OBJECTIVE To compare the extramaxillary approach with the widely used intrasinus approach via finite element method. METHODS A unilateral three-dimensional model of the craniofacial area surrounding the region of interest was developed using computed tomography image datasets. The zygomatic implants were modeled using three-dimensional computer-aided design software and virtually placed according to the described techniques together with one conventional implant and a prosthesis. The bone was assumed to be linear isotropic with a stiffness of 13.4 GPa, while the implants were of titanium alloy with a stiffness of 110 GPa. Masseter forces were applied at the zygomatic arch, and occlusal loads were applied to the surface of the prosthesis. The stresses and displacements generated on the surrounding bone and within the implant due to the simulated loading configuration were analyzed. RESULTS The bone-implant interface and zygomatic implant body for the intrasinus approach produced 1.41- and 4.27-fold higher stress, respectively, compared with the extramaxillary approach under vertical loading. However, under lateral loading, the extramaxillary approach generated 2.48-fold higher stress than the intrasinus at the bone-implant interface. The zygomatic implant in the extramaxillary approach had twofold higher micromotion than those with intrasinus approach under lateral loading. CONCLUSIONS No one technique was found to be superior; however, if lateral loading is used, the intrasinus approach is the most favorable for the rehabilitation of severely atrophic maxillae.