then perhaps orders-of-magnitude more electrically efficient, since the electric consumption will be primarily in the computation and not in the memory. And computation can be radically optimized on an ASIC.
As I predicted when I wrote in the Monero thread about this the other day, it appears the white paper didn't consider electricity as the most important factor. Duh.
The thing with memory hardened vs computationally bound is, on a card like an r9 290, you could get something like 175w power use for something not memory hardened (keccak), while on a memory hardened algo that's also obviously using lots of computation (any scrypt type derivative), power use would be through the roof at something like 250w+. GPUs seem very power inefficient in terms of dealing with memory configuration parlor tricks, so if they ever do get implemented in ASICs, the power consumption advantage is always enormous.
Scrypt is
not memory bound on GPUs. That is precisely the problem. Using the lookup gap technique it became possible for scrypt's (reduced) scratchpad to fit in local memory and not require scratchpad access on every iteration, so it becomes highly parallel and somewhat CPU bound.
Allegedly the equihash algorithm is provably resistant to such tricks. Still, it does have a compute component, and that could potentially be made more power efficient. Possibly the memory access pattern could also be exploited, since it is (I think) not necessarily random like with a cryptographic scratchpad-type algorithm.