Is SSGI hardware accelerated like other forms of raytracing on RTX hardware?

If it is not, why is it not? Why would it not be the same kind of math but confined to a limited amount of screen space information?

No, it can’t be (unless you’re deliberately creating some ridiculous setup just to do so) and it’s fast enough that doesn’t need to be, it’s just the same principle as screenspace reflections or ambient occlusion, but now GI.