Hi Nodemansland, welcome! As is often mention around the forum, to make returns on this project you need to be in it for the long term. You will make very little in the first year. So it is suggested you don’t use a dedicated machine for running nodes as your operating costs will most likely exceed your income. It is best (for profitability) to run on hardware that is already being used for something else.
If you are interesting in running the project long term, the way I would do it is to spin up as many nodes as you have public IPs and start them small with the ability to expand them later. Keep in mind that the public IPs need to be on different /24 subnets to actually make a difference as the network treats all nodes on the same subnet as one node for ingress purposes (example with 5 nodes on the same subnet will collectively receive the same amount of ingress traffic as 1 node). With a single IP, the main reason to run multiple nodes is if a HDD/node is full and you want to offer more space.
The advantage to starting several nodes now, even if you won’t be using all of the space, is that they will all be vetted and pass their holdback period at the same time, so in the future as the network gains clients and traffic increases, you could already have several nodes ready to go.
I believe it was calculated that the maximum practical size of a node (when the network was busy with test data) is around 40 TB. Near this threshold the deletes will outpace ingress and growth will stop. I think it was said that at current traffic levels that threshold is around 20 TB. Of course this is all based on historical trends, and what happens in the future could and likely will be different.
–edit-- Indicated /24 subnets as thoughtfully noted by @Pac below