PeerTube should run happily on a virtual machine with 2 threads/vCPUs, at least 1 Gb of RAM and enough storage for videos. In terms of bandwidth, a lot will depend on which PeerTube instances you federate with and what your relation with them is (more about that below).
If you also to contribute to the growing of PeerTube, you can participate in its funding here: <a target="_blank" rel="noopener noreferrer" href="https://soutenir.framasoft.org/en">https://soutenir.framasoft.org/en</a>
If you have any questions, feel free to use our forum: <a target="_blank" rel="noopener noreferrer" href="https://framacolibri.org/c/peertube">https://framacolibri.org/c/peertube</a>
If you want to support PeerTube, you can <a target="_blank" rel="noreferrer noopener" href="https://soutenir.framasoft.org/en">support Framasoft with a donation</a>, but also by helping others to discover and <a target="_blank" rel="noreferrer noopener" href="https://joinpeertube.org/">learn more about PeerTube</a> and our projects: sharing is caring!
Except for video transcoding, a PeerTube instance is not CPU bound. Neither Nginx, PeerTube itself, PostgreSQL nor Redis require a lot of computing power. If it were only for those, one could easily get by with just one thread/vCPU.
You will hugely benefit from at least a second thread though, because of transcoding. Transcoding is very cpu intensive. It serves two purposes on a PeerTube instance: it ensures all videos can be played optimally in the web interface, and it generates different resolutions for the same video. PeerTube support for offloading transcoding to other machines is <a target="_blank" rel="noopener noreferrer" href="https://github.com/Chocobozzz/PeerTube/issues/947">being discussed</a>, but not yet implemented.
It's best to contact and talk directly with hosting providers, to understand their business model, vision, etc. Because only you can determine what makes you trust such or such host, and thus entrust your videos to them.
1/2 GB of RAM should be plenty for a basic PeerTube instance, which usually takes at most 150 MB in RAM. The only reason you might want more would be if you colocate your Redis or PostgreSQL services on a non-SSD system.
