From cameras to displays, visual computing systems are becoming ubiquitous in our daily life. In particular, holographic near-eye displays promise unprecedented capabilities for virtual and augmented reality (VR/AR) systems. However, their underlying design principles have stagnated after decades of evolution. The image quality achieved by traditional holographic displays is limited, and algorithms for computer-generated holography (CGH) are slow. This shortcoming is due to a lack of co-design between hardware and software, importantly, impeding the delivery of vivid 3D visual experience.   This talk covers a family of Neural Holography advances that apply the unique combination of machine intelligence and physics to solve long-standing problems of computer-generated holography (CGH). With leveraging the advantages of camera-based optimization and neural-network model representation, we are able to deliver full-color, high-fidelity, 3D holographic images. Driven by trending machine intelligence, these hardware-software jointly optimized holography techniques can unlock the full potential of VR/AR/MR applications.