It is still profitable, since no one is willing to opensource the design of the chip. And I think it will be profitable for a long time...
Whats open sourcing have to do with it? Of course selling these chips now is extremely lucrative.
As for how long, it only depends on how much these companies can produce. Because they will produce as much as they can for as long as its marginally profitable, and they will sell (or use) anything they produce. Price and difficulty explosion will just be a result of those.
How fast can this go? For some perspective, TSMC alone does >1000 28nm wafer starts per day. One wafer equals ~70TH using hashfast's numbers. So one day of TSMC production gives 70PH worth of bitcoin asics, which when deployed, might be enough to make it no longer profitable to produce more asics as electricity cost might exceed mining revenue at that point. There are at least a dozen other fabs.
Of course asics have to be packaged, mounted, miners assembled and shipped/deployed. Some/many companies will falter there, but it only takes one to get it right. Its entirely feasible by next summer, prices will have plummeted so far, NRE's for new designs cant realistically be recovered and by the end 2014, we might see stagnation in the network when electricity costs put a limit on the hashrate.
Let's see what will happen. Also, that means, the bitcoin becomes the new driving force for the semiconductor industry, while it is making 'great' contribution to a greenhouse environment:) The flaw of the bitcoin system is that a lot of energy is wasted just for getting a control of inflation. Actually a better system can make better use of these energy.