Although cryo-electron microscopy (cryo-EM) of biological macromolecules has made important advances in the past few years, the level of current technical performance is still well below what the physics of electron scattering would allow. It should be possible, for example, to use cryo-EM to solve protein structures at atomic resolution for particle sizes well below 80 kDa, but currently this has been achieved only for particles at least 10 times larger than that. In this review, we first examine some of the reasons for this large gap in performance. We then give an overview of work that is currently in progress to 1), improve the signal/noise ratio for area detectors; 2), improve the signal transfer between the scattered electrons and the corresponding images; and 3), reduce the extent to which beam-induced movement causes a steep fall-off of signal at high resolution. In each case, there is substantial reason to think that cryo-EM can indeed be made to approach the estimated physical limits.