I'm not sure I follow. If processing the block-chain is cheap enough that everybody can do it, I see no reason to depend entirely on centralized mines. The little guy can still participate, and there's bound to be more little guys than big guys, especially if bitcoin clients do a bit of mining themselves.
Sure, it's profitable and large-scale operations will probably arise, but that's distinct from centralization, per se.
Now we've come full-circle to my original thesis: this is a short-term phenomenon that I predict will slow and eventually reverse.
I'm still unclear as to why this shouldn't be the case.
> Also, computing power per watt requires ASICs to compete.
EDIT (since I can no longer reply): ASICs should only offset the point at which it becomes feasible for common-denominator computers to participate. I fear we may be talking past each other...
Bitcoin mining started on general-purpose CPUs. Then people discovered they could compute more hashes per watt on GPUs. Then people realised that a custom circuit optimised just to perform that calculation would be most efficient, starting with FPGAs and then burning those circuits into ASICs.
All mining that is feasible these days is done with specially-built ASICs, and even then the cost of mining 1 BTC (hardware and electricity) is hovering around 1 BTC.
No, centralisation proceeds because the hashing is subject to economies of scale.
If processing power increases, it increases for all players.