TensorFlow on three platforms - first impression:
Amazon Web Services (AWS) - does require use of Unix script, and if running from Windows, additional download of utility PuTTY for ssh connections (geek speak for cybersecurity key).
https://www.datacamp.com/...learning-jupyter-aws Google Cloud Platform (GCP) - much the same as AWS, and no PuTTY download required.
https://towardsdatascienc...-15-min-61e16da34d52 Microsoft Azure - here is step by step guide and setup Ubuntu (Unix) server
https://blogs.msdn.micros...-step-by-step-setup/ Assuming we seek min hassle, no-geek speak (beyond couple of Unix commands as documented), and shortest path to TensorFlow, I recommend GCP for getting started. (It so happens, among the three, Google offers new sign-up the highest free credit, 50% more in nominal dollars.)
The Dark Horse - Microsoft Azure Machine Learning Studio, which is great for: 1) drag-n-drop GUI setting up (without programming), 2) visual comprehension (flowchart like) of machine learning tasks, and 3) no Unix commands! Current default Windows-based configuration does not support TensorFlow. When it does, it is a serious contender on ease of use. Do give it a try and play with staple ML classifications and decision trees using data examples supplied.
https://studio.azureml.net/ Feel free to share you own TensorFlow experience @TIG, so we all learn and get better at it fast.