Adverse clinical outcomes for total disc arthroplasty (TDA), including subsidence, heterotopic ossification, and adjacent‐level vertebral fracture, suggest problems with the underlying biomechanics. To gain insight, we investigated the role of size and stiffness of TDA implants on load‐transfer within a vertebral body. Uniquely, we accounted for the realistic multi‐scale geometric features of the trabecular micro‐architecture and cortical shell. Using voxel‐based finite element analysis derived from a micro‐computed tomography scan of one human L1 vertebral body (74‐μm‐sized elements), a series of generic elliptically shaped implants were analyzed. We parametrically modeled three implant sizes (small, medium [a typical clinical size], and large) and three implant materials (metallic, E = 100 GPa; polymeric, E = 1 GPa; and tissue‐engineered, E = 0.01 GPa). Analyses were run for two load cases: 800 N in uniform compression and flexion‐induced anterior impingement. Results were compared to those of an intact model without an implant and loaded instead via a disc‐like material. We found that TDA implantation increased stress in the bone tissue by over 50% in large portions of the vertebra. These changes depended more on implant size than material, and there was an interaction between implant size and loading condition. For the small implant, flexion increased the 98th‐percentile of stress by 32 ± 24% relative to compression, but the overall stress distribution and trabecular‐cortical load‐sharing were relatively insensitive to loading mode. In contrast, for the medium and large implants, flexion increased the 98th‐percentile of stress by 42 ± 9% and 87 ± 29%, respectively, and substantially re‐distributed stress within the vertebra; in particular overloading the anterior trabecular centrum and cortex. We conclude that TDA implants can substantially alter stress deep within the lumbar vertebra, depending primarily on implant size. For implants of typical clinical size, bending‐induced impingement can substantially increase stress in local regions and may therefore be one factor driving subsidence in vivo.