This paper investigates how the asset-return variance risk premium changes leverage. I find that the premium lowers leverage by increasing risk-neutral bankruptcy probability and costs in a model where asset returns have stochastic variance with risk premium. Empirically, the model calibrations verify significant reduction in optimal leverage, closer to observed leverage than the model without the premium. In model-free regressions, I also document negative correlation between leverage and the variance premium. The most negative correlation is among investment-grade firms with low asset beta and historical variance but high variance premium because their assets have high exposure to market variance premium.