In this talk we will explore how to measure and reduce resource costs of frontier AI, including compute, energy, environmental impact, and financial costs. The talk opens with an overview of AI's environmental impacts and introduces "AI-waste" as a lens for identifying where resources are spent unnecessarily. From there, it covers estimates of resource costs in frontier AI systems, practical approaches to forecasting energy consumption and carbon footprint, strategies for building more efficient AI systems from the outset, and a look at where Green AI is heading.