All steps in estimating future climate impacts from emission scenarios are computationally expensive: running Earth System Models, downscaling and/or bias-correcting the outputs, and running process-based impact models. Altogether, these processes can take months. The latest evolution of reduced complexity climate models, or simple climate models, can project global climate from the latest emissions scenarios for tens of thousands of physical realizations in seconds. Novel methods are being developed to leverage the outputs from simple climate models to carry out risk assessments, and quantify climate impacts beyond the global mean temperature and even climate extremes. Concurrently, the latest advances in machine learning have enabled end-to-end simulation of climate dynamics at a fraction of the computing cost of physically-based systems. Impacts may be spatially resolved, enabling policy-relevant analyses to be carried out based on emissions scenarios which have never been run through fully-coupled Earth-system models, such as Network for Greening the Financial System (NGFS) scenarios. Applications of impact emulation extend to economic and integrated assessment models of climate change. With the rise in application of machine learning for Earth system model emulation and downscaling, this session aims to bring together research on statistical, physical and hybrid emulators with a focus on climate impacts.