HPCaaS vs On-Premise vs Public Cloud - a comparative outlook
One of the obstacles for large on-premise clusters are the requirements for high-end data centers in addition to the challenge of ever-increasing energy costs.
Large cloud providers have increased their focus on HPC and added more specialized HPC services to their offerings. They leverage the success of their cloud computing ecosystem and are perceived by some to have the most agile and flexible solution available; allowing you to pay on a per-use basis and to adjust your capacity on the fly to meet your needs. But is this really a cost effective long-term strategy for running AI workloads or simulations that can benefit from capacity at scale?