Students of international relations have long argued that large and rapid shifts in relative power can lead to war. But then why does the rising state not alleviate the concerns of the declining one by reducing its expected future power, so that a commitment problem never emerges? For example, states often limit their ability to launch preemptive attacks by creating demilitarized zones, or they abandon armament programs to avoid preventive wars. In a model of complete information, I show that shifts in power never lead to war when countries can negotiate over the determinants of their power. If war occurs, then, it must be that negotiations over power are impossible or too costly. I then show how third parties, domestic politics, and problems of fungibility can increase the costs of such negotiations, and hence lead to war, even under complete information.