There already exist companies working with chips on "better" technology at this exact moment -- Avalon (110nm) and BFL (65nm). So far they have been more an example of how not to approach this business rather than a threat. In the short term, it seems it's quite obviously a case of "size doesn't matter, but what you do with it": proper management, execution and customer service weigh a lot more towards success compared to the technology of the chips.

In the long term, there will come a point when the utter gains in efficiency of better chips will make up for the incompetence of all these competitors -- and it is quite possible that a new and so far unproven company comes up with both the tech and the execution to "take the crown"; it's happened before in very competitive fields. At that point, ASICMiner should be ready with an answer -- but rushing towards the latest and greatest in process nodes might not be that answer. Keep in mind that the more recent/small a process node is, the more difficult and expensive it becomes to work with. The tools are more expensive, the companies providing wafers are fewer, chips take longer to design and any mistakes cost more in both time and money. Personally I believe keeping at least one or two generations "behind" the latest would yield a the optimal balance of cost vs. performance and power consumption, but of course that's something for ASICMiner to evaluate themselves.