Anyone built custom chips for Bitcoin mining?

Hey guys, I'm wondering if anyone has actually made their own integrated circuits for mining Bitcoin. You can find hardware designs for SHA256 pretty easily, but have any of you gone ahead and built your own mining chips using those designs? I'd really like to hear from anyone who has experience with creating a mining ASIC.

You need few million bucks for competetive asic development. It is no more an easy task for small operators like it was in the early days (Avalon1, Antminer S1......). Builders sidehack (Gekkoscience) and jstefanop (Futurebit) have been using Bitmain and Bitfury asic chips for their btc miners.

The only reasonably 'low cost' way to make them to use would be the 28nm node or larger. Even then you are looking at at least several k$ to run the chips at the handful of experimental/small run fabs that would even talk to you about producing a few dozen chips that by today's standards of speed and power eff would be near useless.

Well, I know a way to experiment with custom ICs without millions. Either way, I'd presume that the people who know how to write the HDL, test the chip and get it fabricated to functional CMOS devices won't hand out their secrets because the information is obscure while the profit motive is too high. I'm interested in doing it as an academic exercise but I wonder how people make money doing it currently.

For 16nm and larger node sizes, sure. The lithography equipment needed is very mature and along with a fairly large number of Foundries doing non-cutting edge node sizes, very available and so the cost for masks is fairly low. Go down to 10nm, things are a bit more pricey but still basically use very similar litho tech that 16nm use. Fact: For (mainland) Chinese-owned Foundries 12-10nm is as low as they go because they cannot buy the needed EUV litho equipment & photo resists much less produce it themselves. 5nm is a whole different animal that requires using EUV lithography. EUV litho gear is extraordinarily expensive and certainly not widespread -- you can literally count on 1 hand the number of Foundries using it. Chips using that node are high-value and high demand. There is no sneaking in a few test circuits per wafer because masks are just too freaking expensive and foundry customers simply will not permit 'other chips' to be produced on their production wafers.