Today's CPUs and GPUs are built upon the von Neumann architecture, which has powered modern computing for decades. However, this architecture faces fundamental limitations in achieving higher computational performance with greater energy efficiency. As a promising alternative, neuromorphic architectures, inspired by the neurons and synapses of the human brain, have emerged as a potential solution to overcome the von Neumann bottleneck. Yet realizing this vision requires breakthroughs on multiple fronts, including a fundamental understanding of neural computation, three-dimensional (3D) semiconductor architectures, and analog computing technologies. Can these challenges be overcome to develop neuromorphic chips that achieve the extraordinary computational efficiency of the human brain?