Abstract Prey distribution and density drive predator habitat usage and foraging behaviour. Understanding ecological relationships is necessary for effective management in any environment but can be challenging in certain contexts. While there has been substantial effort to quantify human disturbance for some protected, deep‐diving marine mammals, there are virtually no direct measurements of deep‐sea predator–prey dynamics. We used recently developed techniques to measure deep‐water squid abundance, size and distribution within foraging habitat areas of deep‐diving Cuvier's beaked whales (Ziphius cavirostris) on and around a Navy training range where sonar is often used. Beaked whales are a management priority as both mortal strandings and sublethal disturbance have occurred in association with Navy mid‐frequency sonar. We found large differences in prey (squid) abundance over small horizontal distances. Highest squid densities occurred within a commonly utilized foraging area on the range. Much lower prey abundance was measured in adjacent, bathymetrically similar areas less commonly used for foraging. By combining prey densities with available behavioural and energetic data, we generate relativistic energetic assessments of foraging habitat quality. This provides a simple, yet quantitative means of evaluating fitness implications of spatial prey heterogeneity and associated consequences of disturbance. Synthesis and applications. Given the challenges deep‐diving predators face with limited foraging time in extreme environments, small‐scale prey heterogeneity can have substantial implications for foraging success. Our results provide fine‐scale data within neighbouring beaked whale foraging habitat areas commonly disturbed by sonars. These results have direct management implications and inform population‐level models of disturbance consequences with empirical data on the foraging ecology of these protected species. These issues have been at the heart of recent debate and litigation over spatial management and proposed sonar exclusion zones, which have previously been based entirely on indirect assumptions regarding habitat quality. While limited in temporal and spatial scope, our novel results provide the first direct ecological data to inform such applied decisions. They also highlight broader regulatory implications of different disturbance consequences in nearby areas and demonstrate the value of empirical, biologically based approaches to spatial management of marine ecosystems generally.