Despite their indispensable role in food production1,2, insect pollinators are threatened by multiple environmental stressors, including pesticide exposure2-4. Although honeybees are important, most pollinating insect species are wild, solitary, ground-nesting bees1,4-6 that are inadequately represented by honeybee-centric regulatory pesticide risk assessment frameworks7,8. Here, for the first time, we evaluate the effects of realistic exposure to systemic insecticides (imidacloprid, thiamethoxam or chlorantraniliprole) on a ground-nesting bee species in a semi-field experiment. Hoary squash bees (Eucera (Peponapis) pruinosa) provide essential pollination services to North American pumpkin and squash crops9-14 and commonly nest within cropping areas10, placing them at risk of exposure to pesticides in soil8,10, nectar and pollen15,16. Hoary squash bees exposed to an imidacloprid-treated crop initiated 85% fewer nests, left 84% more pollen unharvested, and produced 89% fewer offspring than untreated controls. We found no measurable impact on squash bees from exposure to thiamethoxam- or chlorantraniliprole-treated crops. Our results demonstrate important sublethal effects of field-realistic exposure to a soil-applied neonicotinoid (imidacloprid) on the behaviour and reproductive success of a ground-nesting solitary bee. To prevent potential declines in ground-nesting bee populations and associated impoverishment of crop pollination services, soil must be considered a possible route of pesticide exposure for bees, and restrictions on soil-applied insecticides may be justified.