Simultaneous interpreting is regarded as one of the most difficult and stressful types of activity. Simultaneous interpreters work in a severe time deficit and have to absorb much information per unit of time, which means that different cognitive processes (e.g., memory, attention, thinking, perception) are engaged concurrently. This leads to a severe cognitive load, which is sometimes compared with that of pilots. In the current study we investigate how the increase in cognitive load, which happens throughout time, affects interpreters’ performance. This surge, we expect, will be observed on two language levels: verbal and nonverbal (gestural). The analysis is based on 10 videos of simultaneous interpreting of a lecture about biodiversity from English (L2) into Russian (L1), approximately 10 minutes each. The results of the study show an increase in speech disfluencies on the verbal level and distribution of various gesture functions on the nonverbal level. In particular, verbal disfluencies are exteriorized in fillers, draggings and truncations, and their number increases during the interpreting. Along with disfluencies we observed the rise of the co-speech gestures, e.g., adapters and pragmatic gestures, that help maintain control over the process of simultaneous interpreting, structure the output and reduce the cognitive load and stress experienced by participants while performing the task.