Need both compute and graphics on cloud hosted Tesla cards.

Our application does graphics (OpenGL) and compute (CUDA). We also use CUDA/OpenGL sharing to do a portion of the interactive rendering. We need to be able to run code on cloud compute instances. We’d like to use the AWS P3 instance (P100 Tesla GPU). This would be a windows instance.

Is there a way to make the cloud hosted P100 Telsa support both graphics and compute? I know this has something to do with TCC vs WDDM mode. What are the steps one would take to make this happen? Does AWS give you enough control to set this up or is this something AWS would need to enable on their end?