The relationship between nutrient intake and wealth of poor households continues to be an issue of huge policy relevance. In this paper, we contribute to the ongoing debate on the nutrient-income elasticity using a sample of ultra-poor households with orphans and vulnerable children (OVC) in Kenya. To estimate the nutrient-income elasticity for these households, we employ panel data techniques that enable us to tackle measurement error and simultaneity bias. In addition, we use semi-parametric panel data models to address nonlinearities. For most of the nutrients considered, we find that income elasticities are significantly different from zero but below unity. Caloric intakes turn out to be less income-inelastic than macro and micro nutrient intakes.