Organizations use hardware accelerators such as GPUs to dramatically improve the performance of AI/ML workloads that may run several hours or longer. IT teams have come to realize that these hardware accelerators are isolated islands—unable to be shared across many parts of the business.