This filies contain the results supported in Rubiano-Buitrago, López-Goldar et al., "Macroevolution of plant defense favors toxin potency over concentration". We found how antagonistic coevolution has long been predicted to leave a signature of escalating defenses in host species. Opposing this classic theory, we previously reported phylogenetic declines in concentrations of cardenolide toxins in milkweeds. Accordingly, we hypothesized a trade-off between toxin concentration and potency, such that plant species with lower investment in toxin abundance show higher potency per unit toxin. Leaf extracts from >50 Asclepias species were tested in vitro against the monarch butterfly’s Na+/K+-ATPase, cardenolides’ physiological target, revealing a trade-off between cardenolide concentration and potency. Machine learning identified metabolomic features associated with potency, revealing triterpenoids that may have a combinatory effect against the butterfly’s enzyme. Finally, we found that potency phylogenetically increased over time. Results support the hypothesis that, in the face of costly defenses and sequestering herbivores, milkweeds evolved low concentrations of more potent toxins against herbivores. Coevolution may generally drive discordant investment in concentration versus potency of chemical defense.