In this work, we present ThermoHands, a new benchmark for thermal image-based egocentric 3D hand pose estimation, aimed at overcoming challenges like varying lighting and obstructions (e.g., handwear). The benchmark includes a diverse dataset from 28 subjects performing hand-object and hand-virtual interactions, accurately annotated with 3D hand poses through an automated process. We introduce a bespoken baseline method, TheFormer, utilizing dual transformer modules for effective egocentric 3D hand pose estimation in thermal imagery. Our experimental results highlight TheFormer's leading performance and affirm thermal imaging's effectiveness in enabling robust 3D hand pose estimation in adverse conditions.