Information transfer between semantic boundaries incurs thermodynamic costs beyond those captured by Shannon theory or Landauer's principle. We introduce Interface Friction as a quantitative framework for these translational energy losses, analogous to mechanical friction at physical boundaries. Through analysis of biological, computational, and organizational systems, we demonstrate that interface friction accounts for ~80% of total information processing costs in complex hierarchical systems. This work provides both theoretical foundations and practical metrics for designing low-friction information architectures.