We challenge the common assumption that current illiquidity is the best proxy for future market conditions by developing a forward-looking framework that forecasts the full distribution of U.S. corporate bond illiquidity using gradient-boosted trees. From these forecasts, we construct expected illiquidity (EI) and expected tail illiquidity (ETI), capturing downside liquidity risk. Relative to current illiquidity, EI reduces forecast errors by 21\% and, in yield-spread regressions, exhibits a coefficient up to four times larger. ETI further strengthens this link. In cross-sectional asset-pricing tests, a simultaneous one-standard-deviation increase in both measures commands a 1.53\% annual risk premium and 0.57\% alpha.