Scholars have long recognized that imminent shifts in relative power may motivate declining states to initiate conflict. But what conditions exacerbate the risk posed by these anticipated power shifts? Building upon existing bargaining models of war, I show that larger initial power asymmetries increase the probability of preventive conflict. Theoretical extensions that account for certainty effects and variable costs of war, both of which are linked to initial dyadic power balances, drive this relationship. It follows that looming power transitions in which rising states approach and surpass parity, long considered war-prone scenarios, are not particularly problematic. Instead, the risk of conflict is greatest when preponderant powers confront conventionally weak but rising states. I test the theoretical predictions in the context of anticipated power shifts due to rivals pursuing nuclear weapons. Extensive empirical tests that relax assumptions employed in prior analyses of preventive conflict offer strong support for this contention. These results shed light on the underpinnings of many pressing contemporary interstate security issues.