Need to be careful of terms here.
Saving money is not the same as saving energy.
Just because at some place and time energy is less expensive in one form than another does not necessarily mean that using the less expensive form is saving energy.
Consider that electricity is mostly generated created from thermal sources. [Main exceptions that come to mind photo-electric, tidal and to some extent wind; just to head off the sticklers - hydroelectric power comes fundamentally from evaportive transport of water, i.e. solar heat.]
That nuclear plant probably heats some fluid, which is then used to spin a turbine to generate electricity. (Likewise gas or coal or oil fired plants.)
So yes, it may be possible to use higher efficiency boilers/etc. to capture more of the energy from the fuel in a power station than in a home, but you also have the losses in conversion to electricity and in transmission.
(Against the losses/energy required to transport fuel.) (i.e. it might be more efficient to eliminate the middle-man and just have a pile of U in the basement, keeping you toasty.
send it down high-voltage lines (% loss?) then use the computers as resistance heaters, or have the gas piped/trucked to your house and burned in your home furnace which is probably only 10-20% efficient?
A bit overly pessimistic on furnace efficiency, according to US Dept of Energy estimates:
High efficiency central furnace 97.0
Typical central boiler 85.0
Minimum efficiency central furnace 78.0"
Electricity distribution loss circa 7.5%
Of course, if you take into account something like heat pumps (where you use the energy not for direct heating, but to effect energy transfer), then your "cooking with gas."
Baseboard, resistance 99.0
Central heating, forced air 97.0
Central heating, heat pump 200+
Ground source heat pump 300+
[Don't know if there are heat pumps powered by other than electricity.]
Also, need to account for the energy involved in creating your computing heater. As noted elsewhere, because of the shorter lifecycle and complicated manufacturing processes involved, computers take significant energy to produce. So frequent new parts are out if one wants energy efficiency. (On the other hand, there is so much cool second hand gear out there.)
A case could certainly be made for thermostatically controlled second-hand computers to replace purely resistive electric spot heating. (In a mild climate here, much of the year my (tiny) office is kept adequately warm by body heat and one P3 with a CRT).
I have been wondering about this for a while - when you count the whole energy input (not just the commodity price of what you buy), what are the relative energy efficiencies of the heating sources?