shocks, which also yields a sufficient statistic for crash risk. The analysis also characterizes what sectors are systemically risky (or conditionally granular): those that produce inputs for a large fraction of final production and have no close substitutes.