Confidence in Bitcoin has never been demonstrated in such a blazing way. A private company announced they designed a full custom 65nm ASIC for mining bitcoins. They claim that consumer devices based on the chip will start shipping to customers this December.
The company, Butterfly Labs, had announced in June 2012, in a very brief press release, that they accepted venture capital from unnamed investors for designing a chip optimized for mining bitcoins. Given that we now know that they manufactured them at 65nm, and that the NRE costs for this process node are roughly half a million USD, the size of the VC investment must have been of the order of at least one million USD.
The only details publicly released at that time were estimated performance numbers for 3 devices that BFL planned to build based on this chip. The Bitcoin community received that announcement with skepticism. Nonetheless, a couple weeks later, when BFL started taking pre-orders, they received a quarter million USD of pre-orders in the first 24 hours.
Since then, not only BFL revised their performance numbers upwards in September (eg. the "Single SC" device was revised from 40 GHash/second to 60 Ghash/second), but they also announced a very impressive power efficiency: 1000 Mhash per Joule. They also released pictures of prototypes.
The performance of these ASICs is far above anything that the current FPGA-based Bitcoin mining devices achieve (10 to 20 Mhash per Joule), including BFL's own FPGA devices. (FPGA, or field-programmable gate array, is a type of technology typically found in high-performance computing devices or clusters, eg. in financial or medical fields.) In fact, because BFL expects the ASIC-based devices to completely obsolete the FPGA-based ones, they have offered to buy back old BFL devices from their customers.
BFL has been growing very quickly in the past year, and not without pain. The order backlog for their current-generation FPGA-based devices peaked at about 3 months in mid-2012, and has never been brought down to acceptable levels, not even until a week ago when BFL stopped selling FPGA-based devices, to focus on their ASIC line. The community estimates that BFL received about 2-4 million USD of ASIC pre-orders so far. So the backlog situation is likely to get worse, even though BFL is working very hard to resolve these startup growing pains, as one of their representatives explained at the London Bitcoin Conference.
As a Bitcoin miner myself since 2010, I am shocked that the mining competition went so quickly through the 4 classical phases of computing hardware evolution:
- 2009: CPU mining on general-purpose computers.
- 2010: GPU mining on graphics cards-equipped computers.
- 2011: FPGA mining on somewhat specialized devices.
- 2012/2013: ASIC mining on completely custom silicon chips!
I totally expected ASICs to arrive at some point (mining is an embarrasingly parallel process based on the SHA-256 hashing algorithm, perfectly suited to ASICs). I just did not expect a venture capitalist to understand the potential of a distributed cryptocurrency like Bitcoin so soon. Either he is a non-technical person who blindly trusted the founders. Or he is a very technical crypto/network person who grasped Bitcoin's concepts instantly. I see no other possibilities. Bitcoin is notably hard to explain. I see many intelligent but non-technical persons (ie. a typical VC) who completely fail to understand Bitcoin. Anyway.
This is a moment of significance for Bitcoin. We are witnessing the birth of a specialized Bitcoin mining industry.