ASIC
/ˈeɪ.sɪk/ · Noun · Development · Origin: 1981
Definitions
An Application-Specific Integrated Circuit, a chip designed and fabricated for one particular task rather than for general-purpose computation. Because the silicon implements the target function directly, an ASIC can be one or two orders of magnitude better in performance per watt than a CPU or GPU running the same workload. The tradeoff is total inflexibility once fabricated, together with enormous non-recurring engineering costs and design cycles measured in years, so ASICs only make economic sense at high volume or where the workload is stable and valuable. Bitcoin mining moved decisively to ASICs, rendering CPU and GPU mining obsolete; Google's TPU is an ASIC for tensor operations. FPGAs occupy the middle ground, being reconfigurable after manufacture at the cost of speed and efficiency.
In plain English: A computer chip built to do one specific job really well. Unlike a general-purpose processor, it can't be reprogrammed, but it's faster and more power-efficient at its designated task.
Example: "Bitcoin mining moved from GPUs to ASICs, and overnight your gaming rig couldn't compete with purpose-built hardware."
Origin Story
The custom chip that does one thing but does it insanely fast
**Application-Specific Integrated Circuits** emerged in the early 1980s as an alternative to general-purpose processors. Instead of a CPU that can run any program, an ASIC is designed for a single task -- video decoding, network packet routing, or a specific algorithm.
The ASIC design process is expensive (millions of dollars for mask sets) and slow (months of fabrication). But the payoff is dramatic: an ASIC can be 100-1000x more efficient than a general-purpose processor at its specific task. The tradeoff is flexibility -- if the spec changes, the chip is useless.
Bitcoin mining drove a dramatic ASIC arms race starting in 2013. Mining went from CPUs to GPUs to FPGAs to ASICs in just four years. Today, Bitcoin ASICs from Bitmain and MicroBT are so specialized they can do literally nothing except compute SHA-256 hashes.
Coined by: Semiconductor industry
Context: Early 1980s, VLSI design era
Fun fact: Google's TPU (Tensor Processing Unit) is technically an ASIC designed specifically for neural network inference. The original TPU was deployed in 2015 and was 15-30x faster than contemporary GPUs for ML inference, proving that custom silicon still beats general-purpose hardware.