Using a unique dataset of trades and limit orders for S&P 500 futures, we decompose the aggregate risk into a component driven by the impact of net market orders and a component unrelated to net orders. The first component—flow-driven risk—is large, accounting for approximately 50% of market variance, and it is not transient. This risk represents the joint effect of net trade demand and the price impact of that demand—i.e., illiquidity. We find that flows are largely unpredictable, and lagged flows have no price impact. Flow-driven risk is time varying because price impact is highly variable. Illiquidity rises with market volatility, but not with flow uncertainty. Net selling increases illiquidity, which amplifies downside flow-driven risk. The findings are consistent with flow-driven shocks resulting from fluctuations in aggregate risk-bearing capacity. Under this interpretation, investors with constant risk tolerance should trade against such shocks (i.e., “supply liquidity”) to achieve substantial utility gains. Quantitatively accounting for the scale of flow-driven risk poses a major challenge for asset pricing theory.