It's a guess. Last time he saw his PC it was on and working. Now it is off and not working.

Horowitz and Hill postulated (*1) that the most unreliable components in an electronic system would be (in order):
1. Cables and connectors.
2. Switches.
3. Potentiometers and trimmers.

I think that the empeg has borne out their postulation rather well, assuming that we class the rotary encoder in the same class as a pot or trimmer, which I believe is reasonable.

But I also think that H&H didn't really go far enough in that chapter. Admittedly at the times that AofE was written and updated, the technology was transitioning from a mix of wire-wrap and PTH pcbs to a mix of PTH and SMT pcbs, and the resulting power densities of the era were far lower than those seen today, but they did actually discuss cooling and the use of fans just prior to that postulation. They didn't go the extra step though - they didn't say anything about _electromechanical_ reliability, eg fans, (or generators and other motors). Ironically they did state that, "in any cooling situation it is a good idea to design conservatively; failure rates for electronic components rise dramatically when equipment is operated hot." (*2)

If you asked Slashdot readers (!) what failed in a routine manner in their PCs most often, I am confident that you would get 50%+ of the answers involving electromechanical devices; Hard Disks, CD/CDROM/DVD drives and fans, and another 30% saying PSU (without knowing why).
That would also sit with Enterprise Server product lines - why do Enterprise servers have hot-plug disks and hot-plug redundant PSUs? Why not hot-plug network cards?
It would also match my experiences pretty well too - my linux server has run 24x7 for over 3.5 years now with only the occasional reboot. In that time the CPU fan has been replaced twice IIRC (maybe 3 times..), and the PSU has been replaced once. It has also had 2 new drives (RAID1, constantly logging so never spun down), but the old drives hadn't failed - I simply ran out of space and it was as cheap & easy to replace them completely than to augment them (which would have required more IDE channels). The only other death in the family has been a mouse, very recently after 40+ months of dedicated service.

I don't believe that the CPU is to blame (other prime heat candidate). Most PCs will crash long before the CPU dies even if the fan dies, and a crashed CPU doing nothing doesn't usually consume or dissipate much energy. Most modern CPUs have thermal protection built in anyway. (This is assuming of course that the heatsink has not become dislodged somehow which could possibly cause the temperature gradient to exceed the protection's ability to deal with it.)

I'm going to discount the PCs power switch as a cause - ATX only uses a momentary switch to signal a power state change, and if it failed it would not shut the machine down. (I suppose that if a lightening strike caused a power outage then a failed switch would prevent the PC being turned on again, but most people would mention that they suspected a power outage may have occured. Besides, I just don't see it being the most likely cause.)

Furthermore, he stated that he had tried to turn it on "to no avail". I would guess from this and the topic title that the machine is _completely_ dead, ie no LEDs, no fans etc as most people would state if it were otherwise. That would suggest that the PSU is not turning on at all.
ATX1.1 specs dictate that the PSU turns on when signal PS_ON is pulled to TTL low, ie 0V, and turns off when PS_ON is pulled high or allowed to float. (Pulled up in PSU). The circuitry required to do this isn't exactly rocket science, even allowing for adding WOL, WOM, Keyboard Wake functionality. Since this signal basically carries little current, this circuitry shouldn't really be prone to overheating either.

I suppose that H&H would suggest checking cables and connectors, ie power lead (it didn't get unplugged by the cat by chance?), ATX motherboard connector and the ATX power switch connector. But connectors are far more likely to fail when they are prone to connect/disconnect cycling, or noticeable physical vibrations. I don't see that being the case here.

Finally I suppose that the PSU could be locked in a shutdown mode due to a short circuit on one of the power lines (+12V, -12V, +5V, +3.3V or +5VSB). But again, this isn't the kind of failure I'd expect to see hit suddenly. This shutdown mode can only be cleared by cycling AC power - I'd expect to see it kick in again as soon as the power switch is pressed, probably with a gentle click or whir from the PSU.

But yes, it is just a guess.

*1 Art of Electronics, 2nd Ed, Chapter 12, Pg858.
*2 Also Pg858.

I am obviously not worthy, Oh Great One!!!

I save those kind of accolades for people who do crazy stuff like designing a radical in-car entertainment device 5 years before all the sheep do, or putting binocular remote vision on an RC car. But thanks anyway.