I analyze a model of coalition war termination in which continued fighting prevents power from shifting in an enemy’s favor but causes power to shift between partners. War ends in equilibrium when both across- and within-side distributions of power are sufficiently stable, but coalition politics can shorten or lengthen the war. When within-side power is shifting but the rising partner’s stakes in the war are low, the war ends “early,” before fighting can solve the underlying bargaining problem. When continued fighting strengthens the rising partner sufficiently, its declining partner acquiesces in continued fighting when intramural discord is costly or coalition exit is unattractive, causing the war to end “late,” well after the enemy’s power has been stabilized. I use the model to account for the early end of World War I and the late end of the Paraguayan War and then discuss implications for research design, war outbreak, and balancing coalitions.