AWS or a strong server for computing with Spark?

Question

Java

Hello, I will probably have a slightly abstract or holistic question.
The fact is that there is an idea to try to develop a service that will analyze the data. Accordingly, a constant question is spinning in my head, how to organize this business.
The approximate essence is as follows, the user comes, enters data, and counts me. As a result, calculations without BigDate can take some time, so I want to calculate as quickly as possible and give the user a result.
Therefore, if you choose AWS, how to calculate the billing for each user, dynamically increase resources as needed from the web interface, and so on. Is this all possible? How hard?
Or will it be okay to start with if you rent a serious server? For example, with 32-64 GB of RAM and pay a certain price every month.
Thanks for the thoughts, opinions, thoughts...