There are two things that they may be speaking of: 1. Causing transient errors to crash the system and cause restarts that may take many minutes. 2. Actually damage the machines. I imagine that only a small fraction of the energy required to damage the hardware is necessary to introduce serious transient errors. IBM used to test the main frames as they installed them. They had to resist several inch sparks drawn between the machine and a one meter frame. Software diagnostics ran during the test. I think that few desk top machines would survive that. EMP (Electro Magnetic Impulse), a side effect of nuclear devices, is purportedly able to damage electronics over distances of many miles. Some weapons may be designed to enhance this. Ordinary high explosives can produce a scaled down result. EMP is strategic only because it damages electronics that are too far enough to be damaged by the blast. I suspect that high explosive EMPs are similar in this regard. The physics behind this is not abstruse. A significant part of the explosive energy can be turned into EMP whether the source is nuclear or chemical. How well it can be directed is probably highly classified. The "antenna" is vaporized in either case and dissipates much of the energy. The energy comes out in 10's of microseconds for high explosives and fractions of a microsecond for nuclear. I don't know how much it takes to fry an IC but judging from the wrist straps that are recommended for installing ICs I would guess that it is a fraction of a Joule. (I once discovered that a one Joule jolt really hurt.) Faraday cages attenuate EMP by the same factor that they attenuate secrets. I think that if a blast doesn't damage the cage then neither will the EMP. Of course the cage may survive but fail to protect the interior (insufficient attenuation). I don't know whether a cage sufficient for tempest is sufficient for EMP protection. Comm lines and power cables go thru the cage and cause problems here as well. Perhaps hefty surge protectors suffice here. Communications equipment outside the cage should at least be equipped to recover quickly upon transient error and not tear donw circuits. Normal error control can then hopefully compensate for the transient. (IP, ATM, Frame relay??) A large capacitor can discharge a lot of power in a short time without causing nearly so much commotion as an explosion. Discharging a one kg 25000 volt capacitor makes a lot of noise, however. I don't know how well it can be muffled.