Interventions targeting early childhood development, such as investment in preschools, are often seen as promising mechanisms to increase human capital and to reduce the intergenerational transmission of poverty and inequality. This paper presents results from a randomized evaluation of a large scale preschool construction program in Cambodia, and indicates a cautionary tale. The overall impact of the program on a wide set of children?s early childhood outcomes was small and not statistically significant, and for the cohort with highest exposure the program led to a negative impact on early childhood cognition. Moreover, for this group, the intervention increased inequality as the negative impacts are largest for children of poorer and less educated parents. The results can be explained by the frequent occurrence of underage enrollment in primary school in the absence of preschools, stricter enforcement of the minimum age for primary school entry after the intervention, substitution between primary and preschool following intervention, and difference in demand responses of more and less educated parents to the new preschools. These results indicate that the design of ECD interventions needs to start from a good understanding of parental and teacher decisions pre-program. More generally, they show how implementation and demand-side constraints might not only limit positive impacts, but could even lead to perverse effects of early childhood interventions.