The obligate mutualism and exquisite specificity of many plant-pollinator interactions lead to the expectation that flower phenotypes (e.g., corolla tube length) and corresponding pollinator traits (e.g., hawkmoth proboscis length) are congruent as a result of coevolution by natural selection. However, the effect of variation in flower morphology on the fitness of plants and their pollinators has not been quantified systematically. In this study, we employed the theoretical morphospace paradigm using a combination of 3D printing, electronic sensing, and machine vision technologies to determine the influence of two flower morphological features (corolla curvature and nectary diameter) on the fitness of both parties: the artificial flower and its hawkmoth pollinator. Contrary to the expectation that the same flower morphology maximizes the fitness of both plant and pollinator, we found that the two parties have divergent optima for corolla curvature, with non-overlapping fitness peaks in flower morphospace. The divergent fitness optima between plants and pollinators could lead to evolutionary diversification in both groups.