As I said, you cannot mine Bitcoin with an Nvidia or a AMD GPU or any retail components really. You need to buy ASIC's. If people are buying them for mining, it's not for mining Bitcoin.
Building an ASIC is very expensive, though there are prototype programmes that allow developers to share wafers with others and to reduce the cost. It also requires most expensive tools and a great deal of know-how, one mistake and the chip won't work. Then if you need to adapt the application for a new type of bitcoin that was not foreseen, or the key length or whatever, you have to make a new chip, you can't reprogram them (hence the name Application-Specific IC).
So I'm not convinced it's a viable solution except if you do it on a massive scale and happen to work in a company that develops ASICs for a living (that's what you say in your other post, just echoing that here).
But I think you can still use FPGAs for that. Most companies developing ASIC chips use them for prototyping, but there are many other companies developing only on FPGA because products which a lower production volume use that technology (higher cost by unit, but much lower NRE). Those boards are largely underused and it's easy to let them run a bitcoin application at night or when they're not used for testing.
Even as an individual, you can buy those boards and the development tools for an affordable price. I wouldn't say it's cheap though, but it's 4 digits instead of 6
Since Intel bought Altera, there might be CPUs with an integrated FPGA that could serve that purpose, if the FPGA is big enough (same for AMD/Xilinx). Perhaps a future alternative.
EDIT: like
here, though this doesn't sound realistic and such a patent is probably more for trolling or protecting potential developments that are more specific. Intel has been onto that way before them, I'd be curious to see if they can get this patent to pass through the first examiner's round.