Deep learning is all the rage, and this enthusiasm is most evident in the explosion of startup companies leveraging this revolutionary model of computation. Training and inference of deep neural networks is fundamentally intensive in both computation and memory demands, so the spread of deep learning may have profound impact on the silicon world. In this talk, I analyze the state of the deep learning startup universe, with particular attention to three distinct dimensions of the change - use of deep learning in mainstream EDA, the potential for EDA-like tool innovations in neural network design, and most especially, the design of new silicon platforms for deep learning. The on-going renaissance in chip startups reveals important trends in computation structures, memory system, parallelism, software environments, business models and venture funding. I conclude with a roundup of the startups particularly worth watching and the emerging formulas for success.