The extraordinary progress in AI over the last few years has been enabled, in part, by modern advancements in computing. Algorithmic ideas that had been around for decades have finally been brought to life thanks to Moore’s law and innovations in microprocessor and computing architectures. As Moore’s Law slows and data volumes explode we can expect to see new types of innovations emerge that will allow the rate of progress to continue. My presentation will cover state-of-the-art computing for AI, as it exists today, as well as a roadmap of innovations that will lead us into the decade(s) to come. This includes the importance of approximate computing (both algorithms and hardware), analog devices for AI, and quantum computing for AI. I will also highlight how these innovations bring forth both challenges and unique opportunities for the design automation community.