It's a question of what you're spending power on. When you cool something there's the action of pumping heat away from it, and then the issue of how easily heat

It's a question of what you're spending power on. When you cool something there's the action of pumping heat away from it, and then the issue of how easily heat can migrate back in. So you're only really fighting the inefficiency of your insulation, and insulation can be very, very good (at the extreme end you could magnetically balance something in a vacuum chamber and then chill it - it'll take a very long time to heat up by radiative emission from the walls (double wall insulated cups use this property though my take is they're just gas filled - same idea though). Essentially the ongoing cost of keeping something cool is purely a function of heat load making it through insulation, and in all cases that's going to be much lower then generating heat in the system - the problem is that engineering and complexity wise, it's non-trivial and failure scenarios can be bad - i.e. an SC magnet quench can be very bad since you amongst other things might immediately boil your coolant. This happened at the LHC during commissioning.