Awesome! This question bears directly on the intersection of thermodynamics, computational complexity theory, economics, and identity instantiation, and hence I cannot resist answering.
The question boils down to the relave scalability of different effects with respect to each other:
1) How much computation can you produce per unit energy?
2) How much utility (or how many happy-people-equivalents) can you produce per unit of computation?
If reversible computing is possible, then the answer to 1) is that you can produce as much as you want, producing arbitrarily small energy losses (due to the inability to achieve perfect irreversibility e.g. due to friction in implementation.)
This would render pessimism about 2 irrelevant, it seems.
Edit: And furthermore, if we accept the notion that a virtual being's subjective experience of time depends soley on the program that instantiates them (e.g. a species of substrate independence that is widely accepted here), then such beings would subjectively live forever, even if reversible computing is constrained by having to be arbitrarily slow. (Although people would have to accept uploading themselves, of course.)
A dialogue discussing how thermodynamics limits future growth in energy usage, and that in turn limits GDP growth, from the blog Do the Math.
I think this is quite relevant to many of the ideas of futurism (and economics) that we often discuss here on Less Wrong. They address the concepts related to levels of civilization and mind uploading. Colonization of space is dismissed by both parties, at least for the sake of the discussion. The blog author has another post discussing his views on its implausibility; I find it to be somewhat limited in its consideration of the issue, though.
He has also detailed the calculations whose results he describes in this dialogue in a few previous posts. The dialogue format will probably be a kinder introduction to the ideas for those less mathematically inclined.