After we get widespread ASICs (3-6 months time) then it is a battle for who can get more on the network ... and the same arguments apply all over again as for GPUs. i.e. can one or several huge installations compete with a distributed network of thousands of motivated individuals? ... go figure, the network wins every time. It is not about the particular type of hardware but what is the most efficient resource allocation mechanism ... and the answer every time is distributed network (for this kind of application).
Bottom line is that the scaling of costs for big, centralised computing centers is a greater power than distributed networks ... you can throw more money, people, hardware at the same site and return in compute power just doesn't go up like it does when those resources are distributed over many individuals on a network. Do the math on that then get back to me.
There is a valid fear of a savvy entity that sees BTC as a threat. If said entity decided it wanted to try and destroy BTC it could invest a few million dollars to protect hundreds of millions of dollars of future revenue. At the moment you could attack BTC legally by funding lobbyists or by attacking the network itself. Attacking the network, given a few days of reading about BTC, would logically happen via an ASIC of comparable efficiency to what is being released to consumers now.
Assuming Avalon made $0 USD on their first batch, I believe it was 300 units @ $1300/each. That's a paltry $390000 USD for 18 TH/s. In reality the fixed costs of the masks are likely a fraction of this and the chips are in the $1-10 USD/chip range. Pulling numbers out of my ass, throw $1-5 million USD in renting rackspace and producing 300+ TH/s to counter the expected production of BFL and Avalon for the next 6 months and you can effectively control the network. If you are wise you'll wait until you have an astronomical amount of hashing power and then flip the switch at one moment and double-triple the network hash rate. Guaranteed within hours BTC miners/developers will notice "oh shit, someone owns the network". Crash user trust in the security of the network and crash the value of BTC. Before the developers can change the protocol and get enough users over to the new network BTC will be trashed. Sure, other more complicated cryptocurrencies may crop up(LTC for example) but the damage is done.
THIS is what I fear and it'll be an issue until the cost to attack the network exceeds what I'd consider to be 'pocket change' for larger entities. At this moment if my net worth was sufficient that I could throw $2-3 million at a private ASIC endeavor as 'fun' I'd do it just to fuck with people. It would be supreme trolling, not casual forum trolling.