Cloud GPUs have become crucial hardware accelerators for training and running neural networks. The popular graphics processing unit (GPU) is the most widely used for these purposes, but serverless GPUs are increasingly gaining popularity. Cloud GPU vendor pricing has been assembled in tables that are filterable to the user’s liking and includes GPU cloud servers and serverless GPUs. A GPU comparison chart is included for those selecting GPUs for their needs. The raw data can be found in a CSV on GitHub. Serverless GPUs are a newer technology, so there is an expectation of growing pains and bugs.
https://fullstackdeeplearning.com/cloud-gpus/