All USB ports are backwards compatible, so what are you actually asking for? Even the latest USB-C ports can usually handle most ancient USB1 & 2 devices with simple cheap physical adapter to convert connection from USB-A cable from old device).
Your description brings an interesting consideration, but needs more details than provided to a meaningful recommendation:
- Issue: depending on devices and chipset, you can overwhelm (overload) a USB Root Hub (search in this forum for "USB RootHub" and/or M$' USBView tool for more info on this issue)
- Implication - number of ports is a too simplistic question.
You need to determine what devices will be using those ports (5 USB cameras? or 2 for keyboard/mouse, 3 for cameras. or ??? etc. And what resolution from camera will impact bandwidth implications. And then evaluate driver maturity (some vendors are junk and cause problems when adding multiple cameras from same vendor), and then motherboard chipset and number of USB Root Hubs, and ability to spread traffic over those hubs to avoid overload, *if* that applies to your situation... it does get technical. Audio interfaces tend to use a lot less bandwidth (small fraction) than video.
However, beware differing data paths for audio and video, due to manual sync considerations.
And then, depending on setup, a USB4 or Thunderbolt port into a hub (which can be small, may provide the ports (and at ~40Gb/s, still not get overloaded). Considering portability is a primary consideration, I'd look at a combination small keyboard/mouse that uses a single USB port. Or are you planning to control OBS via something like a streamdeck?
Depending on your setup, consider how you will control things. I started single monitor, but with multi-camera setup (and using PTZ ) I needed a 2nd monitor. With DisplayPort MST I can drive 2 monitors (or more) with single DP cable coming out of PC.
Streaming over WiFi will be a crap-shoot as best, and unless wireless network engineer level knowledge, Murphy's Law says likelihood of network glitches over WiFi goes up with importance of event. There are ways to avoid a problem, and use WiFi, but not using consumer stuff.
A GPU for encode offload helps lower CPU requirements. And there is a base graphics processing requirement for OBS. But you don't need a dedicated physical GPU. Depending on other workload/tasks, especially any Video inputs to be handled, an integrated GPU or even Intel QuickSync may be plenty. depends ...
So, to get a real answer and by that I mean not a flippant response that over-specs a system to be safe, or low-balls making unstated assumptions, you'll need a lot more info than what you've posted to get an answer that isn't overkill, or possibly under-powered. Also, be sure to indicate how long you expect to use the system (CPU) before upgrading? Is something like next-business-day onsite support important (not cheap, but available).
There are numerous 1Litre (mini/small form factor SFF) and NUCs that would most likely suffice. Depends on other requirements as I mentioned, and budget/expectations on being able to get more sophisticated down the road with same system (for example, adding color correction filters/processing, audio filters, etc). It is easy to start adding effects which greatly increase computational workload for real-time video encoding