Early mainframe computers were probably the size of a nuclear sub power plant. And yet now we can place that much computing power into something smaller than a phone. Who knows how small we can made a nuke?
Presumably you meant nuclear reactor, not nuke (weapon).
If nuclear power plants had the same failure rate as subs do, no one would consider them. The Navy accepts a much higher level of risk than society does. They have to make that tradeoff to avoid the limitations of WW2 era subs.
Ultimately, energy produced is based on how much fission is going on. That requires an inherent space. It's not like the shrinking of ICs, or the lowering of voltage (and thus heat). The SMRs so far suggest the usual inefficiency as size shrinks.
I don't think there is any shortage of physics postdocs trying to find new methods for cold fusion, or energy from nothing. If they do someday figure out fusion, then hydrogen fuel becomes a reliable replacement of most of the battery in a car. It wouldn't matter that it takes 3x more energy than just charging a lithium battery.