Why have militarized interventions to curtail violence by drug cartels had wildly divergent results? In the past six years, state crackdowns drove a nine-fold increase in cartel-state violence in Mexico, versus a two-thirds decrease in Brazil. Prevailing analyses of drug wars as a criminal subtype of insurgency provide little traction, because they elide differences in rebels’ and cartels’ aims. Cartels, I argue, fight states not to conquer territory or political control, but to coerce state actors and influence policy outcomes. The empirically predominant channel is violent corruption—threatening enforcers while negotiating bribes. A formal model reveals that greater state repression raises bribe prices, leading cartels to fight back whenever (a) corruption is sufficiently rampant, and (b) repression is insufficiently conditional on cartels’ use of violence. Variation in conditionality helps explain observed outcomes: switching to conditional repression pushed Brazilian cartels into nonviolent strategies, while Mexico’s war “without distinctions” inadvertently made fighting advantageous.