20 nm this year, 16nm next year (most TSMC customers are skipping 20nm since 16nm goes into production this year) and then it will take until 2016 at least before next shrink. If by magic Intel opens up their fabs to anyone with enough cash we could see 14nm during 2015 (not gonna happen unless Intel runs into cash/capacity utilization concerns). Intel alone spends more per year than the entire BTC network is worth on r&d (over 10b 2013) , a substantial part of that is directed towards fab tech. The costs of r&d and fab costs has risen exponentially for each new node and there is a good reason very few companies has their own fabs these days (and even fewer of them are on the bleeding edge).

Until the day btc mining becomes a multi billion dollar industry it will have little effect on the rest of the tech industry. The days of rapid improvements are over and soon we'll be stuck with the same 24~ month cycles as the rest of the industry. Sure there most likely are design optimizations that can be done, the question is who is going to fund a new chips designs on a bleeding edge process for (relative) small gains in a already saturated market.