Even if you manage to steal a porta-reactor, you can't make a bomb with it
Given that there's a sci-fi feel about this plan, then, it's no surprise that it comes to us courtesy of DARPA, the Pentagon tech bureau which is always seeking to achieve regime change in the state of the art. The US warboffins, however, say that this is no longer the stuff of sci-fi.
Recent advances in high temperature materials science suggest that an investigation of compact, deployable nuclear reactor technologies may be timely. The goal would be to create a fieldable design that could be deployed to maritime and/or ground based forward operations to provide on-site power and fuel production capability in regions not connected to a robust grid and/or not easily accessible for fuel resupply. Preferred reactor designs would allow for several years of operation without refueling.
DARPA would like to see porta-reactor proposals able to carry a 5 to 10 megawatt electrical load in addition to producing 15,000 gallons of JP-8 or road fuel daily - enough to fill up a Chinook helicopter around a dozen times. They also want designs which are "inherently safe (negative temperature coefficient)", ie ones which can't suffer a runaway if damaged and melt their way down out of the base into a glowing tunnel in the Earth's crust.
As for size, DARPA doesn't offer any detailed requirements, but it does say that the reactor and its associated generators, fuel plant etc should be "readily deployable". That probably won't be too hard; existing submarine reactors in the same power range are only dustbin-sized.
The Pentagon tech-heads also aim to deal with proliferation fears by specifying that neither the reactor's fuel nor its wastes should be of any use in making weapons: thus, even if enemy forces managed to overrun a US base or steal its power'n'jetfuel plant they'd be no further foward toward making a bomb (or anyway no further than any nation with a civil nuclear programme).
Even so, given the general level of fear and misunderstanding that exists around nuclear technology, you have to say that the kit might never take the field with the landbased US military even if it succeeds technically. Naval deployment on ships which already have reactors would be much more plausible.
And, of course, for countries such as France or Switzerland which make serious use of civil reactors (which often produce a lot of waste power at times of low demand; reactors don't like being throttled up and down much) the fuel-making tech could offer major carbon cuts and freedom from fossil imports.
It'll be interesting to see how this pans out: though there's the usual caveat here that not many DARPA projects ever succeed. ®
*Methane is a hugely more powerful greenhouse gas than CO2, and one molecule of methane burns to make one of carbon dioxide: thus reducing its ecological burden by a factor of 25 as it does so.
This is why anyone who is truly concerned about the environment should always, without exception, light their farts. Failure to do so is colossally irresponsible in a global warming context.
**Reprocessed fuels, unlike freshly-mined uranium, are often of a grade that could also be used in weapons as well as reactors. The fear is that such material might be lost or stolen, therefore it should not be created. The USA follows this policy rigidly, and as a result has a very large waste stockpile. Other nations sometimes reprocess, but in general the cost - in large part resulting from the proliferation fears and associated precautions - are so large that it's generally cheaper to mine new fuel.