What determines whether militaries will defect from authoritarian incumbents during regime crises? Variance in military behavior in the Arab Spring has given rise to a debate around this issue. This article highlights weaknesses of the dominant explanation and develops an alternative account of military behavior in ‘endgame scenarios’. If militaries are politicized institutions that play a major role in regulating access to power under authoritarianism, they are more likely to intervene during normal times, but less likely to defect during mass uprisings. I quantitatively test this argument against data on military coups between 1975 and 2000 drawing on a new variable that allows me to explicitly model the impact of major regime crises. I illustrate the emergence of different forms of political–military relations and their consequences in the Arab Spring by drawing on evidence from Syria, Egypt, and Tunisia.