The first workshop on Neuromorphic Architectures (NeuroArch) aims at exploring novel ideas and research opportunities in design, programming, and application of neuromorphic and brain-inspired accelerators. In the current realm of processor design, where energy and power constraint has shifted the designs toward heterogeneity, hardware neural networks are emerging as candidate accelerators with attractive characteristics and broad application scope. In addition to the power-efficiency and fault tolerance of neural accelerators, we are at the junction of time where: 1. As technology scales down to the atomic levels, the increasing process variability causes the designers to pay a high tax in performance and efficiency to provide fault-free designs; the intrinsic robustness of neural networks may lead to fault-tolerant accelerators. 2. Novel neural network algorithms such as Deep Belief Networks outperform many alternative machine learning algorithms across a broad set of applications. 3. Significant progress in neuroscience sheds light on the operating principles of biological neural networks, which can thus be partially replicated in hardware. 4. The landscape of computing has changed toward providing a more personalized and more targeted experience for the users, thus increasing the importance of applications that require learning. Therefore, we believe it is imperative and timely for the computer architecture community and the design of next generation computing systems to explore and research neural models of computing.