I agree with this, but unit-value is not going to be hard to enumerate on the fiat side. I suggest you go with the estimates in the original ppt linked to in this thread.
Are you referring to
http://weis2012.econinfosec.org/presentation/Breuker_presentation_WEIS2012.pdf ?
I'd appreciate if you could explain the ppt / pdf linked earlier. It looks fairly strange to me. The assumption there is to use 0.3% of all USD payment transactions as budget in electricity for the PoW network. He arrives at ~17600 Petaflops of power, so around 1385 Terahash of SHA256 compute power (according to the ratio on bitcoinwatch.com). While all other arguments of the presentation stay valid with BFL ASIC technology rated at 1W/GH it should be possible to build this size of network at 1,4 MW, give it 2,8 MW for the hosts computers involved. He states himself that interest in PoW would trigger innovation - well it already has.
Which other conclusions does the paper arrive at that I'm missing or are we talking about another paper?
There is the really important point that if you reduce electricity costs and replace them with hardware costs, you are just substituting one form of waste for another.
I think the most appropriate comparison is between bitcoin and electronic banking. After all, bitcoin is not a replacement for in person exchanges of cash. (Though we could print bitbills which would seem to be identical to cash and probably have the same electricity requirements)
If I swipe my debit card, how much electricity do I use?
If I use my smart phone to pay with bitcoin, how much electricity do I use?