HPC workloads traditionally run on a fixed-size in-house cluster. For users, this means waiting in queue and constraining their jobs to fit within the bounds of the cluster. But what could happen if the end user got the resources they needed at the scale required? Azure has the capabilities and ease-of-access to accelerate innovation by giving HPC and AI jobs the right resources on demand. This talk covers seven lessons from a decade of HPC in the cloud and how the scale, flexibility, and elasticity offered by Azure can transform how HPC is done.