What I was referring to by asic product life cycle was the time to next "upgrade" as we are just getting started with asics and assuming we are here for the long haul(I hope) then we will be seeing a march down the nanometer ladder with ever more powerful asics.
There are several obvious precedents in history.
We have the scramble for the desktop, with MIPS, CMOS, Zilog, HP, IBM, DEC, Intel, Motorola, NatSemi, SUN, and others, wrestling for various slots of the market with different ISAs. People pretty much settled on x86 for the desktop, with the rest getting pushed to the margins above and below the consumer/business desktop, with a few consolidations (HPPA and Itanium, PPC/POWER, etc.).
Because of a number of factors (mainly Microsoft), the x86 PC platform emerged and a number of variants of ISA compatible clone chips, some licensed and some not-so-much so, were developed by a range of competitors. AMD, Harris, Cyrix, IDT, Transmeta, VIA, and others have wrestled in this space, to various degrees of success. Some of that continues today, but we see elimination and consolidation in the space. It's basically Intel and
AMD.
Other examples of innovation, starvation, and consolidation include mobile platforms, computer audio subsystems, and various GPU cycles, at the high (SGI) and low (Nvidia, ATI, 3dfx, S3, Paradise, Matrox, etc.) ends. Success has pivoted on application support and maintaining position within the pack.
I suspect we'll see the same sort of thing with BTC ASICs. We're seeing an evolutionary step with the first gen, from new architectures to masked FPGA descriptions that are burned in silicon.
What happens next can follow a number of paths.
First is the evolutionary migration to smaller transistor sizes, which will permit faster, lower-power operation, higher hasher density per die, etc.
Second, macro-scale architectural innovations may be applied to gain much more significant advantages.
Third, software support will be critical. A great piece of hardware is only as good as how much power it can put onto the network.
None of these may happen (I'm looking backward to point forward) or all of them may happen. They're expensive. Generally, a leader will emerge at some point. Because software makers usually want to target the largest market, you'll start to see less-supported platforms either adopt the interface (if not the architecture) of the leader and attempt to differentiate themselves somehow, whether it be price, clock speed, improved availability, or marketing. Those who don't follow consolidation in such a small market will exit the market.
While I'm interested in seeing what the first generation does, it's really not that important. New players may emerge or others may fail completely out of the gate. No one has even demonstrated a prototype yet.
What I am excited about (ok, strong language) is what generations two and three look like. Once everyone has an ASIC in their hands and the market is validated, it's going to get sexy. I would say that any of the current ASIC makers who don't have their next gen chips in the pipeline (hah!) will probably not survive, even if it's just a process improvement "tock".
