Amicable social interactions can enhance fitness in many species, have negligible consequences for some, and reduce fitness in others. For yellow-bellied marmots (Marmota flaviventris), a facultatively social rodent species with demonstrable costs of social relationships during the active season, the effects of sociality on overwinter survival have yet to be fully investigated. Here, we explored how summer social interactions, quantified as social network attributes, influenced marmot survival during hibernation. Using social data collected from 2002 to 2012 on free-living yellow-bellied marmots, we calculated 8 social network measures (in-degree, out-degree, in-closeness, out-closeness, in-strength, out-strength, embeddedness, and clustering coefficient) for both affiliative and agonistic interactions. We performed a principal component analysis (PCA) to reduce those attributes to 3 affiliative (connectedness, strength, and clustering) and 4 agonistic (submissiveness, bullying, strength, and clustering) components. Then, we fitted a generalized linear mixed model to explain variation in overwinter survival as a function of these social components, along with body mass, sex, age, weather conditions, hibernation group size, and hibernation group composition. We found that individuals with stronger amicable relationships were more likely to die during hibernation. This suggests that social relationships, even affiliative ones, need not be beneficial; for yellow-bellied marmots, they can even be fatal.